AMD Preparing To Give Intel a Run For Its Money

jfruh writes: "AMD has never been able to match Intel for profits or scale, but a decade ago it was in front on innovation — the first to 1GHz, the first to 64-bit, the first to dual core. A lack of capital has kept the company barely holding on with cheap mid-range chips since; but now AMD is flush with cash from its profitable business with gaming consoles, and is preparing an ambitious new architecture for 2016, one that's distinct from the x86/ARM hybrid already announced."

Comments

Only the great Master of Paper can save AMD

[VTBlue]

With Papermaster running the show, I expect nothing short of excellence...on paper.

Re:Only the great Master of Paper can save AMD

[binarylarry]

I stick to Intel because they're the best CPU you can buy right now.

But I'd love to see AMD back in the game. I bought the first X2 Athlon series, what a beast that was.

Sadly that was also the last AMD CPU I've purchased.

Re:Only the great Master of Paper can save AMD

[drinkypoo]

I bought the first X2 Athlon series, what a beast that was.

Sadly that was also the last AMD CPU I've purchased.

The Phenom II X3 was also an absolute monster for the price, as was the Phenom II X6.

Re:Only the great Master of Paper can save AMD

[ganjadude]

I was a strong AMD guy from the K6 days up until the phenom days. up to that point I found better bang for my buck with AMD Its only with the haswell processors* that I have made the switch back to intel

* just happened to be the generation I was on when i was ready to upgrade my 1st gen phenom

Re: Only the great Master of Paper can save AMD

[Anonymous Coward]

Got a Phenom II x4 as it was the best bang for the watts and I was building an always on multi-purpose rig.

Re:Only the great Master of Paper can save AMD

[haruchai]

AMD made it easy to upgrade incrementally; not sure if the same would have been true of Intel as I've not had an Intel desktop in over 10 yrs.

Bought a Athlon X2 with nForce-based mainboard & DDR2 RAM in 2006.
Maxed out the RAM & upgraded to Athlon II X4 in 2009 while keeping same mainboard.
In 2011, bought new 990FX- based board to get SATA 3 / USB 3 & DDR3 RAM but kept the same 4-core CPU.
Just last week, got a 8320 Black Edition 8-core at a good price and might soon get my first "AMD" videocard as an upgrade for my GeForce 9600GT.

Re:Only the great Master of Paper can save AMD

[Nemyst]

AMD arguably still is pretty competitive... for the price. Having to append those words to anything about AMD is what they hate, though. They can compete in the low-mid end by undercutting Intel (and even then their CPUs are usually more power hungry), but as soon as you get closer to the high end AMD's just left in the dust except for those rather rare use cases which don't need much FPU performance but can run in massively parallel systems, where their 8-core CPUs shine.

Re:Only the great Master of Paper can save AMD

[drinkypoo]

AMD made it easy to upgrade incrementally; not sure if the same would have been true of Intel as I've not had an Intel desktop in over 10 yrs.

No, intel changed sockets more than AMD did in that period. I got an AM3+ board, so I went from Phenom II X3 720 to Phenom II X6 1045T, which I still have. If you're not expecting massive single-thread performance, it is still a fairly beefy CPU. I mean, sure, half as much as an intel chip, but I paid a hundred bucks for this (and for my original CPU) and you'd have to spend $200 to get an intel chip with this much horsepower today. AMD-chipset motherboards are cheaper than intel-chipset motherboards as well, so the total savings was at least $200 if not more. Today, I still have more than enough CPU for anything I want to do; It's the 240GT that's holding me back now. Been thinking about a modest upgrade to a newer nvidia card pulled from a Dell on ebay for $60.

Re:Only the great Master of Paper can save AMD

[Anonymous Coward]

I'd advise against unless it's a very new card or you're not intending to game; Nvidia's just dropped support for all pre-fermi cards and will be dropping support for Fermi (400 and 500 series cards) with their next hardware revision.

That said, given you said you're coming from a 240GT, I guess gaming isn't really a concern for you anyway.

Re:Only the great Master of Paper can save AMD

[Anonymous Coward]

Yeah, its not like they intentionally gimp their processors with shoddy construction methods, LIKE PUTTING THERMAL PASTE DIRECTLY ON THE PROCESSOR. I know I'll never own a processor that requires delidding. Thankfully AMD processors are still made correctly.

Re:Only the great Master of Paper can save AMD

[drinkypoo]

I'd advise against unless it's a very new card or you're not intending to game; Nvidia's just dropped support for all pre-fermi cards and will be dropping support for Fermi (400 and 500 series cards) with their next hardware revision.

That's OK, I'm looking at a 640GT. I do game, but I mostly play older games and I don't demand full detail at 1920x1200, my display res.

The 240GT is enough to play many older games at full res with full options, but of course, no newer games. And it won't even play most newer games with even half the options turned on. Even Star Trek Online taxes my 240GT.

I chose the card on the basis of power consumption. The 250GT was the hot card when I bought it (I've had it long enough to replace the GPU fan!) and the 240GT had 3/4 the performance with 1/2 the TDP. I have a 480W power supply, no problems. The 640GT is another one of these no-SLI, no-additional-power-connector cards, so it appeals to me. I can get one for fifty bucks.

Re:Only the great Master of Paper can save AMD

[Lonewolf666]

Possible but not attractive if the time frames between upgrades was a bit longer.

Bought an Athlon X2 with DDR2 RAM in 2007.

Wanted an upgrade in 2011 and found that the price per GByte of DDR2 RAM was much higher than for DDR3 RAM. CPU and GPU needed changing anyway. 4GByte of DDR3 RAM were not more expensive than buying another 2GByte of DDR2 RAM would have cost. Some Athlons and Phenoms for socket AM2+ were still available but the socket AM3 CPUs looked considerably better.

So I settled on a Phenom II X4 on a relatively inexpensive new socket AM3 Board. Bottom line, I paid some extra for the new board but got better memory bandwidth and a more power efficient CPU out of it. The old board remained operational, as I did not rip out CPU, GPU or RAM.

I also still had a nice case from 2004 hanging around so I shelled out a bit additional money for a PSU and a new hard disk, and put the new system into the old case. Overall, I paid maybe 150 euros more than with maximum reuse of the old stuff. But that way I kept the 2007 PC usable, which is still useful from time to time :-)

Re:Only the great Master of Paper can save AMD

[neokushan]

I do wonder what the future has in store for the humble CPU. With a huge market shift towards tablets and phones in the consumer area, where power savings are more important than raw oomph, as well as a similar shift in a good portion of the server market, are we starting to reach an era of CPU's being "good enough" for most people and performance to begin stagnating?

Hopefully some good competition between AMD and Intel will keep things fresh and fast.

Re:Only the great Master of Paper can save AMD

[Anonymous Coward]

I remember back in the day when nerds loved AMD for the Athlon XP for being the smarter, nerdier choice. Then when the Athlon 64 and its dual core version were dominating everything Intel could throw at it, Nerds still bought it because it was the best CPU even though AMD was charging ridiculous prices. But then.... suddenly, Intel comes out with the Core 2 Duo which was an unbelievably monumental leap in performance... and on top of that they charge reasonable prices for them. Suddenly you can get a Intel CPU for $300 that offers 6x the performance of an AMD CPU that was selling for $1200 just a week ago. That turned a lot of nerds against AMD, if they hadn't been busy ripping people off when they were on top they'd have held onto our heartstrings and we'd have kept buying.

Re:Only the great Master of Paper can save AMD

[ne0n]

The best thing AMD got right was the upgrade path, S939 kept my machine current for years before the glorious AM3 socket. Every Intel we've built OTOH only has one good CPU per socket and no upgrade path after that.

Re:Only the great Master of Paper can save AMD

[evilviper]

I stick to Intel because they're the best CPU you can buy right now.

I'm sure your method of choosing a car is, no doubt, similarly insightful...

Re:Only the great Master of Paper can save AMD

[OneAhead]

This written on an Athlon 64 X2 "Toledo". It runs XFCE 4.8 on a 64-bit Linux kernel 3.5 pretty snappily (my experience is not too much worse than on my Ivy Brigde) and keeps up admirably with the (modest) workload I'm throwing at it. It's a day-and-night difference with the 1-year-newer P4 Cedar Mill I threw out because it was nearly unworkable in the exact same function, even though it was paired with faster memory. I'm thinking the early EM64T implementations were not up to snuff, which wasn't such a big deal at the time because no-one was using it much.

I guess I'm giving everyone a blast from the past with this post :)

Re:Only the great Master of Paper can save AMD

[OneAhead]

Core2 Duo 6x faster than Athlon 64 X2? What are you smoking? Yes, it was somewhat faster, and AMD was a bit more sluggish adjusting its prices than it should have been, but the Core 2 Duo's big success was largely due to its much lower power draw in laptops, and more importantly, the market's "Intel inside" bias. The Core2 later got significant competition from the Phenom xx50 and newer, but that was only half a year before the Core i7 came out. Only then, you could really say the Intel CPUs ran circles around their AMD counterparts, at least in desktop and mobile applications.

Re:Only the great Master of Paper can save AMD

[Anonymous Coward]

The Phenom x3 was alright for the price, especially if you could unlock the 4th core. But the Phenom X6 is where AMD started sucking. Lots of slow cores sitting around doing nothing but sucking away power and creating heat. That's where they went wrong. Now they sell 8 core CPUs that get slaughtered in most everyday tasks by a $45 dual core Celeron G1840 which uses 1/4 of the power or less.

Re:Only the great Master of Paper can save AMD

[Anonymous Coward]

Agreed, Phenom II's were amazing - most of my home servers still run them, 1/4 the price of anything Intel had at the time around the same speed.

Honestly, if they were still developing Phenom II's w/ die improvements / power reduction - I'd still be buying them (my latest round of build servers will probably end up being AM1s instead).

Re:Only the great Master of Paper can save AMD

[ooshna]

I love my Phenom X3 720. I got lucky and got one that was a 4 core with the forth core disabled. I have it unlocked through the bios and am run at 3.4 on air. Not bad for under $200 a 4 years ago.

Re:Only the great Master of Paper can save AMD

[Anonymous Coward]

Star Trek Online taxes high end GPUs with you have all details maxed and you're in a warzone, such as the Undine space battlezone. Even a dual GTX 780 can crawl to only a few FPS.

Re: Only the great Master of Paper can save AMD

[JamesHassinger]

During that period? I dropped intel because every time I wanted a new CPU I needed a new motherboard. Didn't use to be that way in the pentium II and pentium III days, but sometime mid 2000s intel changed. Sorry intel I don't wanna buy a $200+ CPU AND a $200 motherboard everytime thanks. My 6 core 1055t AMD uses a 4 year old motherboard and I'm thinking about getting a new motherboard for usb 3.0 and ten USB ports and 4x PCI-E x16 that can use the same CPU. It's nice not having to change both CPU and motherboard at the same time when you're happy with what you have

Re: Only the great Master of Paper can save AMD

[Anonymous Coward]

With Intel CPUs, I have to upgrade less often than with AMD. The cost evens out and I end up with top of the line gear instead of settling for AMD's second best.

Re:Only the great Master of Paper can save AMD

[Anonymous Coward]

Even AMD's 8 core CPUs get left in the dust by Intel's 4 core CPUs. They haven't been competitive since the Athlon vs Pentium III days.

Re:Only the great Master of Paper can save AMD

[Anonymous Coward]

I wouldn't say 6x faster, but easily 2x to 3x at the same clockspeed.

Re:Only the great Master of Paper can save AMD

[Anonymous Coward]

Although it's three times the price, might I recommend the 750Ti? (Maxwell architecture)

Sub-60W TDP (no additional power connectors) and it's fast enough to run almost all modern games at medium/high (It's faster than my 560Ti, which can draw 220W at load with the overclock needed to stay relevant). It's on a more recent CUDA revision as well which means it should be useful if you're doing video encoding or grid computing of any kind.

Re:Only the great Master of Paper can save AMD

[drinkypoo]

Although it's three times the price, might I recommend the 750Ti? (Maxwell architecture)

Nope. Sounds nice, but $100 is my absolute upper limit for a computer component these days. I don't think I've ever spent more than about $300 for any one part, but I try to keep it a little cheaper these days. And I don't really care about CUDA either. My whole six-core PC has about $600 in parts in it. If I buy this video card that will actually decrease, and I'll be able to kick the 240GT down to a machine that has a fairly pathetic video card in it now and my old Phenom II X3 720 which gets up to 3.2 GHz on air with a $20 thermaltake heat pipe/big fan cooler. Though, I think that's in my system now, so I might have to drop another twenty bucks on another cooler. The OEM cooler was fairly pathetic.

Like I said, I'm not really picky about performance. I just want adequate to play games with, say, medium settings but full resolution. Since I mostly play older games, I'll get fantastic performance in most of what I play.

Re:Only the great Master of Paper can save AMD

[sound vision]

I've been running with a board from the same era (ca. 2007) straight through today. It's an AM2 board, so you can't put a *new* CPU in it anymore, but their backward-compatible AM2+ chips extended the life significantly. I haven't gamed much in the past year, but as of 2012, I hadn't come across a game it wouldn't play. That's a lifespan of 5 solid years, which is great for a gaming rig, and (coincidentally?) about the length of a console cycle.

The only thing that could get me to upgrade at this point would be if some new game comes out that I *have* to play, and the game can't run with a less-than-stunning CPU. However, given that most games today are developed to console specs, I have yet to come across such a game.

Re: Only the great Master of Paper can save AMD

[Anonymous Coward]

Same here. Sadly I stuck around longer until my first Intel build, Q6600.

Re:Only the great Master of Paper can save AMD

[OneAhead]

No way; the Core2 was essentially a thoroughly reworked Pentium III. Moore's law has stalled terribly when it comes to single-thread performance; what we did get is more cores. Even more so when looking clock-for-clock; in that department, both Intel and AMD only had incremental increases of 5~20% since the Athlon 64, with the big exception of the Nehalem architecture, which was the only truly big jump in single-core per-clock performance of the x86-64 era.

Just like Bulldozer?

[jandrese]

Put up or shut up AMD. I don't care how awesome your chips will be in 2 years, I care how good they are now, and right now they aren't so good. 2 years is a long time in the CPU world, I hope your gains aren't completely eaten away before you launch the chip, again.

This is really a low quality article. "In the distant future we plan to release faster chips!" What a scoop!

Re:Just like Bulldozer?

[thaylin]

The story is more about the come back of their great designer who has done a lot, and how they are betting on him.

Re:Just like Bulldozer?

[Anonymous Coward]

All the article is saying is that now they have the capital for R&D so they expect to be able to push new territory that they otherwise wouldn't have. Intel will just match this new capital expenditure level, and everyone will be happy.

Re:Just like Bulldozer?

[Anonymous Coward]

All the article is saying is that now they have the capital for R&D so they expect to be able to push new territory that they otherwise wouldn't have.

Citation needed. Last I looked, Intel's R&D budget was larger than AMD's revenue.

Re:Just like Bulldozer?

[Anonymous Coward]

Yup, and the BS about them being first to 64-bit...maybe in the consumer sector, but Intel, IBM and DEC all had 64-bit chips before the Athlon was even designed let alone shipped.

Re:Just like Bulldozer?

[TheRaven64]

Last I looked, Intel's R&D budget was larger than AMD's revenue

That certainly was true (probably still is), but it's misleading. AMD no longer owns fabs and the majority of Intel's R&D spending is on process technology. By spinning off GlobalFoundaries, AMD is able to share that R&D cost with other SoC makers and go to other companies if they happen to be able to do it better at a specific time.

Re:Just like Bulldozer?

[Anonymous Coward]

Where are the same demands for realism when it's a 3D printing story? There it's always "in 50 years we'll 3D print houses on Mars!" and everyone sucks each other's dicks!

Re:Just like Bulldozer?

[Rich0]

Yup, and the BS about them being first to 64-bit...maybe in the consumer sector, but Intel, IBM and DEC all had 64-bit chips before the Athlon was even designed let alone shipped.

They invented the architecture that you probably typed your post on. That was the point. Heck, on my linux distro it is still called amd64...

Re:Just like Bulldozer?

[Anonymous Coward]

Good point. Would be interesting to know how their CPU R&D budgets compare, though AMD will still struggle if Intel continues to be at least one generation ahead of the fabs AMD are using.

Re:Just like Bulldozer?

[rogoshen1]

how is spinning off your fabrication capability 'good' in the long run? (not trying to be flippant, it's a serious question)

Re:Just like Bulldozer?

[bhcompy]

Last year I had an R&D budget of $1. This year I have an R&D budget of $100. It doesn't matter what my competitor has, it matters that I can now do a lot more research and development.

Re:Just like Bulldozer?

[bhcompy]

Yes, and Itanium is dead. Alpha is gone. And I don't even know what IBM did. Meanwhile, x86-64 is here to stay.

Re:Just like Bulldozer?

[Anonymous Coward]

But Intel has been spending on research. AMD gave up a while ago, and settled for making money on mobile devices.

Re:Just like Bulldozer?

[jandrese]

Because I want the news from last week of course.

Re:Just like Bulldozer?

[Anonymous Coward]

Oh! The mighty Linux says? Wow. This must be fact!

Re:Just like Bulldozer?

[AdamHaun]

how is spinning off your fabrication capability 'good' in the long run?

I don't work at AMD, but I do work at another company that relies partly on foundries.

Basically, it's economies of sale and competition. Semiconductor fabrication processes keep getting more expensive. Foundries specialize in process development and spread the R&D across many, many customers. Unless you're willing to spend a fortune keeping up (as Intel is), have special requirements, or need a ton of volume, you have little to gain and a lot to lose from rolling your own process. Remember, you don't just have to make transistors, you also have to have good enough yield to turn a profit and good enough reliability to keep your customers. If you fail, you have to spend even more money to fix the fab on top of the money you're losing on the stuff you manufacture. Meanwhile, TSMC is cheerfully cranking out wafers for your competitors.

Re:Just like Bulldozer?

[haruchai]

That's a good question. After all, Intel was able to use their vastly superior fab capabilities to fend off AMD's enhanced tech for years until they released their Nehalem architecture to definitively take back the desktop CPU performance crown.

They're not called ChipZilla for nothing.

Re:Just like Bulldozer?

[Anonymous Coward]

It isn't, but they needed survive and the fabs were draining too much cash.

Re:Just like Bulldozer?

[kasperd]

Yup, and the BS about them being first to 64-bit...maybe in the consumer sector, but Intel, IBM and DEC all had 64-bit chips before the Athlon was even designed let alone shipped.

That is true. However AMD were the first to make a 64-bit architecture, which was x86 compatible. And it was also the first 64-bit CPU to be in a price range that was acceptable to average consumer. But most importantly, AMD designed an architecture so successful that Intel decided to make their own AMD compatible CPU. Today Intel probably earns most of its money on CPUs using AMD's 64 bit design.

But if AMD now want to go and build an entirely new design, which is nothing like x86, they may very well be repeating the exact same mistake Intel made to let AMD64 get the lead.

By now it might be safe to ditch all 8, 16, and 32 bit backwards compatibility with the x86 family. But AMD64 compatibility is too important to ignore.

Re:Just like Bulldozer?

[Anonymous Coward]

Not the architecture, that belongs to Intel, AMD extended it to support 64 bits. Intel actually had 64 bit extension in the closet, they didn't want to cannibalize the Itanium. When the AMD extensions became mainstream, the defacto standard, Intel licensed it. But the extension is still tied to the Intel architecture and Intel is still in control.

Re:Just like Bulldozer?

[drinkypoo]

Intel does that by spending massive scads of money on process technology. They are able to spend that because they massively overcharge for their product if you take only the design, manufacturing and distribution costs into account. You also have to count paying for their processes.

AMD let a bunch of designers go, it's not clear if intel would have eaten their lunch so aggressively otherwise. Guess we'll have some inkling here soon.

Re:Just like Bulldozer?

[drinkypoo]

Not the architecture, that belongs to Intel, AMD extended it to support 64 bits.

What are you on about? amd64 is not an architecture, nor is x86. They are instruction sets. The underlying architecture may be informed by the instruction set, but it's also only loosely coupled in modern CPUs.

Re:Just like Bulldozer?

[bobbied]

Yes, and Itanium is dead. Alpha is gone. And I don't even know what IBM did. Meanwhile, x86-64 is here to stay.

And we are all better off? Um, NOPE!

The ONLY thing that kept the X86-64 going was that it ran X86 instructions. As a processor the X86 family is really horrid both from the electrical and programming sides. Lucky for the motherboard makers most of the sticky "design" is abstracted away by the chip-sets, but for the programmer, the architecture was not designed to be expandable, certainly didn't port well into 64 bits.

If you ask me, Alpha and other processors where of better design. My personal favorite was the Motorola 68000 (aka power PC) stuff. That was an engineer's dream to work with both from the hardware and software sides of things. Certainly the instruction set and programing model was a whole lot better than what we generally see today...

So why did the X86 win out? You can think both IBM and Bill Gates for that. (I suppose you could blame Motorola for not accepting the license terms too.) It is the processor in the PC, which have been sold by the millions, making production costs lower.. Apple Mac stuff was initially Motorola CPU's but finally even Apple had to ditch them for cost. X86 is in use for business and historical reasons, not because it was the right choice technically.

Re:Just like Bulldozer?

[Anonymous Coward]

Who was the first search engine - no one cares and most don't know. Being first is nice, staying first keep you in business.

Re:Just like Bulldozer?

[Anonymous Coward]

Motorola 68000 and PowerPC have nothing whatever to do with each other except that Apple used both in its products and Motorola was involved in the design of both processors.

Re:Just like Bulldozer?

[timeOday]

2 years is a long time in the CPU world

Well, not so long as it used to be. I recently got a Macbook Pro and under "About This Mac / Processor" it says "2.3 GHz Intel Core i7" - the same thing it says on a Macbook Pro I got 3 years ago. The CPU is not actually identical of course - it has much-improved battery life, which is good. But the performance increase, if any, is not noticeable. Times really have changed.

Re:Just like Bulldozer?

[oh_my_080980980]

I think the point was even with Intel's massive cash and infrastructure they couldn't bring 64 bit to the desktop - hell they couldn't do it on the server end either; thet Itanium chips were huge flops. And what killed Itanium was AMD's chip!

" Itanium failed to make significant inroads against IA-32 or RISC, and then suffered from the successful introduction of x86-64 based systems into the high-end server market, systems which were more compatible with the older x86 applications." http://en.wikipedia.org/wiki/I...

So the point is that AMD was more than capable of producing a chip to beat Intel.

Re:Just like Bulldozer?

[oh_my_080980980]

You mean like Windows phones.

Re:Just like Bulldozer?

[Anonymous Coward]

My personal favorite was the Motorola 68000 (aka power PC) stuff.

Just a minor correction. 68000 has very little in common with the PowerPC except that they were both great architectures to program for. PowerPC is RISC and was originally spec'ed out for a 64 bit architecture. Take any FreeScale CPU manual (the spin-off company from Motorola) and you will see that the WORD datatype is 32-bits long (hence DWORD is 64 bits). On top of that PowerPC was the product of a consortium of companies AIM (Apple IBM Motorola). 68K on the other hand was a 16/32 bit CISC CPU and was designed and manufactured by Motorola only.

You can think both IBM and Bill Gates for that. (I suppose you could blame Motorola for not accepting the license terms too.)

As for thanking Mr. Bill G, well you don't need to. A 32-bit CPU also required an expensive chipset. So, m68k was definitely the choice of high-end HW of its day (e.g. Unix servers and workstations) and would not have been a good choice for the PC platform just due to the cost of the chipset and the peripherals.

Re:Just like Bulldozer?

[aliquis]

Yeah. This year you can get the PSU. .. :)

Re:Just like Bulldozer?

[Guy Harris]

Not the architecture, that belongs to Intel, AMD extended it to support 64 bits.

What are you on about? amd64 is not an architecture, nor is x86. They are instruction sets. The underlying architecture may be informed by the instruction set, but it's also only loosely coupled in modern CPUs.

This is why the term "microarchitecture" is useful. In addition to the term "instruction set", the term "instruction set architecture" is also used, so "architecture" is used for both.

Re:Just like Bulldozer?

[__1200333]

Instruction Set Architecture (ISA) is also a common name for it. That and tons of linux distributions and people refer to it as "arch" or "architecture". GNU coreutils even has a program named arch that prints out the "machine hardware name" (ew...).

The fact is lots of people call it architecture and it can reasonably be called architecture in the given context. I'll allow it!

Not denying it's important to know the difference, though!

Re:Just like Bulldozer?

[aliquis]

IBM did Power and PowerPC (IBM PowerPC 970 is the "G5" in Apple products.)

Re:Just like Bulldozer?

[Grishnakh]

I think the point was even with Intel's massive cash and infrastructure they couldn't bring 64 bit to the desktop

Wrong. They could have if they had wanted to, but they didn't want to. They wanted 64-bit to remain in the realm of big-iron, so they could sell their big, overpriced Itanic chips. Whenever anyone asked about 64-bit chips, Intel said "buy our Itanic!". When anyone complained about the 4G memory limitation inherent with 32-bit chips, they pointed to their crappy PAE extension.

Then AMD came out with the X86-64 ISA, and then suddenly Intel looked stupid. They tried to say things like "people don't need 64 bits on desktop systems", "you can use PAE to use more than 4G", "no one needs more than 4G", until they trotted out their hastily-made "EM64T" version.

Re:Just like Bulldozer?

[aliquis]

What I've read is that the AMD64 stuff is actually good. Personally I don't have an opinion since I'm not educated and experienced enough to have one really.

Motorola 68000-series isn't PowerPC. Their PowerPC chips are called PowerPC 74x, 75x, 74xx and 86xx. ... and I assume you're wrong there before a friend of mine back in the days was a demo coder on the Amiga 500 and later got a i386 and afaik he thought it was better (why else would he had switched?) Then again he never experienced the M68040 or M68060 (or the 020 or 030 I guess) and there was improvements like MMU, floating point units, I guess possibly more registers in those.

Apple used both M68k, PowerPC both from Motorola and IBM and Intel x86 and whatever Intel calls AMD64. I guess Apple business was the major player of higher end PowerPC chips and as such they had to fund a large part of the research and manufacturing whereas they didn't had to do that with Intel chips. The 32-bit chips was pretty lame but I guess they knew Intel had 64-bit chips coming soon and that they would be very competitive vs what they could get from IBM.

Re:Just like Bulldozer?

[aliquis]

s/before/because/g :)

Re:Just like Bulldozer?

[Adriax]

Helps to bribe system builders to keep AMD out of most consumer's machines.
http://hardware.slashdot.org/s...

Re:Just like Bulldozer?

[drinkypoo]

Yeah, but ISAs are all but over, and they are over in x86-land. We haven't had an x86 core defined by its instruction set since the 80486. Even the Am586 was internally RISCy.

Re:Just like Bulldozer?

[vakuona]

I would also add that unless AMD plans to have a business where it can migrate cheaper chip business to use the older fabs, AMD might well find itself having to manage fabs i.e. run a foundries business to recoup its investment. AMD doesn't really make the cheap as chips chips (the kind of stuff that Broadcom makes), then they should not be in the business of fabs. They ought to let the likes of TSMC who can manage that migration much better than AMD be in that business.

Re:Just like Bulldozer?

[aliquis]

When you say a 32-bit CPU what do you mean really? The Intel one or the Motorola one?

From what you're saying afterwards it sound like the m68k was ok on more expensive stuff but not on PC because it was cheaper?

But lots of designed for home gear like the Mega Drive, Atari and Amiga home computers and so on used the m68k.

Re:Just like Bulldozer?

[Guy Harris]

Yeah, but ISAs are all but over, and they are over in x86-land.

They're not "over" to compiler writers and assembler-language programmers.

We haven't had an x86 core defined by its instruction set since the 80486. Even the Am586 was internally RISCy.

So? They (and the latest z/Architecture chips) might translate native instructions into microops and schedule and execute those microops, but the only way in which those microops - or other implementation details of the processor - are visible to code and people or software that generate code is that they may affect the performance of particular sequences of instructions, so that, for example, a compiler might optimize differently for different processors.

Re:Just like Bulldozer?

[Lonewolf666]

Well I guess Intel have the volume too, so that is another thing going for them. AMD with its lower volume may be (more strongly) forced to go with a foundry.

At the same time, progress in fabrication processes seems to slow down a bit, and cost advantages are no longer so obvious with a new generation. Maybe the gap between Intel and AMD in manufacturing will shrink due to that.

Re:Just like Bulldozer?

[shizzle]

What are you on about? amd64 is not an architecture, nor is x86. They are instruction sets. The underlying architecture may be informed by the instruction set, but it's also only loosely coupled in modern CPUs.

AMD64 and x86 most certainly are architectures. Have you heard the term "instruction set architecture", i.e., ISA? The underlying implementation you refer to is usually referred to as the "microarchitecture" to distinguish it from the ISA.

The term "architecture" is often tossed around to refer more broadly to the general organization of an implementation, but it's not wrong to use it in the more specific sense of ISA as well.

Re:Just like Bulldozer?

[Anonymous Coward]

EMT64 was in the labs for several years before it was released to market, just like "Jackson Technology" or "Hyper-Threading" was in the labs several years before it's introduction to market, so "hastily-made" is definably false.

As someone stated earlier, EMT64 was already cooking in the labs at the same time Itanium or IA64 was cooking, Intel gambled on IA64 and wanted to start to move away from IA32, but fortunately for the rest of the world, AMD forced their hands.

I know a lot of this first hand, has I am a former member of one of the Intel Validation teams

Re:Just like Bulldozer?

[drinkypoo]

Yeah, but ISAs are all but over, and they are over in x86-land.

They're not "over" to compiler writers and assembler-language programmers.

IS != ISA, HTH HAND

Re:Just like Bulldozer?

[khellendros1984]

IBM introduced the 64-bit version of the PowerPC architecture in 2002, and I think that some of their mainframe processors were 64-bit about 2 years before that.

Re:Just like Bulldozer?

[Anonymous Coward]

As a processor the X86 family is really horrid both from the electrical and programming sides.

Nowhere near as horrid as Itanium, at least on the programming side. Yeah, let's throw all the complexity into the compiler, not the CPU, that'll work.

Re:Just like Bulldozer?

[Guy Harris]

Yeah, but ISAs are all but over, and they are over in x86-land.

They're not "over" to compiler writers and assembler-language programmers.

IS != ISA, HTH HAND

It would only help if it were true. What are your definitions of "instruction set" and "instruction set architecture", and what citations can you give that would make those definitions worth taking seriously, as opposed to, for example, Intel's use of "instruction set" and Intel's use of "instruction set architecture"?

Re:Just like Bulldozer?

[drinkypoo]

AMD64 and x86 most certainly are architectures.

Nope. I will correct you one time and then I'm done with this stupid thread.

AMD64 and x86 are instruction sets. x86 isn't even an instruction set, it's a name for families of instruction sets. Once upon a time, instruction sets were related directly to architectures. In PC-land, that time ended with the 80486. All x86-compatible processors since have been some other kind of core internally, with decode and encode on the way in and out of the CPU to make it look like an x86 processor.

We no longer have instruction set architectures in mainstream computing, because we have moved up to a level of complexity as well as a level of feature churn where that makes no sense. Developers depend on instruction sets remaining fairly constant, but customer continue to demand (with their dollars) ever faster processors. The instruction set is now divorced from the processor internals.

That is what I mean when I say we no longer have instruction set architectures: we have micro-op architectures with translators strapped to them. Meanwhile, architectures shift around beneath them with most users (and programmers!) none the wiser. Architectures are now known not by the associated instruction set (which is now simply a line item in the specifications) but by code names like "Hammer", "Clawhammer", "Bulldozer", "Nehalem", etc etc. Today these architectures are designed to implement multiple instruction sets.

Re:Just like Bulldozer?

[bhcompy]

The best technical choice isn't always the best practical choice. x86-64 won out because it was practical and manageable from a cost perspective. Ran existing code natively while at the same time being an upgrade for the future. Technically speaking, it would be great to have the nation run completely on nuclear power, but there are costs to that, coupled with natural disaster risks, so we moved in other directions. It wasn't practical.

Re:Just like Bulldozer?

[Lonewolf666]

By now it might be safe to ditch all 8, 16, and 32 bit backwards compatibility with the x86 family. But AMD64 compatibility is too important to ignore.

Dropping 16 bit backwards compatibility is probably OK by now, and I don't think there is such a thing as 8 bit programs on x86 at all. But 32 bit software is still widely used and backwards compatibility to it is an important feature of AMD64. AMD would be crazy to drop that in an AMD64 compatible CPU.

At the same time however, they are developing ARM-based server processors which are not x86 compatible at all. So there seems to be a market for that. There certainly is in the tablet world. I just don't see it for the desktop yet.

Re:Just like Bulldozer?

[drinkypoo]

What are your definitions of "instruction set" and "instruction set architecture", and what citations can you give that would make those definitions worth taking seriously, as opposed to, for example, Intel's use of "instruction set" and Intel's use of "instruction set architecture"?

Instruction set, the set of instructions. ISA, the part of the architecture which handles the instruction set. Just the decode and encode stages of the CPU, and the instruction set itself. But originally, the ISA was defined by the very architecture of the processor, and it related directly to the architecture of the underlying processor.

Re:Just like Bulldozer?

[shizzle]

AMD64 and x86 most certainly are architectures.

Nope. I will correct you one time and then I'm done with this stupid thread.

Since drinkypoo is done with this thread, I won't bother to try and change his/her mind---but for the benefit of other readers, you may safely assume that drinkypoo is wrong. If you really care to learn more, there's plenty of information out there, say here or here.

Or if you really want to go back to the start of it all, the IBM 360:

The design made a clear distinction between architecture and implementation, allowing IBM to release a suite of compatible designs at different prices.

All the code names listed above are different implementations of the AMD64 architecture.

Re:Just like Bulldozer?

[InvalidError]

AMD may have shed the risk associated with owning fabs but process-wise, they have also slipped almost two process generations behind Intel - Intel is about to start shipping 14nm products while AMD is still on 28nm.

And Intel is also sharing some of their fab costs by leasing some of their capacity to other fab-less chip designers such as Altera.

I bet AMD never expected TSMC and GF to slip so far behind Intel when they spun their fabs off.

Re:Just like Bulldozer?

[Guy Harris]

Instruction set, the set of instructions. ISA, the part of the architecture which handles the instruction set.

Who, other than you, uses that definition of ISA? Intel doesn't, as per my previous post. IBM doesn't, either, and neither does ARM, nor does AMD, nor does Sun^WOracle, for example; they're all using "instruction set architecture" as "instruction set", with "architecture" perhaps given to signify that the instruction set is not just a characteristic of a particular processor, it's something that's specified separately from particular implementations of the instruction set.

Just the decode and encode stages of the CPU

So what does the "encode" stage of a CPU do? Take various internal chip signals and write out instructions to memory?

But originally, the ISA was defined by the very architecture of the processor, and it related directly to the architecture of the underlying processor.

If you mean "in the very early days of computers, the connection between the instruction set and the design of the CPU was straightforward", that might be true, but, dating at least back to the IBM System/360, the same instruction set was implemented by extremely different internal processor designs in many families of computers.

Re:Just like Bulldozer?

[shizzle]

Instruction set, the set of instructions. ISA, the part of the architecture which handles the instruction set..

That may be what you think it means, but that's not what it means. Instruction set and ISA are synonyms. In fact "architecture" by itself can mean "instruction set", but people also use "architecture" in other ways, so the term ISA came around to clarify that we're talking about "architecture" in the sense of "instruction set".

As in my other post, I'm not really trying to argue with drinkypoo, just hoping to keep others from being misled.

Re:Just like Bulldozer?

[Robear]

Intel and AMD followed, not led. MIPS, SPARC and ALPHA all preceded them to fully 64-bit chips. And the Sun MAJC chip was the first dual core chip on the market.

Re:Just like Bulldozer?

[Darinbob]

IBM did Power, but PowerPC was jointly designed between IBM, Motorola, and Apple.

Re:Just like Bulldozer?

[Guy Harris]

EMT64 was in the labs for several years before it was released to market

But was the project started before, or after, AMD published the first AMD64 spec? My guess is "after", unless Intel either 1) somehow managed to come up with the same extensions as AMD or 2) discarded their extensions and went with AMD's after AMD announced them.

Re:Just like Bulldozer?

[Darinbob]

Intel's problem here was that they tried to break the x86 chain by not being backwards compatible in their 64-bit solution. Even Intel knew that x86 wasn't a great architecture and they wanted something better. There were a set of problems with public acceptance of itanium though; their emulation of x86 wasn't very good, and for the high end systems that they targeted there was already a lot of competition, and the price was steep. AMD succeeded by being an extension of x86 instead of being a new design and by being cheaper. In that sense, Itanium was a better architecture all around and the x86-64 was just more of the same.

Think of it this way, when you've worked on code that's 10 years old and you think "this would be so much better if we could throw it away and start from scratch" imagine that Intel thinks the same way with x86 only it's dealing with a 40 year long chain of incremental improvements.

Re:Just like Bulldozer?

[aliquis]

Regardless they have made 64 bit PowerPC processors which was the point.

Re:Just like Bulldozer?

[Guy Harris]

IBM introduced the 64-bit version of the PowerPC architecture in 2002,

Well before that, and, yes, the PPC 620 was used in some systems shipped by Groupe Bull. IBM's first systems using 64-bit PPC processors came out in 1997.

and I think that some of their mainframe processors were 64-bit about 2 years before that.

2 years before 2002, yes.

Re:Just like Bulldozer?

[Anonymous Coward]

You have no idea what you're talking about.

Re:Just like Bulldozer?

[Anonymous Coward]

Frequency is not the same thing as performance. Your new 2.3 GHz i7 is faster in benchmarks and real-world use than your 3 year old 2.3 GHz i7. The performance gains have slowed down a bit, since Intel has been putting more focus on power (as you note, battery life is greatly improved), but performance gains are not completely gone by any stretch of the imagination.

You might not notice it, however, because a lot of people don't really have a use for all the performance modern CPUs have to offer.

Re:Just like Bulldozer?

[Anonymous Coward]

The ONLY thing that kept the X86-64 going was that it ran X86 instructions. As a processor the X86 family is really horrid both from the electrical and programming sides.

Like you know anything about either side.

Lucky for the motherboard makers most of the sticky "design" is abstracted away by the chip-sets, but for the programmer, the architecture was not designed to be expandable, certainly didn't port well into 64 bits.

You base this on... what, precisely? 64-bit x86 seems to work pretty well to me.

If you ask me, Alpha and other processors where of better design. My personal favorite was the Motorola 68000 (aka power PC) stuff. That was an engineer's dream to work with both from the hardware and software sides of things.

ROFL. 68000 and PowerPC are entirely different instruction sets, not the same thing in any sense at all. And anyone who tries to claim that x86 is ugly "electrically" while 68000 is beautiful clearly has no clue. Despite all the elegance which appealed to assembly language programmers, it turned out that 680x0 (particularly post-68020) is actually much less friendly to high performance, pipelined hardware implementations than x86. Ironically, the super powerful addressing modes and high orthogonality of the ISA made hardware engineers' lives significantly more difficult. This led to Motorola's 680x0 designs falling progressively further behind Intel's x86 designs, both in performance and time-to-market, and thence to Apple deciding to shitcan 680x0 and switch to PowerPC, a completely different architecture derived from IBM's POWER design.

So why did the X86 win out? You can think both IBM and Bill Gates for that. (I suppose you could blame Motorola for not accepting the license terms too.)

Motorola lost the IBM PC deal because the 68000 literally was not ready, and was too expensive for IBM's system price targets. Not because Motorola balked at licensing terms (for what?). The 68000 ended up losing in the market because, although it was clearly superior in the early 1980s and was successful in many markets, particularly high-priced UNIX workstations, by the late 1980s it had lost the high end markets to even more clearly superior RISC CPUs. And then it also began falling behind x86 performance, leading even Apple to abandon it.

Re:Just like Bulldozer?

[Anonymous Coward]

For clarity, I am not the AC who was on an Intel validation team.

It's an open industry secret that Intel's 64-bit x86 project began before AMD first announced or published AMD64. It was not an identical extension by any means. When Intel realized that AMD64 was going to force them to release 64-bit x86 products, they actually tried to get Microsoft to support Intel's own 64-bit x86 in Windows in addition to AMD64. This would have made AMD64 irrelevant, in much the same way that MMX had previously made AMD's 3DNow! irrelevant (on average, independent software vendors faced with similar effort to write software for two similar chips are going to prioritize the one with higher sales volume).

Microsoft refused. They didn't want to fragment 64-bit x86, and they felt that if Intel had wanted to have a voice they should've beaten AMD to the punch, or at least participated when AMD published first. And, of course, it was in Microsoft's best interests to resist Intel's attempts to monopolize x86.

Intel was then forced to go back to the drawing board and rework their extensions to mostly match AMD's, which significantly delayed their time-to-market. It's known that at least one Pentium 4 generation shipped with dark silicon implementing Intel's original, never-disclosed-in-public 64-bit x86 extension. There are still differences between AMD64 and EM64T, but they're confined to privileged VM-related instructions which only kernel-level software can use. The userspace ISA got fully harmonized, which is what Microsoft needed in order to unify the software market.

Most of the changes to Intel's processor designs would've been in the decoder front end, rather than the actual execution pipeline: a 64-bit add is a 64-bit add. So it wasn't as huge a task to rework the design to support AMD64 as you might think, though obviously it cost Intel a product generation.

Re:Just like Bulldozer?

[haruchai]

Right. Had forgotten about that, my bad.
Still very glad I never spent a penny on Dell hardware.

Re:Just like Bulldozer?

[Anonymous Coward]

Technically speaking, it would be great to have the nation run completely on nuclear power, but there are costs to that, coupled with natural disaster risks, so we moved in other directions. It wasn't practical.

You left something out - Included in the working definition of 'practical' at the time was "can the design be used to make Pu239?" If a design, like Liquid Thorium Salt or Pebble Bed, couldn't make bombs, it didn't get funding. To maintain the ability to make megaton explosions, we accepted higher (forget about nature) man-made disaster risks like Chernobyl, Fukishima, Hanford, and Savannah River and blame nature for the result of our global myopia.

Re:Just like Bulldozer?

[LordLimecat]

Except the chip you have now is roughly
  * 66% faster across the board, per thread
  * 70% faster at multithreaded ops
  * 800% faster at AES operations
  * 300%+ faster on video transcodes
and to top it all off,
  * 66% less power draw at any given time, and
  * statistically speaking, probably 2x as many cores.

Other than that, yea! No real changes.

Re:Just like Bulldozer?

[khellendros1984]

Oops, should've worded that differently. I was thinking of the G5 (or the 970 series, if you want to stick with the IBM names). I was just looking for a well-known 64-bit IBM-produced chip from before 64-bit x86.

Re:Just like Bulldozer?

[WuphonsReach]

Itanium 64-bit failed for one reason. It could not execute 32-bit workloads at the same speed as an equivalently priced 32-bit Xeon chip.

With an AMD Opteron chip, you got the best of both worlds. A chip that could do 64bit, and it could run your existing 32bit software as fast as your old CPU.

That made moving to Opteron a no-brainer decision. You got better performance from having a newer chip, even if you weren't ready to jump to a 64bit OpSys. PLUS, when you finally did move to 64bit operating systems, your CPU chip was ready and waiting.

The other reason that AMD ate Intel's lunch for a while was that they were the first ones to drive the cost of dual-core chips below $150. Intel was still charging $300+ for a dual-core CPU while you could pickup AMD Athon x2s for under $150. And dual-core makes a huge difference in how responsive the machine feels compared to a single-core CPU.

Re:Just like Bulldozer?

[Anonymous Coward]

3 years ago was already the 2nd core ix architecture. If anything, performance is at best 30% faster per thread. He has an i7, so exactly the same number of cores and threads. Yes the 4th generation ix do draw much less power, but the CPU is only part of the story.

Essentially, performance-wise, Desktop CPUs have stalled. I bought my core i7 920 (1st generation) in 2009, and I can find no reason to upgrade. 40% faster across the board at best is not noticeable. In comparison, in 2004 I had bought an Athlon 64. Single core, 2GHz. The performance improvement between the Athlon and the core ix were astonishing.

Re:Just like Bulldozer?

[smash]

Intel do not massively over-charge for their product. If they did, people would buy AMD instead.

Re:Just like Bulldozer?

[smash]

Um.... Itanium failed for 2 reasons. The second one being price. It wasn't just Xeon priced, it was way more expensive than that. It wasn't just outperformed by an equivalent priced Xeon (that didn't exist, they were cheaper), it was destroyed performance wise by consumer non-xeon CPUs on x86 code.

Which happens to be all of the code people actually cared about running.

Re:Just like Bulldozer?

[smash]

The question is why would you bother? A modern x64 CPU has billions of transistors. The 486 CPU (which has 16/32 bit compatibility) was implemented entirely in around 1 million transistors. You're going to save a fraction of a percent of the CPU die, whilst breaking virtually all software written prior to 1992, and probably a lot of niche software written after that. It's not worth it.

Re:Just like Bulldozer?

[smash]

Nah, the original guy posting about his Macbook pro is correct. I also have a 2011 macbook pro with Sandy bridge, an i7-2720 quad core, 2.2Ghz. GPU performance is different, power consumption is better now, but in terms of raw performance on CPU, it's a wash. The 1990s style gains of 2x performance every 18 months or so are well and truly gone.

But yes, i agree, desktop CPU performance has pretty much stalled since the Sandy Bridge machines. I suspect it is because AMD simply dropped the ball and intel hasn't seen any need to push the envelope yet.

Re:Just like Bulldozer?

[Bite The Pillow]

Performance is good enough that they spend effort making it more efficient for battery life, instead of letting the CPU idle .1% more time. And you say this is not progress?

What if the about page displayed estimated power consumption instead of speed? Progress would be evident.

Re:Just like Bulldozer?

[LordLimecat]

Wasnt sandy bridge just coming out in may, 2011? A computer owned at that point would have a nehalem architecture, which i was comparing with haswell (which is in the wild now, and has been for a year).

Re:Just like Bulldozer?

[kasperd]

Think of it this way, when you've worked on code that's 10 years old and you think "this would be so much better if we could throw it away and start from scratch" imagine that Intel thinks the same way with x86 only it's dealing with a 40 year long chain of incremental improvements.

At least AMD64 did do some preparations for ditching some of the cruft. The 64-bit mode of the AMD64 architecture left out some of the features of the original x86 design. If we can get the 16 bit BIOS interfaces replaced with 64 bit interfaces, then it would make sense for the next generation of CPUs to switch on in 64 bit mode. After that, it won't be long before you can completely drop the 16 and 32 bit support from the CPUs. Support for 32 bit user mode on a 64 bit kernel may need to live for a little longer though.

Re:Just like Bulldozer?

[K. S. Kyosuke]

Let's hope he won't get run over by a bus. ;-) A bus factor of one would be bad for any company's success. AMD's size just multiplies the risks.

Re:Just like Bulldozer?

[Lonewolf666]

Depends on how much that percent of the CPU die holds the rest back in terms of complexity and maybe performance limitations (not really my area of expertise). You may be right that it does not really matter.

On the other hand, "prior to 1992" means DOS and maybe Windows 3.x software. I'm aware that there are still a few DOS-based maintenance tools for the PC around, but otherwise I don't know anyone who still works with DOS software.

I used to work for a company that was really backwards that way, until a few years ago they produced a medical device with DOS-based software as "implicit real time system" (no other thread that can steal the CPU). But even they have given up on DOS, as the technical limitations became too bothersome. The successor of that device, now on the market, uses Windows 7 with a real time extension to the OS.

Re:Just like Bulldozer?

[Rich0]

When the AMD extensions became mainstream, the defacto standard, Intel licensed it. But the extension is still tied to the Intel architecture and Intel is still in control.

What does it even mean to be "in control?" AMD could apparently fork the last architecture/instruction-set/whatever-you-want-to-call-it (AIW for short), and create their own AIW which everybody adopted, just as Intel routinely adds new instructions to newer chips.

I don't question that Intel is usually in the lead with advancing the x86 world. However, it is fair to say that by introducing amd64/x86_64/whatever-you-want-to-call-it (AXW for short) AMD brought 64-bit to the masses. They certainly didn't invent the first 64-bit chip, but they created the first one that anybody was willing to use.

I suspect that today moving to ARM or some other architecture would be easier than it would have been 10 years ago. Back when the Athlon 64 came out virtually all workstations ran Windows, with all software being win32. On the server front Linux was popular, but even in that world there was a lot of non-portable software around. Using AXW made it a lot easier to migrate, and at least in the Linux world I'd like to think that people have learned to avoid reading a 32-bit record using an int data type. In the non-linux world things are much more heterogeneous than they were a decade ago as well. That means that changing instruction sets entirely is more practical than it used to be.

Re:Just like Bulldozer?

[TheRaven64]

Fab building is expensive and so is fab R&D. You want to spread the cost as much as possible. For the R&D costs, you want to build multiple fabs for the same process. For the fab costs, you want to make sure that the fab is running at full capacity for a very long time, even when it is one or two generations old. Intel can do this, because they produce a lot more chips than AMD and because they produce a large range of ICs that can be produced on older fabs (AMD does this to a far smaller degree).

For AMD, they can spread the R&D costs by allowing GlobalFoundaries to produce chips for ARM SoC vendors (for example) as well as for them. GlobalFoundaries can then keep their fabs open and producing things for smaller customers that don't need the latest generation for many years after AMD no longer has anything that they want to produce in that space.

Re:Just like Bulldozer?

[Theovon]

Actually, a Microsoft engineer developed the 64-bit extension, and AMD adopted it.

Re:Just like Bulldozer?

[Theovon]

I've seen a number of papers that squeeze more performance out of x86 processors by being micro-op aware. Some just more carefully choose among instruction sequences, given awareness of the micro-ops that will be generated, while others consider what-if scenarios over what more performance you could get if you could have more control over the micro-ops themselves. For instance, given an ISA sequence and the corresponding micro-op sequence, is there a functionally equivalent alternative micro-op sequence? To some degree, compiler writers have to reverse-engineer the micro-op generation in order to generate micro-op sequences that schedule better. Even Intel compiler writers do this; although they have access to exactly what micro-ops are generated it's still not trivial to get the optimal micro-op sequence.

First to 64-bit

[Anonymous Coward]

You mean first to x86-64. Intel had a 64-bit processor before that (Itanium). 13 years later, Itanium is dead and x86 is holding us back, so much that servers are turning towards ARMv8 (inferior design to Itanium, but tons of momentum from mobile/embedded).

Re:First to 64-bit

[danomac]

First in the sense that Apple made the first tablet. ;-)

Re:First to 64-bit

[toejam13]

While the Itanium ISA may be dying, a lot of the redundancy, error detection and error handling from Itanium has made its way into the Xeon line.

And x86-64 isn't holding all of us back. Different processor architectures excel in different places. ARM is good for mobile and clusters. x86-64 is good for desktop and big iron.

Re:First to 64-bit

[jon3k]

ARM servers excel at some workloads. The only people looking at ARM are people running massive web server farms. Xeon still crushes ARM in performance per watt and absolute performance for almost every server workload. It's definitely an interesting area and somewhere I expect ARM to really grow but right now it's not much of a contest.

Re:First to 64-bit

[bobbied]

You mean first to x86-64. Intel had a 64-bit processor before that (Itanium). 13 years later, Itanium is dead and x86 is holding us back, so much that servers are turning towards ARMv8 (inferior design to Itanium, but tons of momentum from mobile/embedded).

You do realize that this run towards ARM is not a full stampede, and is driven by price and operating costs and only useful for Unix/Linux systems as windows server isn't really interested in supporting ARM yet. This is more like a trickle of some large specialized systems off onto Red Hat (or similar) systems where one can afford to just change processors and recompile everything in an effort to same a bit of operating power and hardware costs. But you have to be looking at enough servers to make this worth the labor cost.

So, where I don't care for the X86 family and would love everybody to switch to ARM, I know it's not going to happen in my career without there being that "killer" app that pushes everybody off of Windows. Right now, with "Office" being the "killer got to have" application of all time, and that generally only running on Windows, guess what? X86 is here to stay.

Re:First to 64-bit

[iggymanz]

Itanium not quite dead yet, about 8% of the "Unix" big iron market last year. dying yes.

Re:First to 64-bit

[Yunzil]

I've been hearing "x86 is holding is back" for, oh, over 20 years now.

Re:First to 64-bit

[smash]

The newer atom style CPUs, and even recent core i series have massively dropped x86/x64 power consumption. I suspect ARM in the server room is a flash in the pan, and intel is going to prevail in that space simply due to the huge legacy of x86 code. ARM will face the same problem as Itanium in that respect - performance emulating x86 code will suck.

Sure, for basic web services on open source software ARM may be fine, but with cloud services being the current big thing, and companies wanting to migrate their x86 based applications to the cloud, ARM just isn't going to cut it. Atom is similar in power consumption to ARM now and provides a relatively seamless transition to/from full-fat x64 Core series or Xeon as performance requirements dictate. ARM simply can't scale to that degree at the moment.

Re:First to 64-bit

[smash]

Agreed. And it hasn't really been true since the 386 introduced protected mode.

Re:First to 64-bit

[Anonymous Coward]

Office on Windows? You are thinking desktop. the discussion was about server.

Re:First to 64-bit

[macpacheco]

ARMv8 an inferior design ? On what grounds ?
For its intended market ARMv8 is the best tech available. It's not meant to replace traditional Intel markets. Its about creating new markets that will eventually replace the typical Intel market (desktop computer mainly).
Very happy typing on my Cortex A15 Chromebook. My main issues with this machine are the limitations of ChromeOS and having only 2GB of RAM. Nothing a 4GB RAM, Cortex A53 Chromebook wouldn't solve (plus re installing a new Linux replacing ChromeOS). Haven't replaced ChromeOS yet, since the alternatives haven't figured out how to do video acceleration yet.
In another 5 years ARM will really be killing Intel. What they need is to stay the course, instead of succumbing to the temptation of trying to compete with the Core iX CPUs. The vast majority of the users have zero need for a Core i7, or even a Core i5, as long as they get rid of their WinBloat OS and use Linux instead.
What I want is an ARMv8 cpu that has the performance of the latest Core i3, even if they need 8 or 12 cores to do that.

Re:First to 64-bit

[unixisc]

You mean first to x86-64. Intel had a 64-bit processor before that (Itanium). 13 years later, Itanium is dead and x86 is holding us back, so much that servers are turning towards ARMv8 (inferior design to Itanium, but tons of momentum from mobile/embedded).

If you're talking about the first 64-bit CPU, that was MIPS R4000. Itanium came relatively late. And like everything else - MIPS, SPARC, POWER, Alpha, they too were completely incompatible w/ the Pentium line

Re:First to 64-bit

[Anonymous Coward]

Office runs on ARM. Windows Server can definitely run on ARM64, will it ship that's another question. MSFT does some amazing stuff that ends up not shipping for business reasons. They also have significant input on AMD64 and ARM64 ISAs. MSFT refused to support an alternative to AMD64 from Intel after Itanium, the message was "we have support for 64-bit x86 ISA thank you".

RTFA

[Anonymous Coward]

DId you RTFA?

His task will be a new microarchitecture to overcome some of the shortcomings in AMD's current generation microarchitecture, called Bulldozer. Bulldozer adopted clustered multi-thread (CMT) designs[...] Bulldozer is inherently less efficient than Intel's chips[...]What Keller will do, no one knows. And AMD would be nuts to tip its hand. The most logical move for Keller would be to dump the CMT design in favor of a design with simultaneous multi-threading (SMT)

More of the same? Probably not.

Re:RTFA

[jandrese]

"The current design is bad, are you going to use a better design?"
"We won't say, it's a secret!"

What a compelling article.

Re:RTFA

[nigelo]

What a maroon.

Re:RTFA

[Anonymous Coward]

When intel realized that the netburst architecture was a dead end, they took the best thing about it (the bus) and grafted it on to a modified version of the architecture they'd abandoned for netburst. Do a die shrink, pack in multiple cores, and the result was the core/core 2 architecture. It's pretty clear that core 2 was the foundation of intel's climb back to relevance and then dominance among enthusiasts (obviously mainstream and business never left, but that's not what most people reading this article care about).

Even though intel currently powers all but two of my 8 machines (and one of those is only by circumstance), I really hope that AMD can finally do the same thing intel did. We need real competition again. The market SUCKS when one company is dominant. not counting the nehalem -> sandy bridge jump, we haven't seen an increase in performance from one chip generation to the next of more than 8%-10% or so since... well... core 2.

Re:RTFA

[Pinhedd]

Hyperthreading (Intel's implementation of SMT) is what gives Intel's i7 series microprocessors a huge advantage over AMD's FX series equivalents.

In terms of pure scalar arithmetic an i7-4700 series microprocessor Haswell and FX-8300 series Piledriver microprocessor have nearly identical clock-for-clock capabilities. Haswell has 4 scalar ALUs per core, whereas Piledriver has 2. SMT allows for the backend execution resources (which includes the scalar ALUs) to be balanced between the two SMT frontends, whereas CMT does not. Disabling or idling one of the SMT frontends improves the instruction throughput of the complementary frontends as they will have greater access to the shared resources due to reduced fighting. Disabling one of the CMT frontends only reduces competition for resources that are shared, which on AMD FX series microprocessors includes some of the cache and floating point hardware. The unshared arithmetic hardware, which is responsible for the bulk of the instructions in most programs, is idled along with the frontend that it belongs to.

CMT is inherently less efficient than SMT. It's also a simpler design that's easier for a smaller company to implement.

Re:RTFA

[OneAhead]

Not everyone runs workloads that are poorly vectorized and parallelized, you insensitive clod.

Hyperthreading (Intel's implementation of SMT) is what gives Intel's i7 series microprocessors a huge advantage

The P4 had hyperthreading too. If that really would be such a huge advantage, one would think it would have been a bit more competitive than it was...

Disabling one of the CMT frontends...

...assuming the workload is not keeping all the frontends busy most of the time.

...only reduces competition for resources that are shared, which on AMD FX series microprocessors includes some of the cache and floating point hardware.

Not with AVX-intensive workloads; there, a single thread can keep the whole shared FPU busy with AVX instructions.

CMT is inherently less efficient than SMT. It's also a simpler design that's easier for a smaller company to implement.

{citation needed} on both accounts.

Re:RTFA

[Bengie]

Once you account for L1 bandwidth and load latencies, AVX FPU "sharing" still isn't an issue. L1 is only so fast and 32byte loads takes a bit, and the scheduler can let the hyper-thread do work while the other thread is loading.

Re:RTFA

[Pinhedd]

Not everyone runs workloads that are poorly vectorized and parallelized, you insensitive clod.

Very few consumers actually run workloads that are properly vectorized and parallelized. Explicit vectorization requires manual creation of multiple codepaths based on the level of vector support in the hardware. Autovectorization is substantially more flexible but requires a decent compiler to pick it up; ICC is by far the best for this, and GCC still struggles after many years of development. Furthermore, the proliferation of virtual-machine based languages means that consumer application developers have largely absolved themselves responsibility for writing code that is vectorized much less suitable for auto vectorization by a JIT that actually does so. Heck, SIMD support in Javascript is just beginning to materialize.

Hyperthreading (Intel's implementation of SMT) is what gives Intel's i7 series microprocessors a huge advantage

The P4 had hyperthreading too. If that really would be such a huge advantage, one would think it would have been a bit more competitive than it was...

Netburst had a ton of issues with it that crippled performance across the board. The HT design was also rather immature. The implementation of HT in later releases of the Itanium series and Nehalem were vastly improved.

Disabling one of the CMT frontends...

...assuming the workload is not keeping all the frontends busy most of the time.

There are only a handful of common consumer applications that keep 6 or even 8 frontends busy at all times. AMD's FX series microarchitectures tends to keep up with Intel Core microarchitectures in such applications, yet fall behind in the ones that consumers spend most of their time running. Javascript, the language that powers the web for some strange reason, is inherently single-threaded.

...only reduces competition for resources that are shared, which on AMD FX series microprocessors includes some of the cache and floating point hardware.

Not with AVX-intensive workloads; there, a single thread can keep the whole shared FPU busy with AVX instructions.

That's correct. The vector unit in both AMD's FX series and Intel's Core series microprocessors are shared between two front ends. although on Intel's architecture the add and multiply vector EUs are on separate ports and can accept issues from separate threads in the same cycle (albeit in lieu of two scalar arithmetic instructions), I'm not sure if AMD's architecture works the same way (although I think that the instruction latency is longer). What I was discussing is that under AMD's CMT design the architecture is unable issue instructions to ALUs on the module's paired core, whereas SMT allows this by virtue of having a completely common backend with a unified reservation station. If one of the frontends on an AMD FX series microprocessor is disabled, the two ALUs are disabled along with it and the result is a typical 4-way SMP with 2 ALUs per logical processor. If Hyperthreading is disabled, the result is a 4-way SMP with 4 ALUs per logical processor as ALUs can still be issued instructions from the unified reservation station. SMT allows for flexibility that simply doesn't exist under CMT.

CMT is inherently less efficient than SMT. It's also a simpler design that's easier for a smaller company to implement.

{citation needed} on both accounts.

There are piles upon piles of benchmarks out there demonstrating this. Intel's architecture excels in instruction throughput, transistor budget, and power efficiency.

Look at the price of AMD's microprocessors on any online retailer's website. Intel's i7-3930k still sells for around $600 and its successor is around $630. AMD's flagship FX-9590 fell from $1000... to $600... to $300 in a matter of weeks as it just can't keep up where it counts.

Re:RTFA

[smash]

The p4 had hyper-threading added in an attempt to save the botched design that caused unused CPU resources due to pipeline stalls.

Intel are coasting at the moment. Why release something faster than the already fastest consumer CPUs available on the market they are already making when AMD are so busy shooting themselves in the foot? Let AMD spend billions on a new architecture, then release the faster CPU that intel no doubt already have waiting, to compete with that.

Re: RTFA

[Anonymous Coward]

Yeah, almost to the point of chestnut!

The word you wanted was "moron".

Re: RTFA

[Anonymous Coward]

Whoosh

You little kids think you're so smart.

Re: RTFA

[Anonymous Coward]

Way to prove the point.

What an ignoranimus...

Re:RTFA

[OneAhead]

Yeah, I guess I'm a bit biased in that we're constantly running stuff that keeps as many threads busy as we want, and we do indeed see that this dramatically decreases Intel's advantage. I do see your point that this is not a typical use case.

CMT is inherently less efficient than SMT. It's also a simpler design that's easier for a smaller company to implement.

{citation needed} on both accounts.

There are piles upon piles of benchmarks out there demonstrating this. Intel's architecture excels in instruction throughput, transistor budget, and power efficiency.

Look at the price of AMD's microprocessors on any online retailer's website. Intel's i7-3930k still sells for around $600 and its successor is around $630. AMD's flagship FX-9590 fell from $1000... to $600... to $300 in a matter of weeks as it just can't keep up where it counts.

For the "less efficient" part, I would accept "less efficient for single-threaded applications" or "less efficient for consumer applications", but inherently less efficient is strong wording and I'm not convinced it is true. Just as you haven't given me any evidence that CMT is simpler and easier to implement. AMD has shown in the past it can match Intel in design sophistication (as opposed to fabbing, where it was usually behind, except for the short period of time it had SOI and Intel didn't). It was my understanding that its choice to not do SMT is due to philosophical differences, not design complexity.

Re:RTFA

[Pinhedd]

SMT on a dynamically scheduled architecture requires resolving and tagging data dependencies between instructions from two or more contexts as those instructions enter the reservation station, are dispatched to execution units, and eventually enter the reorder buffer. Speculative and cancelled instructions also need to be resolved from two or more contexts at once. That's not particularly easy to do, and the difficulty grows with the number of execution ports and accompanying execution units. Intel has been working at this for many, many years. Even when HT was not used in x86 (Core 2 era) it was being developed into the Itanium family of microprocessors.

If the ideality condition for efficiency is having all backend execution ports busy on every cycle CMT can theoretically reach parity with SMT by having a demanding thread running on each logical processor and without other factors such as cache latency becoming a bottleneck (assume ideality on these). However, outside of synthetic benchmarks this is incredibly hard to accomplish. As soon as one thread blocks (IO syscall for example) or enters a long-latency event (page fault for example) the operating system can either toss the thread on the waiting queue and context switch, or simply do nothing and issue stalls until the event is resolved. Excessive context switches cause overhead, and should be avoided, and stalls are inefficient by definition. If no context switch is performed, the CMT frontend must stall which means that the backend execution units will be idled. A SMT frontend must also stall, but the backend execution units may still be used unless the complementary thread also encounters a long-latency event.

Intels performance advantage most certainly does erode when highly-concurrent tasks are employed, but AMD's microprocessors require significantly more transistors and significantly more power to obtain the same level of performance.

Intel Inside, inovation outside.

[Anonymous Coward]

It was never technology or the lack of it that kept AMD out and Intel in. It was clever marketing, FUD and just plain ignorance of the customer. The "Intel Inside" ads and the "what if something is not compatible with AMD" feeling that the marketing gurus created kept Intel on the top outselling even the more superior AMDs. The real Intel killers are the ARM processors and mobile computing that is giving Intel a run for its money. This is what happens if you refuse to innovate!

Re:Intel Inside, inovation outside.

[Anonymous Coward]

There was also the contracts PC makers signed with Intel, keeping the chips out of the wildly selling (at the time) Dude, you're getting a Dell market.

Re:Intel Inside, inovation outside.

[Defenestrar]

Yeah? Well the original Athlon 64 dual core isn't compatible with modern Window's operating systems. The instruction set which prohibits it from being used was around at the time, AMD just didn't put it in.

Re:Intel Inside, inovation outside.

[Anonymous Coward]

Lol as if windows 8 wasn't enough of a clusterfuck already.

More reason to stick with 7 or go to Linux.

Re:Intel Inside, inovation outside.

[khellendros1984]

If you want to be picky, it *is* supported by Windows 8.1, but only if you use the 32-bit version of the OS.

Re:Intel Inside, inovation outside.

[OneAhead]

Wow, just wow. I can understand they want to start using newer instructions, but not with a release that amounts to a service pack! They should either have done this with windows 8, or waited till windows 9. What the hell were they thinking?

Re:Intel Inside, inovation outside.

[ChunderDownunder]

Could they not just patch that in software - like emulating an FPU on 486SX ?

Buh?

[drinkypoo]

But the real fight of a decade ago, when AMD was first to 1GHz, the first to 64-bit, the first to dual core, seemed missing. It's not surprising since the company was facing a real threat to its survival. But with a gravy train from the gaming consoles, it looks like the company is ready for a fresh battle, with a familiar face at the helm.

Uh, wait. No. It was surprising when AMD was the performance leader. It was surprising because they were broke. It's not surprising to see AMD pushing out a new architecture now that they have money. It takes a lot of money to do that. So we start out completely ass-backwards here.

Much elided, then

The most logical move for Keller would be to dump the CMT design in favor of a design with simultaneous multi-threading (SMT), which is what Intel does (and IBM's Power and Oracle's Sparc line).

Wait, what? Why? Why wouldn't it make more sense to just fix the lack of FP performance, perhaps by adding more FP units? Why would it make more sense for them to go to a completely different design? It might well, but there is no supporting evidence for that in the article.

Re:Buh?

[dagamer34]

It wasn't that surprising that AMD was king around 2003-2004, the problem was that Intel was playing very dirty, signing deals with OEMs like Dell to specifically NOT use AMD chips. The fines Intel got from the EU are never going to do as much to help AMD as actually gaining more profits during that period (and who knows, they may not have sold their mobile Radeon group to Qualcomm in an effort to raise cash). It's the domino effect of unknowns that hurts the most.

Re:Buh?

[jandrese]

AMD was dominant while Intel was chasing dead ends (Netburst and Itanium). Once Intel woke up and started working on sane chip designs again AMD's goose was cooked. They just can't compete with Intel's R&D budget. Plus, AMD made some boneheaded decisions of their own, like firing a bunch of their R&D staff in the belief that computer automated chip layout would prove superior to human designed layouts.

Re:Buh?

[drinkypoo]

AMD was dominant while Intel was chasing dead ends (Netburst and Itanium). Once Intel woke up and started working on sane chip designs again AMD's goose was cooked. They just can't compete with Intel's R&D budget.

Well, that was my point, AMD can afford to have an R&D budget right now. But you're right, intel spent a lot of time dicking around with nonsensical architectures that they might well have been able to spend crushing AMD sooner. On the flip side of that, though, is the question of whether they could have actually been more effective. Too many cooks, and all that. Spending more money doesn't necessarily result in getting where you want to go sooner. You tend to go somewhere, but not necessarily in your chosen direction. Itanic illustrates that better than anything else mentioned so far in this discussion.

Plus, AMD made some boneheaded decisions of their own, like firing a bunch of their R&D staff in the belief that computer automated chip layout would prove superior to human designed layouts.

Yeah, I think that will eventually happen, but not now. Right now, human insight is still king. On the other hand, maybe they spent their staff budget on R&D :)

I don't think intel is immune to competition. AMD with some cash is still dangerous to them. I won't hold my breath, but I would like to see them remain relevant.

I would also like to see them figure out graphics drivers, so I can buy their integrated processors...

Re:Buh?

[oh_my_080980980]

That a Microsoft dragging it's heels with a 64 bit OS.

Re:Buh?

[aliquis]

Also thank Microsoft for the useless monitor resolutions I guess.

Re:Buh?

[Darinbob]

Itanium was a sane design. Extending the tired old x86 system wasn't a good design, although it was a good business choice.

Re:Buh?

[evilviper]

AMD was dominant while Intel was chasing dead ends (Netburst and Itanium). Once Intel woke up and started working on sane chip designs again AMD's goose was cooked.

Those were only dead-ends BECAUSE AMD was killing them. Intel would have been happy in a world where P4s sold, and Itanium was the only 64-bit upgrade path.

AMD got a bigger lead thanks to Intel's missteps and slow response, but that's about all you can say. AMD has been the innovator in the past, and has had similar surges in market share, when Intel was going down more expensive roads were enthusiasts didn't want to follow.

The old K6-2 was doing pretty well, offering an option for OEMs to continue to use old Pentium motherboards, while Intel was trying to push everyone to use expensive Pentium-IIs in a Slot-1 motherboard that everybody hated. AMD's 3DNow extensions did have Intel at a disadvantage for a while. Oddly enough, I don't think you can look back at the Intel PII and say it was a "dead end" despite its problems.

Throw in all the other CPU-orthogonal problems like Intel trying desperately to force Rambus memory on everyone. Or previous "dead-end" CPUs like the PentiumPro, which won't really die, and we've seen this dance dozens of times over.

I suppose you're right... If Intel NEVER makes ANY mistakes, then AMD won't see a significant bump in their sales. But if AMD can just get closer to competitive with Intel, while keeping prices lower, they could claw back the market share they've lost since the "Core" processors debuted.

Re:Buh?

[amorsen]

No. Itanium was completely insane. Itanium took everything ever invented in computer architecture and tried to fit it onto one chip. At the same time it added every feature from x86 and PA-RISC, apart from the actual ISA of course, in order to simplify porting operating systems and other software. Making e.g. 10W Itanium chips is out of the question unless you software-emulate the whole thing at 10MHz on a sane architecture.

Anyway, Itanium succeeded in killing off the Unix market for MIPS and PA-RISC and almost killed off SPARC (but Sun veered at the last moment). It was a great success, politically. I do not believe Intel management was smart enough to see that ahead of time though; if they did it was one of the most shrewd business moves in history.

The x86 ISA has a much worse reputation than it deserves, and I say that as a RISC lover. The encoding takes up less space than almost all RISC ISAs, which is helpful if you are power constrained. Saving state is really quick on x86, so context switches and interrupts are fast. In contrast, Itanium encoding is horrendously verbose due to all the NOP instructions, and saving state on Itanium is extremely slow.

Re:Buh?

[AmiMoJo]

You seem to be confused. Itanium uses very long institution words and is not so good for code density. The architecture is simplified compared to x86, forgoing things like out of order execution and branch prediction. The idea was that the compiler would optimise so well it wouldn't be needed, but that turned out to be really hard.

X86 is a mess. Even the ISA varies between manufacturer (SIMD extensions). It only runs fast because it pretty much rewrites code on the fly. That's why early Atom CPUs sucked. It took them years to get that stuff working at low(ish) power.

Re:Buh?

[Anonymous Coward]

You forgot the biggest bonehead AM2 w/ABSOLUTELY NOTHING new, well it had DDR2 but BFD and a new socket.

Meanwhile Intel had taken the pentium-m, which was already giving k8 troubles, and turned out core at the same time that AM2 appeared.

AMD had NO excuse for NOT doing better as they WERE making money in those days, just, perhaps, not quite as much as they could have been, but their failure since k8 is ENTIRELY their own fault. Crap x86 core designs.

I built an FX-9590(same price as 8350, so WTF not) system(just because I remembered the good old days of socket 939/4800+ x2) and was surprised at how sluggish it felt compared to my several year old i7-3930k. Both have similar amounts of DDR3 memory(32GB), similar OS(i7 win7 x64/fx win 8.1), identical wd 10k hdds, and GPUs that should've been close in relative power(i7 GF 670/fx r9 280x(going all AMD/ATI oon bigredone)), and both havegood mobos(i7 ASUS p9x79/fx asrock extreme9(only have 3 choices)). So yes, it still feels sluggish to use.

(I parked it in the corner for now as I wasted about an hour attempting to install the latest beta catalyst in April, but could only get the base driver to actually install, all the other pieces phailwhaled, e.g. control just phail install, appears to install correctly but nope phail to run, etc. and so I got tired of guru3ding DDU and decided to just park it for a while... might drag it out later today and prepare to waste another hour on those kwality "beta"(alpha would be more accurate) drivers...)

Re:Buh?

[amorsen]

There is absolutely nothing simplified about the Itanium. When it comes to branch prediction, it has both a traditional hardware branch predictor, but also predicate bits, branch hinting and 8 branch registers.

Yes, out-of-order execution was cut, but that is not really an ISA feature but more of an implementation detail. The Itanium ISA is almost impossible to execute out-of-order due to its enormous internal state among other misfeatures. The difficulty of going out-of-order was just another nail in the coffin for the slow and expensive Itanium.

Anyway, a minimal Itanium core is enormous compared to a "small" out-of-order core like the Pentium Pro, so simple is not an appropriate word for it.

Atom was just the Pentium MMX reheated. For integer workloads, it runs at pretty much original Pentium MMX instructions-per-clock, but at much higher frequencies of course. This year Intel finally made the Atom out-of-order, probably due to increasing pressure from out-of-order implementations of ARM.

Re:Buh?

[Anonymous Coward]

The Bulldozer is more lacking in integer performance than FP, though that could also use some improvements...

I'm Still Rooting for AMD

[Jaborandy]

I was so proud of them when they kicked IA64's ass with their amd64 architecture, beating Intel at their own game by choosing to be x86-compatible when even Intel didn't go that way. Then I was sad when amd64 started getting called x64, since it stripped AMD of the credit they deserved. Go AMD! A world without strong competition for Intel would be very bad for consumers.

Re:I'm Still Rooting for AMD

[jandrese]

I have to admit, I found the AMD64 moniker a little confusing when I first read it. I had to Google around to make sure my Core2 chip would support it and that it wasn't using some AMD proprietary extension top of the 64 bit extension. x86_64 is less confusing, even if it is more awkward to type out.

Re:I'm Still Rooting for AMD

[maz2331]

AMD has historically made huge gains in a stair-step fashion where they are flat or declining for a long period, then leapfrog way ahead of Intel, who then catch up and pass them in a more linear fashion.

Re:I'm Still Rooting for AMD

[Anonymous Coward]

As others have pointed out, AMD have historically beaten Intel when Intel fscks up. Intel needed the P4 to keep ramping up clock speed because it had sucky IPC, and it hit a brick wall, so AMD beat them because they had significantly better IPC at similar clock speeds. Intel wanted everyone to switch to Itanium, so that was their 64-bit push, while AMD pushed 64-bit into the x86.

As soon as Intel realized they needed good IPC on a 64-bit x86 you couldn't fry eggs on, AMD was back to second place, and have been stuck there since.

Re:I'm Still Rooting for AMD

[evilviper]

I was sad when amd64 started getting called x64, since it stripped AMD of the credit they deserved.

AMD called it "x86-64" when they brought it out. They only rebranded it as AMD64 about the time they licensed it to Intel, and Intel started calling it i64 or some such. It was, still is, and always will be: x86-64

Re:I'm Still Rooting for AMD

[JenovaSynthesis]

How could AMD64 "kick IA64's ass" when they were never in competition with each other? IA64 refers to Itanium which was never released for consumer markets but rather for high-end servers. It was to compete against Sun's SPARC, DEC's Alpha, and the SGi MIPS chips.

Re:I'm Still Rooting for AMD

[Anonymous Coward]

So, it's not so much that AMD goes through periods of incredible success on its own merits, but more that Intel goes through periods of mental retardation? Sort of like passing on both of the current-gen consoles, especially after screwing AMD out of the previous generation, and knowing how that played into their success of recent years?

I'm really not sure you are supporting your case in the way that you think.

Re:I'm Still Rooting for AMD

[Anonymous Coward]

Because the architecture has largely been retired, with major manufacturers dropping it altogether from most product lines, OS support for the platform nearly nonexistent, and most entities engaged in converting to x64 where possible? We just finished a project tasked with converting 3,000 HPUX Integrity VMs from Itanium-based superdomes to RHEL VMs running on x64-based ESX hosts.

Itanium may not be dead yet, but it's on life support. If that doesn't count, then what does?

Re:I'm Still Rooting for AMD

[JenovaSynthesis]

No, it does not count. Itanium's target market changed and they did not switch to AMD. If you want to use the "kicked the ass of" phrasing, Itanium got its ass kicked by Xeon\ because that is what the high-end servers, like the ones SGi produces now, are using that require strong number crunching if they are still using chips in the Intel/AMD ecosystem.

Re:I'm Still Rooting for AMD

[Anonymous Coward]

It's target market shrank, you mean. AMD64 is a standard, not a chip, and it won out over IA64. End of fucking story.

AMD Atanium

[Anonymous Coward]

Come on, you know you want AMD to produce something not-backwards-compatible. The drama, it would be so entertaining.

Re:AMD Atanium

[viperidaenz]

Liquidation isn't entertaining...

Drivers?

[Bigbutt]

Honestly they need a better team writing the drivers. You can have the best CPU/GPU in the industry but if the drivers suck, no one will want to buy them.

[John]

Re:Drivers?

[jeffmflanagan]

I've had plenty of issues with AMD video drivers, though those problems seem to be behind them, but is there really a problem with their CPU/chipset drivers?

Re:Drivers?

[Anonymous Coward]

If they keep dumping code into the open source driver, and linux gaming takes off, it's moot what they do.

Re:Drivers?

[jones_supa]

Not really, as far as I know. The CPU power management drivers and chipset drivers work well.

Re:Drivers?

[hobarrera]

This is incredible true and has been over a decade. They don't seem to realize that they have better hardware for the same price, but people refuse to buy them because it's hard to appreciate the betterness without proper drivers.
I'm sure the FLOSS crowd would also start embracing AMD if they did decent OSS drivers like Intel does.

They're fools

[Baldrson]

This was their opportunity to dominate the CPU market with the MIll CPU architecture and they blew it.

Re:They're fools

[Anonymous Coward]

Actually, they'd be fools if they tried to be the most dominate processor manufacturers instead of being the best processor manufacturers... but that kind of logic doesn't work in most peoples' minds.

Re:They're fools

[Anonymous Coward]

This was their opportunity to dominate the CPU market with the MIll CPU architecture and they blew it.

So...you pitched AMD on your unknown, untested architecture that has never been implemented and only exists on paper... and the went with ARM instead? The fools

I mean it's not like ARM has mature compilers (multiple) OS support, strong developer mindshare, tons of performance data to drive design on real silicon to drive CPU design tradeoffs, like MIII already does...oh..wait minute...

2 years is a long time to wait

[edxwelch]

I think profits will be from consoles, GPUs and low end APUs for the time being

wrong

[Charliemopps]

Sorry AMD, you're heading in the completely wrong direction. CPUs are already plenty fast. They have been for years. 3D gaming is starting to look like just another "Gold plated speaker wire" guy hobby as everyone moves to mobile devices.

The real winners in the future are going to be the very cheap, very efficient chips. Do you want one very powerful computer to run everything in your house? Or do you want everything in your house to have its own dedicated, highly efficient CPU that does just what that device needs?

Re:wrong

[Bruinwar]

Why did I still wait if my CPU is "plenty fast" enough? My dream is to some day work on a machine that waits for my input rather than me always waiting for it. This means faster everything, including the CPU.

Re:wrong

[mlts]

What AMD should consider are FPGAs and different power cores on the same die. This isn't anything new, but done right, it can go a long way in the server room.

The FPGAs can be used for almost anything. Need a virtual CPU for AES array shifting? Got it. Need something specialized for FFT work? Easy said, say done. Different power utilization cores would be ideal for a server room where most of the hosts see peak load use, then after quitting time, end up idle.

Re:wrong

[Anonymous Coward]

I heard a saying that the most reliable way of making money in a gold rush is to sell picks and shovels. The 21st century equivalent is probably in tech bubble the most reliable way to make money is selling servers and network hardware. I would guess AMD is chasing the server farm market.

Re:wrong

[Anonymous Coward]

The real winners in the future are going to be the very cheap, very efficient chips.

Maybe they should hire you as CEO? Or maybe that's why they are working with ARM as mentioned in the summary?

Re:wrong

[viperidaenz]

Anything with an FPGA is always going to be in a niche market.
It won't let you multi-task since you can't reprogram it when ever you context switch.

Users don't want messages say they need to wait for X to finish before they start Y.

They're also very expensive because they use a lot of silicon.
They also consume a lot of power too.

Re:wrong

[Anonymous Coward]

ladies and gentlemen - an idiot (or just poorly informed with a penchant to draw premature conclusions... so I suppose... an idiot).

"CPUs are already plenty fast"

feel like ive heard this before... then this stuff called software comes out and makes CPUs feel not so fast...

"3D gaming is starting to look like just another "Gold plated speaker wire" guy hobby as everyone moves to mobile devices"

lol ok well gold plating has a specific anti-corrosive usage, so... not the best comparison. and uhhh.. angry birds, for a lot of people, isn't quite on par with say, mass effect. a bit like saying 'why watch a movie when I have all these digital shorts?!' and while luddites (or old people) might still cling to the near-sighted notion that 720p is "good enough", 4k and 8k have objective benefits, and require tons of vram bandwidth to run relative to games...

"The real winners in the future are going to be the very cheap, very efficient chips"

cheap and fast is best? YOU DONT SAY! the real winners will be whoever can adequately run the in-vogue software the cheapest. derp.

Re:wrong

[Loki_1929]

CPUs are already plenty fast. They have been for years.

Incorrect. CPUs are plenty fast and have been for years for doing many common tasks. The fact is that they aren't nearly fast enough (particularly for single-threaded items) and almost certainly won't be for another decade or more. There's a limit to what and how much you can multi-thread, and even then, you're still limited by single-thread performance x number of threads.

So yes, for grandma playing Blackjack on Yahoo, today's CPUs are plenty fast. For me and many others? The fastest stuff available is 100x slower than "fast enough".

Do you want one very powerful computer to run everything in your house? Or do you want everything in your house to have its own dedicated, highly efficient CPU that does just what that device needs?

I want computers (and servers, especially) which are able to perform their particular function without me having to wait on them. Ever. I want usable speech recognition feeding into a responsive AI that behaves as expected without delay (and God help you if you answer "Siri" to this). I want Eve Online to be able to stick 50,000 ships in one fight with full collision and damage physics modeling with zero lag. I want to be able to transcode, store, tag, and index 20 hours of home movies and a year worth of pictures without waiting. I want to run realtime and faster simulations of complex systems.

Are these common, everyday needs? Moreso than you might think. A lot of the back-end servers struggle to keep up with workloads that either expand or change over time. While much of what's right in front of your eyes seems pretty happy with the CPU that's there today, there's a lot of stuff happening behind the scenes that isn't. This causes server admins and developers to have to spend inordinate amounts of time, money, and cranial energy figuring out how to make it functional, giving the limited computing power available.

A lot of things need very little power, and they should have very little computers with very little CPUs to make them go. Some things - things you don't think about - need tons of power, either serially or just overall. I'd pay good money if Intel and AMD would stick with 4-12 cores and concentrate on making those cores enormously powerful. As it is, they're risking going the route of SPARC, and obviously that isn't working out well for SPARC. Interestingly enough, Oracle's trying to make SPARC more like x86 even as Intel and AMD are trying to make x86 more like SPARC.

Re:wrong

[Anonymous Coward]

I'm not sure about that. You come off seeming pretty slow. Maybe a tough of down syndrome or a heavy diet of paint chips. Do your eyes point in the same direction? Just curious.

Re:wrong

[jbolden]

I don't think CPUs are plenty fast at all. The stagnation in CPU speeds for the last 15 years has been dreadful for PC applications and the industry. Prices of computers have fallen, and software isn't much more capable than it was in 2000. If we had growth in CPUs like 1985-2000 for the 2000-2015 period ... it would be awe inspiring how terrific our machines would be today.

Re:wrong

[aliquis]

Bullshit.

You're one of those who would be happy with horses back in the day. Or the cave. Or whatever.

Re:wrong

[aliquis]

"I just don't want to have to kill my browser every day to free up resources" :D

(Then again older machine but I can't consider an Intel mini-ITX system today because 16 GB of RAM simply is waaaay too little RAM. And that's kinda always been the case. More RAM is better =P)

Re:wrong

[Anonymous Coward]

Siri...

Re:wrong

[Solandri]

3D gaming is starting to look like just another "Gold plated speaker wire" guy hobby as everyone moves to mobile devices.

Only for the short-term. In a 25-year timeframe, the tech needed for 3D gaming is going to become the most important branch in the computer industry. Why? Because we'll need it to drive holographic display technology. To generate a hologram in real-time, you need to convert a virtual 3D scene to a 2D interference pattern. Transmit that pattern to a display, shine the appropriate light at it, and you have yourself a hologram. Holographic film is about 4000 lines per mm, so the 3D hardware will have to be good enough to drive a display on the order of a million by million pixels in order to generate a hologram. We still have a lot of R&D to do in 3D graphics to make this a reality.

Re:wrong

[Kjella]

There's no doubt that Intel and ARM will clash and the winner will be looking at billions of dollars in profit. That does not mean that AMD's best choice is to get caught in the crossfire, when giants clash they're more likely to be stomped on than anything else. Their strength is that they're the only other one except Intel that can produce x86 chips and can compete with Core/Xeon for running "normal" desktop/workstation software and and a lot of closed source server software. In the ARM market, AMD is nothing special. They don't have their own fabs, they don't have particularly much experience with ultra-low power designs. In fact I was impressed with their Mullins performance but they're together with Intel on the wrong side of the ARM/x86 divide when it comes to tablets. Either x86 tablets take off which is mostly a win for Intel or x86 tablets flop and AMD loses.

AMD won the current generation of consoles, but we all know they won't be designing the next generation until 2020+ and the sales will fall long before that. It gives them a reprieve but it's not a permanent solution and their "traditional" CPU/APU business is failing fast, but I'm not sure how you'd fix it. I mean they could go all in against Intel again but that's been a tough nut to crack. Or they could aim for ARM's territory just like all the established players and with Intel gunning for the same too. Or they could try going sideways into some kind of specialized computing, but really all that have tried have been crushed by commodity hardware. I don't think AMD has many good options left, they've got enemies closing in from all sides.

Re:wrong

[Osgeld]

I have a FX4170 and its not the fastest chip in the barn, but its a damn fast lower midrange chip that is plenty quick for everything I do, including current games, what made me go with intel for the next machine (my wifes) was the obscene amount of power these thing suck down compared to the i7

I dont really need faster if its going to jackhammer a 700 watt power supply

Re:wrong

[Osgeld]

using CPU speed to tell you how fast it is, is like using horsepower to tell you how fast a car is

and if you honestly think software is not much more capable than it was 15 years ago then you use nothing that requires a faster CPU

Re:wrong

[Darinbob]

Because software keeps expanding to consume all available resources. Really, much of the software we use is crap but no one cares because the computers are fast enough and big enough that it feels roughly the same speed as the older computers (ie, Word 2013 feels no faster than than Word 95).

Re:wrong

[Darinbob]

Anything not a PC is a "niche" market to some, however those niche markets combined are vastly larger than the PC market and use vastly more CPUs.

That said, there are existing quality FPGA+CPU chips on the market that are popular, so that getting in today as an "also ran" newcomer is a poor idea.

Re:wrong

[Bengie]

Lnyx doesn't seem to load web pages any faster than it did 15 years ago on a 56k.

Re:wrong

[OneAhead]

In the ARM market, AMD is nothing special. They don't have their own fabs, they don't have particularly much experience with ultra-low power designs.

I would, however, consider it within the realm of the possible that AMD were to release a processor that uses the ARM instruction set but is meant to compete in the laptop (not tablet) market and/or the (currently) niche market of Linux servers using ARM. I'd imagine the thing architecturally having a lot in common with their x86-64 offering (except of course things such as instruction decoder), with power draw and speed both far above the tablet market, but with a performance/W that is very attractive for the laptop/server market (maybe even low enough power for set top boxes or high enough performance for consoles - perhaps even a crossover of both). It could make the ARM server niche grow quite a lot, given that its popularity is limited by poor single-thread performance, which the hypothetical AMD part would cure.

Of course, this is all wild speculation, and AMD has been disappointing on its promises of late, but they must be up to something...

Re:wrong

[SemiChemE]

Sounds like what you really want are NVIDIA CUDA on CAPI: http://www.prnewswire.com/news...

... Only on Power8 from IBM.

Power 8 gives:
30% higher single-thread FPU performance compared to x86 at the same clock speed (based on SPECfp_rate2006)
50% higher single-thread integer performance compared to x86 at the same clock speed (based on SPECint_rate2006)

Plus the Power 8 clock speed will scale up to 25% higher for even more performance. Add in significantly more L3 cache, 8 threads/core (compared to 2/core for x86) and up to 12 cores/chip. IBM has a monster on their hands. It will never find it's way into a laptop, but for servers and HPC applications, it's game on.

Re:wrong

[Anonymous Coward]

What AMD should consider are FPGAs and different power cores on the same die. This isn't anything new, but done right, it can go a long way in the server room.

The FPGAs can be used for almost anything. Need a virtual CPU for AES array shifting? Got it. Need something specialized for FFT work? Easy said, say done. Different power utilization cores would be ideal for a server room where most of the hosts see peak load use, then after quitting time, end up idle.

They've been kicking that idea around for 20 years without much success. The average code monkey doesn't have what it takes to get any use out of it.

Re:wrong

[Anonymous Coward]

Very true. IMHO, there is going to come a time when diminishing returns on CPU speed (we already hit a wall with clock speed) are going to mean going back to actual software engineering to make the most out of a machine and demands. So, it might be that FPGAs might be what is needed several generations from now. Right now the H-1B fresh off the boat won't be able to do much with it, but once you can't just keep throwing hardware at a problem, it will take actual engineering (not "engineering" as in banal coding, but actual design and old school hacking) to keep improving.

Re:wrong

[smash]

Try encoding high def video.

Re:wrong

[smash]

No, 50% improvement is nowhere near what we "want". Order of magnitude improvements are what people really want/need.

Re:wrong

[smash]

Nah, it's more like using engine capacity as a measure of speed... horsepower, when combined with weight can be a rough indication. Engine capacity, not so much. It very much depends on the efficiency of the engine.

Re:wrong

[viperidaenz]

There's also already a laptop with an FPGA
http://www.bunniestudios.com/
It's still a niche market
It's something cool to have that's only going to appeal to a tiny share of the market.
That FPGA on that laptop costs more than the SoC and RAM combined

Re:wrong

[jbolden]

Certainly CPUs are somewhat more efficient. But raw speed matters. If CPUs were 15ghz today plus all the other improvements software would be a lot different. And that's what the 15 years before these were like.

Re:wrong

[Osgeld]

I benchmarked a 2.25 ghz amd barton vs a 2.5ghz i5 and the difference was much more than huge, it almost obliterated the barton off the scale

thats a 2004 cpu vs a 2011 cpu, raw speed is not anywhere close as a performance measurement

CPU Rating:
AMD A-XP 2.25GHZx1 517 (year 2005)
AMD Athlon II 2.5GHZx2 1342 (year 2009)
Intel i5 2.5GHZx2 1752 (year 2011)

Re:wrong

[jbolden]

That's a triple in 6 years. There were times in 6 years we were looking at 16x performance improvement or better.

Re:wrong

[jp10558]

4k and 8k have objective benefits

Mind telling us what those are? I imagine ever bigger screens perhaps, but I feel like many of the computer tech (not mobile yet) has reached "good enough" for the mass market (non-niche) that very few people are going to be motivated to replace already working devices for "better resolution" at this point. Certainly not at the price points on offer.

Also, as is constantly pointed out, in the United States anyway, the infrastructure costs kill many higher bandwidth options. Even TV via satellite or cable is compressed, to the extent that HD doesn't look HD in many reports. Netflix probably isn't going to be able to stream 8K anytime soon, either due to unsustainable "agreement" costs to even try and pass that over the last mile networks, or the fact that unless Netflix is building their own FTTH - the network bandwidth just physically isn't there, and in many cases it seems like it will never be (in any reasonable time frame).

Beyond that, most connectors I'm aware of (though better must exist given MBP retina displays) seem to max out at 2650 x 1600.

Anyway, all that aside, it seems like much of the entertainment market and non-mobile computing market, and heck, mobile too is about bringing prices down, not increasing resolution or functionality. At least, that's what people I talk to talk about - making stuff cheaper. They already do everything they want and then some.

Re:wrong

[Osgeld]

sigh there's just not pleasing people, yea ok 16x of damn near jack shit is much better than 3x of a lot

Re:wrong

[Anonymous Coward]

Oh, perhaps you didn't look at the article I linked. Here, let me quote it:

"NVIDIA GPU Accelerators – NVIDIA is adding CUDA software support for NVIDIA GPUs with IBM POWER CPUs. IBM and NVIDIA are demonstrating the first GPU accelerator framework for Java, showing an order of magnitude performance improvement on Hadoop Analytics applications compared to a CPU-only implementation. NVIDIA will offer its NVLink high-speed GPU interconnect as a licensed technology to OpenPOWER Foundation members."

You want "Order of magnitude improvements", you got it!

The 50% performance enhancements I mentioned are what Power 8 provides alone (without the NVIDIA GPU or CUDA on CAPI as I called it). Sorry for the confusion.

Best low-cost CPU with half-decent GPU?

[ArcadeMan]

I'm looking at the new Intel G3240 with Intel HD 4000 and I was wondering if something around the same price range (70$CAD) from AMD had an equivalent CPU with a better GPU.

Re:Best low-cost CPU with half-decent GPU?

[ArcadeMan]

I'm not even sure the G3240 comes with the HD 4000 because Intel makes it near impossible to know which GPU is used inside a lot of their CPUs, listing only "Intel HD".

Re:Best low-cost CPU with half-decent GPU?

[jandrese]

I don't know about your case specifically, but the rule of thumb is that for low to mid range stuff you can get an AMD solution for about the same price that is going to have a slower CPU and faster GPU. It's pretty easy to beat a HD4000 GPU in any case. Of course this shoehorns you in a bit. If you went with the Intel solution, then you could drop a discrete GPU in later. If you go with a CPU that is too slow you often have to change more of the base system to upgrade (memory and motherboard).

Re:Best low-cost CPU with half-decent GPU?

[washu_k]

No, there really isn't an equivalent. Which is more important, CPU power or GPU power?

The closest AMD in price with a GPU is the A6-6400K. It would be quite a bit better in the GPU department, but MASSIVELY worse in the CPU department. Not even close in CPU power. To get something that wont cripple you on CPU you would need to go up to the A8-6600K, but that is over $110 at the CAD stores I checked and would still be way worse in single thread CPU.

There are also the new Kabini CPUs and the top end of those, the Athlon 5350, is around $70. It would save you money on the MB (AM1 boards are cheap), but would be even worse than the A6-6400K in CPU and might not even match the G3240 in GPU.

Re:Best low-cost CPU with half-decent GPU?

[ArcadeMan]

(reply for both jandrese and washu_k)

Thank you for your comments. I guess I'll go with the G3240 since it's a better CPU, endure the Intel HD GPU for now and add a GTX 750TI later.

Re:Best low-cost CPU with half-decent GPU?

[ChrisSlicks]

AMD A6-5400K. 3.6GHz (3.8 Turbo) and Radeon HD 7540D. $65

As others have said it is a slower processor than the intel but with faster graphics. The AMD only gets a 2100 CPU Mark (Passmark software) which is about the same as an old Phenom II X2 or a few year old Intel i3 mobile chip.

I assembled it as a low-end system for a parent that basically does email and web surfing along with some basic image editing and cheesy games.

Re:Best low-cost CPU with half-decent GPU?

[Anonymous Coward]

It would also depend on whether you felt it was OK to overclock. For roughly the same price, you can get an unlocked FM2 processor and overclock the X86 cores. Since the G3240 isn't hyperthreaded and the CPU clock is locked, you get the benefit of being able to run 4 threads at a higher clock speed.

To be completely honest, the performance difference at this level is minimal. Depending on the level of work you are putting the computer through, you are barely goign to notice the difference if all you plan on doing is surfing the web, checking email, working on a couple office documents or writing a couple scripts. And the money you didn't spend on an Intel CPU can go into more memory, storage, or even a better discrete GPU.

Re:Best low-cost CPU with half-decent GPU?

[Anonymous Coward]

The A6-6400K is the rough equivalent. On paper it will look faster (higher clock rate), but in practice the CPU is about the same speed. The A-series parts almost always have better GPUs than their intel equivalents though.

The part numbers don't make much sense for end users though. The 'upgrades' to the A6-6400K are the A8-6500, A8-6600K, and A10-anything that doesn't end with a T.
As a general rule the first number after the A tells you how good the GPU is (higher means more cores on the GPU, the A6 parts are 192, A8 is 256, A10 is 392) and the second part (the -6400K) tells you how fast the CPU is. K parts are unlocked (overclockable), T parts are low-power, so slower on both the cores and memory bus. For gaming, avoid the Ts at all costs - the slower memory will bottleneck the GPU even if the CPU doesn't.

The catch is the AMD parts eat more power than the Intel ones overall. If you're only going to use the integrated GPU, then that's not so much of a problem, but if you're going to pair it up with a dedicated GPU you'll be better off with intel.

Re:Best low-cost CPU with half-decent GPU?

[aliquis]

I won't bother with checking specific models, you do that if that interest you.

Normally the AMD parts has had slower CPU and better GPU but some/one(?) of the very latest Intel processors with integrated GPU seemed to be about equivalent with the best AMD offerings also on the graphics side.

I don't remember what it was called but it may have been within the latest 47xx/48xx series or something such and as such maybe in a different price range than what you're talking about.

I would only be watching the A10 7850 or those Intel chips with similar performance but it's still not very competitive with a "medium enthusiast" graphics card so I would likely not settle for either. But that's me.

Re:Best low-cost CPU with half-decent GPU?

[Anonymous Coward]

It doesn't, this shows you exactly what's in it, while it is a haskwell GPU it has half the EUs of a HD4000, and is clocked much slower for less than half the performance of a better CPU: http://en.wikipedia.org/wiki/Intel_HD_and_Iris_Graphics#Haswell

Re:Best low-cost CPU with half-decent GPU?

[Anonymous Coward]

Your rule of thumb is stupid. I've been running a Phenom II X6 1090T Black Edition for years, have only upgraded the graphics once, and I can still run anything that drops at max settings without the slightest problem.

The good rule of thumb is to disregard benchmarks, because you aren't playing fucking benchmarks. Yes, the Intel CPUs are more efficient, and score higher on benchmarks optimized for their products, but the question is where that higher performance will pay off. If you're already committed to dropping in a decent graphics solution, and you aren't planning to build in a bottleneck like a slow HDD, then your choice of current-gen CPU is largely inconsequential -- you aren't going to be running it at peak anyway, unless you're running multi-threaded encodes or something, and even then you'll do better offloading that to the GPU.

Hmm, 1 2?

[Anonymous Coward]

With openpower systems coming with such great performance numbers from IBM. AMD jumping back hard into the arena at this time might be good timing. ARM/Mobile is growning everything else is shrinking at this point. Its going to be an interesting couple years with the Tegra chips in there.

Compaq was afraid to use AMD chips FOR FREE

[Kartu]

Compaq was afraid to use AMD chips given out for free, because Intel would "retaliate", ok?
What kept AMD's market share low was not "clever marketing" of its competitor, it's crime.

Back in P4 Prescott times, Intel's more expensive, more power hungry, yet slower chip outsold AMD's 3 or 4 to 1.
Not being able to profit even when having superior products, it's really astonishing, to see AMD still afloat.

Re:Compaq was afraid to use AMD chips FOR FREE

[Kjella]

While Intel did a lot of shady things, part of it was also that AMD didn't have nearly enough fab capacity to supply the market. Those kinds of decisions are made years in advance, you don't just pop up a sub-100nm processing plant on demand. So AMD got a huge winner, they surely produced everything they could and got a nice premium on their products but the remaining demand had to go with Intel. It's just not the sort of market battle you can win quickly.

Re:Compaq was afraid to use AMD chips FOR FREE

[phantomfive]

Compaq was afraid to use AMD chips given out for free, because Intel would "retaliate", ok?

Let's be honest though, Compaq wasn't known for making good business decisions.

Re:Compaq was afraid to use AMD chips FOR FREE

[nbritton]

it's really astonishing, to see AMD still afloat.

No, I think Intel keeps them around to get around anti-trust regulators. AMD should get into more consumer markets. For instance, I use Macs, but I'd like to give AMD some of my money so they can keep the competition up with Intel, because ultimately this will mean better Intel chips for my Mac.

No kidding

[Sycraft-fu]

I would -love- to see AMD truly competitive with Intel on every level because it is only good for us consumers. It would be great if both companies made chips so fast, efficient, stable, and capable that you didn't buy AMD or Intel based on anything but who had the better deal that week.

However I'm not interested in hype and bullshit. As you say, "put up or shut up." I get tired of hearing about how great your shit will be in the future. Guess what? Intel's shit will be great in the future too, probably. It is great right now.

So less with the hype, more with the making a good CPU.

Re:No kidding

[evilviper]

I get tired of hearing about how great your shit will be in the future.

Reading a tech news site may not be for you...

Or at least, don't click through and read EVERY story, including the ones that infuriate you...

Re:No kidding

[Anonymous Coward]

What would be great is if when we buy the CPU that is the best deal that week is if we didn't have to purchase a special motherboard just for that chipset. What the hell is so hard about making CPU chips have a standard socket? Seriously just over engineer it to have more than what they really need then get Intel and AMD to make chips with that socket/pinset.

This is what baffles me about the enthusiast pc market. You can buy motherboard X which works with AMD chipset socket XXXX. Then you have motherboard Y that uses the Intel Xbridge chipset. It seems like it would be advantageous to everyone to standardize it and update the standard when needed. That way motherboard manufacturers would be able to save money and it would be much more simple and cheaper for consumers. I think that I'm right when I say that 0s and 1s go into a CPU and math happens and 0s and 1s come back out right? So why can't they standardize the architecture that transports those 0s and 1s around the motherboard?

I know that they the companies optimize the chipsets to work with their CPU but if there was a standard why couldn't both be optimized to work on that chipset. Would it be that expensive for the Intel and AMD to do that? Is it just because that's the way they have always done it and they are going to keep doing that? Is it because they want bragging rights on whose chipset moves the 0s and 1s around the fastest? It seems like a waste of resources to keep developing them this way.

Re:No kidding

[Anonymous Coward]

Consumer mindset has to change before "who has the better deal this week" determines sales volume. That didn't determine sales in the past, and doesn't now, and I don't see it changing just because "it would be great". Consumers don't switch chip choice on a dime, the moment one company creates the better product. It takes a long time to build up consumer readiness. Whether in the form of news, unbiased opinion, or company PR, it may be an unfortunate necessity that consumers are buttered up and ready to embrace new chips well in advance, because it will take a long time to get them ready and it may be too late if that starts when the chip is already available.

target foot acquired!

[slashmydots]

Way to shoot yourself in the foot, AMD. I don't want or need a new architecture. I want x86 (and x64) for my PC and laptop, the end. In 2016, you'll have some stupid new chip that won't work in any PC or laptop in the world just in time for people to realize that tablets are netbooks with no keyboard and a higher failure rate and stop buying them. I don't think they will make a 1/8 watt cell phone chip when they're this far behind so they're making a completely useless product that nobody will use in anything.

Re:target foot acquired!

[NotInHere]

Intel tried to obsolete x86 with its IA-64 architecture, and they failed. With old architectures its like with old software: lots of backwards compatibility, and a huge mess. And x86 /is/ old. x86 still has features from the 1970 8080 chip. It is time for an architecture to success x86, and remove that clutter. However, with a new architecture, AMD takes a high risk. The PC success of their architecture depends on whether microsoft wants to support three architectures or not.

Re:target foot acquired!

[viperidaenz]

Windows NT used to run on DEC Alpha.

Re:target foot acquired!

[oh_my_080980980]

Not anymore :)

Re:target foot acquired!

[Guy Harris]

Way to shoot yourself in the foot, AMD. I don't want or need a new architecture. I want x86 (and x64) for my PC and laptop, the end.

Another reason to avoid the unqualified term "architecture" when speaking either of instruction sets or chip designs; person A may read "architecture" as "instruction set architecture" and person B may read it as "microarchitecture". I suspect they're talking about a new microarchitecture, implementing the x86-64 instruction set architecture, here.

Re:target foot acquired!

[viperidaenz]

Windows CE runs on ARM, x86 and MIPS. Probably more too.

Re:target foot acquired!

[smash]

I think you over-estimate the clutter required to support old 1970s instructions in an x64 chip. They have billions of transistors and an instruction decoder. The entire 486 CPU (the last CPU with the instructions not decoupled from the internal design) was implemented in 1 million transistors or so. Less than 0.1 percent of a typical modern x64 CPU if it was implemented in hardware, and it is most likely NOT implemented in hardware, but microcode...

Re:target foot acquired!

[NotInHere]

You are right that the "original" instructions may take only a small part of the chip's microcode memory. But if you have to comply not just with a 1970s chip but also with last year's chips, you can't make such big changes as EPIC in a "native" x86 chip without a huge effort.

AMD lost money in 2013 and 2014Q1

[Anonymous Coward]

now AMD is flush with cash from its profitable business with gaming consoles

And that cash is going to other parts of the business to offset their loses. AMD lost $83M last year and $20M last quarter. At this point, the only part of the busines that is viable is the graphics division; they just need better drivers. AMD is grasping for straws at this point. Personally, I think their should ditch their x86 products, it is dragging them down.

Re:AMD lost money in 2013 and 2014Q1

[viperidaenz]

Why do they need to make better drivers? Their graphics division makes all its money selling GPU's to Microsoft and Sony. No driver issues there.

ha ha

[p51d007]

I remember when the AMD K6 came out. "AMD is preparing to give Intel's pentium a run for its money with the new K6".

Re:ha ha

[Anonymous Coward]

Except K6 was competing (and in some benchmarks beating) the Pentium II, not the Pentium.

And with the K7 (Athlon), AMD truly wiped the floor with the Pentium III. Intel didn't manage to get back on top until the Core2.

Re:ha ha

[viperidaenz]

K6, the Celeron killer!

Re:ha ha

[drinkypoo]

The original K6 was a bit of a turd, because of its 24-bit FPU. That was a horrible, terrible mistake.

The K6/2 was a peach of a processor, clock for clock it will beat a Pentium II at many operations not least because the low-end P2s that it was competing with had crippled cache. Sadly, most of the motherboards it was coupled with were pure shit.

By the time that was fixed and the K6/3 came out with onboard L2 cache (and the motherboard cache, if any, became L3) it was too little, too late. The K6 was known as a failure.

Then the K7 came out and it beat the living crap out of the P3 dollar for dollar, watt for watt, clock for clock, any way you wanted to measure. And then it went on to beat the P4 on all the same bases. I won't count AMD out yet.

Re:ha ha

[Red_Chaos1]

I've been in the AMD camp for ages, but I have to admit that even the K6-2 was not all that. It was decent, but it still suffered a weak FPU, which became apparent when MP3 made its big hit.

Re:ha ha

[drinkypoo]

I've been in the AMD camp for ages, but I have to admit that even the K6-2 was not all that. It was decent, but it still suffered a weak FPU, which became apparent when MP3 made its big hit.

MP3 doesn't tax the K6. What does is 3D gaming, which was just becoming a major thing at that time, and which is pretty much all fp math. The K6 had badass int performance, but not so much fp. The K7 had great fp.

Re:ha ha

[Red_Chaos1]

You could argue it's all anecdotal, but the evidence says you're incorrect. A similarly clocked PII would encode faster than the AMD counterpart. Same for video, gaming, etc. It's not like it was terrible, but it wasn't as good as its opponent either.

Re:ha ha

[drinkypoo]

You could argue it's all anecdotal, but the evidence says you're incorrect. A similarly clocked PII would encode faster than the AMD counterpart

Sure, but it's not like it took aeons to do on the AMD. But what really didn't work was 3d gaming. It kinda worked, for a while. But at first the 24-bit FPU caused problems which required AMD-specific patches which were often an afterthought, and later the lack of enough FPU for 3d gaming caused a decisive shift away from the K6. If your mp3 encode takes longer, so what? You're encoding from a batch, right? But if your CPU doesn't have enough FPU to play a game at a decent resolution, it just doesn't have enough.

Re:ha ha

[aliquis]

Yeah I don't either get the people who say K6 to Phenom.

Those where processors with a touch of "unless" and "if" attached to their value and performance.

Athlon-XP and Athlon64 was nice but that's kinda it.

K6 II vs Pentium MMX for gaming and you'd still lose out.

And sure AMD chips had lots of cores later but the performance / core sucked, most games and such wasn't multi-threaded and one very fast core is (imho at least..) nicer to have than multiple slower ones. For whatever reason I haven't really understood some people seem to think lots of threads is good now, maybe it's a latency thing?

Re:ha ha

[smash]

And it was still inferior to the Pentium II, because AMD (literally) couldn't make an 8087 compatible GPU to save their life until the Athlon series.

Re:ha ha

[smash]

The athlon was a hit because AMD had bought a huge amount of technology and know-how from Alpha personnel. There is no DEC Alpha to borrow technology from this time around.

Re:ha ha

[smash]

So what if I'm encoding from a batch? If its going to take 24 hours on the K6 or 12 (or 18, or 22 - whatever) hours on the Pentium 3, it still means I'm waiting longer to be able to do anything with the end result.

Just like Bulldozer?

[Anonymous Coward]

Don't like news about what tech will be like in 2 years? Then why do you read slashdot?

LOL

[Anonymous Coward]

Intel has so much tech on the roadmap, anything AMD were to come out with, Intel would just push stuff to come out sooner that would immediately be better.

AMD should just stop bothering with processors...

Re:LOL

[oh_my_080980980]

Yeah douche bags learn to post.

Re:LOL

[Anonymous Coward]

LORD DOUCHBAG

Yay!

[buggsdummy]

Can't wait for a new day to dawn for AMD!

SecureBoot

[Anonymous Coward]

It better have a configurable secureboot feature, because ARM has the real nightmare of TCPA/TC through the use of a non-configurable secureboot.

2016

[Anonymous Coward]

By 2016 AMD will have a CPU that beats the sh*t out of Intel's 2014 best offerings.

Re:2016

[rrohbeck]

That would be good since Intel's CPUs are only getting 10% faster each generation.

Re:2016

[CajunArson]

Intel gets 10% faster each generation?

SOMEBODY SLOW THEM DOWN WE CAN'T KEEP UP WITH THAT PACE!
      -- AMD Management (where odd-numbered generations get slower and even numbered generations just look good because of the odd numbered ones)

Re:2016

[smash]

Newsflash: lack of competitor sees intel hold off on new technology release!

Run

[Anonymous Coward]

The only running Intel is gonna do is to the bank with armfuls of money.

Let there be an FX-8550

[Anonymous Coward]

I hope they will release a Steamroller based FX series until that, many would like to put their hands on those improvements. No one really expects them to be present in the high-end market, but we sure want them to remain strong in the medium segment.
http://semiaccurate.com/2014/01/23/kaveri-versus-richland-performance-per-clock-comparison/
http://www.extremetech.com/computing/177099-secrets-of-steamroller-digging-deep-into-amds-next-gen-core

Free? Keep in mind they'd lose Intel Payola

[rsborg]

Compaq was afraid to use AMD chips given out for free, because Intel would "retaliate", ok?
What kept AMD's market share low was not "clever marketing" of its competitor, it's crime.

Back in P4 Prescott times, Intel's more expensive, more power hungry, yet slower chip outsold AMD's 3 or 4 to 1.
Not being able to profit even when having superior products, it's really astonishing, to see AMD still afloat.

Intel's Payola [1] (which basically kept Dell profitable for several quarters of the past decade) is something you have to factor in when looking at these "deals". I'm just sad that Intel didn't pay a bigger price for their purely anticompetitive corrupt practices.

[1] http://www.theatlantic.com/tec...

wrong

[Anonymous Coward]

They already work on mobile chips, they also want to be back on par with Intel on the desktop and server markets. Not everybody needs a server or workstation but the market isn't going away and the cash Intel makes is a proof that the market is still relevant. Let's hope AMD succeed, so that there is again competition in this market segment.

To be fair to Intel

[Sycraft-fu]

Netburst did seem like a reasonable idea, in testing. While it was low IPC, it looked like it would scale bigtime in the speed area. They had test ALUs running at 10GHz.

So I can see the logic: You make an architecture that can scale to high frequencies easily, and that gets you the speed.

Obviously it didn't scale, and wasn't a good idea, but I can see what they were going for. It wasn't like it was completely nuts.

Re:To be fair to Intel

[Grishnakh]

Well they completely forgot that higher clock frequencies translate to higher power consumption and higher thermal dissipation, and you can only remove heat from a chip in a consumer computer so fast since resorting to submerging it in coolant obviously isn't feasible for a consumer device. I remember some Intel presentations (I worked there at the time) where they were proudly bragging about how much power these chips consumed. Did they not think people might not like their computers generating that much heat just to do some word processing, and jacking up their electricity bills?

Re:To be fair to Intel

[Anonymous Coward]

Yeah, if you kept that pipeline full it ran like a raped ape. They tried it out and it didn't work. Hindsight is 20/20 but at the time it was a perfectly cromulant approach.

Re:To be fair to Intel

[jandrese]

It's one of those things where it makes amazing scores on benchmarks, but struggles in the real world.

The longtime complaint about P4s is that they were a marketing ploy. People buy the chips with the most Hz, so we're going to give them the biggest numbers ever! Oh, and they'll look amazing in benchmarks. But then AMD's marketing guys came up with the crazy notion of just making up numbers to compare against Intel's mostly meaningless Ghz number and the marketing advantage never materialized. Plus sites got really into doing "real life" benchmarks where they ran real life applications with scripted inputs instead of synthetic benchmarks.

Re:To be fair to Intel

[Anonymous Coward]

Well they completely forgot that higher clock frequencies translate to higher power consumption and higher thermal dissipation

No, they did not forget this. I'm sure you "worked at Intel", but that doesn't mean you have any insight into what the designers were thinking, particularly at high levels.

Bob Colwell was Intel's chief architect for x86 projects from the Pentium Pro through about one or two generations of Pentium 4 (there are substantial microarchitectural differences between generations of P4). I've read some of his comments re: P4. He said that he felt a bit conflicted about the P4's direction even when they were doing it, but the main reason it turned out badly was something which wasn't totally clear when the project began.

P4 originated back in the semiconductor industry's perfect-scaling days, when every process shrink delivered predictable gains in several parameters: transistor switching frequency, transistor switching power, area per transistor, and so forth. The first Pentium 4 generation was built on Intel's 180nm node, which turned out to be the last node which delivered such gains. Thereafter, at 130nm and below, the industry went through the leakage power crisis, aka the "power wall", which (among other things) played merry hell with a lot of the core assumptions about how the P4 microarchitecture was going to scale on new process nodes. Suddenly, if you wanted to push frequency, you were doomed to ridiculously high power consumption levels, unless you kept the amount of high frequency circuitry very small. But P4 was both high frequency and a huge, complicated chip for its time...

Re:To be fair to Intel

[smash]

Yup. What many people forget is that netburst was intended to scale way, way beyond the clock-rate it was released at. Intel fully expected us to be running 10Ghz on silicon (and beyond) by 2004, so they optimized the Netburst architecture to run at that rate. It wasn't low IPC because it was a bad design, it was designed to be low IPC to enable it to scale to ridiculous clock-speed that was expected to be possible (because at the time, the P6 based Pentium 3 had pretty much hit a wall around 1-1.3 Ghz).

The current situation, where we have stagnated around 3Ghz or so for the past decade was unexpected, and not predicted to be a scaling barrier as manufacturing process shrunk.

Netburst may actually end up making a comeback if we move to a different CPU technology to silicon, and are able to ramp up clock-speeds to the level where it makes sense.

Re:To be fair to Intel

[smash]

They didn't "forget" this, because the extent to which this was a problem beyond 3Ghz was not yet known when Netburst was under development.

Re:To be fair to Intel

[jmv]

The boneheaded part was not realizing that clock speed was about to stop increasing very very soon.

Re:To be fair to Intel

[Grishnakh]

It didn't help that they explicitly designed the P4 to work with RAMBUS memory, which was single-source and hideously expensive, and had very high latency. That was a rather stupid decision, as lots of customers switched over to AMD with its much cheaper SDRAM memory because of the prices.

Re:To be fair to Intel

[Anonymous Coward]

And now AMD went down the same dead end... Ironic.

Lol AMD

[Anonymous Coward]

Every.. Single.. AMD System that I have seen has blown caps or a bad power supply. Every.. Single.. One..

My Intel systems at work that were purchased at the SAME time from the same vendor are still running today. After my old Athlon 1.3XP back in 03 started eating power supplies and caps, I bailed on the AMD bandwagon. Way too power hungry for the processing power. Intels just seem to run "smoother" also. Just my .02c AND I don't have computers that sound like leaf-blowers or shop-vacs in my house anymore! Bonus!

Re:Lol AMD

[Anonymous Coward]

Then you should blame the mobo and power supply manufacturers that made the things that blew up. What does AMD have to do with this? I can tell you that ever since I bought an AMD processor, I have never been attacked by tigers. Not once!

Re:Lol AMD

[Anonymous Coward]

Stop buying cheap components.

Re:Lol AMD

[smash]

No, this isn't actually irrelevant. It is the primary reason I abandoned non-intel systems actually. Not specifically blown caps, but simply because there are so many shitty motherboard chipsets out there for alternative CPUs.

CPU isn't so much what drives my machine purchases since about 1996 - it's motherboard chipset: driver support, bugs, compatibility, etc.

Not to say that intel chipsets never have bugs, but it is more likely that the software I run has put in sufficient work-arounds and testing is done that I end up with a stable system. Yes, this in anecdotal, but since I've gone exclusively intel for my motherboard chipsets in 1997, i can count the number of non-RAM related blue screens I have had on one hand (and these were Nvidia GPU driver related).

Re:Lol AMD

[Sardaukar86]

Very scientific.

My advice is to send your failed mobo and PSU back to AMD, I'm sure they'll be very keen to rectify the problem with their processors that are causing capacitors to fail in your system.

Re:Lol AMD

[Anonymous Coward]

Hahahahaha http://news.slashdot.org/comme...

Re:Lol AMD

[Sardaukar86]

Wow APK, I'm really touched that you're watching my postings so closely, ready to stalk at a moment's notice. Just the 12-year-old mentality I expect from you.

Readers: sorry about this off-topic response, I'll keep it brief:

I can take your trolling, APK, I'm a big boy. However I ask you to reel in your behaviour because the rest of Slashdot doesn't need to be punished with your off-topic stalking. I'm not going to respond to your tantrums, so have a heart for the rest of the site, APK! People aren't here to read your posts, so don't try to force them to, please? To do otherwise is just selfish.

Re:Lol AMD

[Anonymous Coward]

Cool story bro.
If only there was such a thing as a 1.3 GHz Athlon XP...

Re:Lol AMD

[Anonymous Coward]

Whatever Wladimir

Re:Lol AMD

[Anonymous Coward]

Yeah, you're full of crap. I worked a beige box outlet from the late 1990s through the mid 2000s, right in the middle of the "blown caps" era. I built, quite literally, thousands of machines, and serviced thousands more; the blown-caps issue was universal. There were a lot of people who, like you, wanted to blame it and the Western Digital HDD failure streak on AMD, because they both directly followed the original K6 CPU problems, but I'm here to tell you that Intel motherboards -- all of them -- had the same issues.

Need to reconsider your hardware news sites...

[Pollux]

According to xbitlabs, Kaveri has worse CPU performance than its predecessor.

AMD got lucky. It's found a dependable stream of revenue in game consoles. Better yet, no matter whether Microsoft or Sony wins the next generation console wars, both have AMD under the hood. Now that's hedging your bets. Whoever at AMD was in charge of negotiating these deals deserves a paid vacation to Necker Island with all the trimmings.

But lets get serious. AMD's current processors suck. And I hate saying that. A decade ago, AMD was the hero in the processor wars. If it wasn't for AMD, we'd be stuck with Rambus RAM, using Itanium processors, and have PCs running so hot we could cook breakfast on the case. But AMD's desktop processors are inefficient, almost two generations in fab technology behind Intel, and just cannot compete at any level.

Unlike 10-12 years ago, Intel's making great strides in microprocessor technology. It is thanks to AMD's competitiveness that Intel finally got its act together, and for that, I will always be thankful. If they can find a way to improve on Intel's product line, I'd be amazed at their comback. But do they really need to?

Fight the good fight

[Dega704]

One could argue that the reason Intel's products have advanced as far as they have is because AMD was there to keep them on their toes. The game has changed since then, with mobile and whatnot, but I am still rooting for a comeback. Rory Read has played his cards well so far; and with Jim Keller back, it will be interesting to see what they have in store for us.

I'll believe it when I see it.

[Chas]

AMD has pissed away massive leads over Intel in the past.

AMD single-handedly created the x86-x64 market from NOTHING.

Then they fell back on their laurels.

Then they bought a graphics company.

Their last effort in the market was basically a fizzle. Forgoing a custom chip designed to eake the maximum efficiency and power from the device, they went with a crappy computer-designed monstrosity that basically was the worst of all worlds, and a flame-throwing power hog to boot.

Sure, they can kick out a processor that says "I can throw *insert a number here* cores at you!*
Are they FULLY FUNCTIONAL cores? And are users actually going to be using anything that can take advantage of the massively multiple architecture?
The problem is that AMD doesn't know and has long since stopped caring.

So, until they actually deliver a complete platform that can top Intel, I'm just going to ignore them. They're not worth dealing with otherwise.

Re:I'll believe it when I see it.

[serviscope_minor]

AMD has pissed away massive leads over Intel in the past.

If by "pissed away", you mean suffered awful monopoly abuse and illegal actions from Intel then yeah sure. Pissed away.

So, until they actually deliver a complete platform that can top Intel, I'm just going to ignore them. They're not worth dealing with otherwise.

The 4 socket servers are still very competitive.

The APUs are certainly interesting. More HSA ench marks have started to creep out. The resutls are promising in that when it works it seems to destroy the intel chips performance wise.

Don't bother

[Anonymous Coward]

The article is just a lot of vague and mostly clueless ramblings from a "freelance journalist" who clearly a) has no insider information from AMD and b) knows nothing about CPU design (as his comments on the 2 INT + 1 FP approach make perfectly clear).

I wonder how much he gets paid for that; I bet I could write an automated "article generation" script that would produce more relevant articles.

SMP?

[Overzeetop]

It would be nice to see AMD offer a 4-8 processor chipset that would allow you to highly parallelize their chips. Intel can do it, but the premium for Xeon silicon is outrageous. Not sure if AMD has enough business in that market that they're willing to chuck it in hopes of getting a leg up, but I sure as hell wish I could drop a second CPU into my desktop so I don't have to chuck the entire thing and buy a whole new board/CPU from Intel just to get a 50% boost in performance every 3-4 years.

Re:SMP?

[Anonymous Coward]

Without a redesign, you can't even power those chips. Imagine the electric bills at 125 - 250 watts per chip!

AMD needs several things:
1. A die shrink
2. A new redesign of their x86 CPUs focused on energy. The FX series is their equivalent to P4 chips. You can cook eggs on those things.
3. A new marketing department. When's the last time you saw a good ad for AMD? They need a marketing budget too
4. Drop the ARM crap. Sure, ARM is cool right now and it will continue to be for some time in phones, but tablets are going Intel.

I've lived through the Sparc is cool, PPC is cool era. ARM is no different. It's awesome but it is just starting to hit the bottlenecks in the architecture. If you compare a windows 8 tablet running arm vs intel, there is a massive difference. Even looking at the cheap ChromeBooks, you can see it. Intel is making headway on power consumption and they have a significant lead on performance. Intel is good at this.. they will win.

I say this as a guy with sparc64 servers, ppc macs and various other goodies in my home. I was a big sparc fan for a long time. I ran macs for a long time. Intel always wins in the end. ARM fans are in a fantasy world.. i think they have embedded for at least 10 years, but computing devices like tablets and maybe phones are going intel.

Re:SMP?

[serviscope_minor]

It would be nice to see AMD offer a 4-8 processor chipset that would allow you to highly parallelize their chips. Intel can do it, but the premium for Xeon silicon is outrageous. Not sure if AMD has enough business in that market that they're willing to chuck it in hopes of getting a leg up

Tough call. Apparently HT is routable so making a 8x system would require a HT router. It would be getting into quite exotic territory. If you've ever opened up a 4x 1U server you'll see almost every inch of the space is used. It's either the PSU, CPUs or RAM slots.

For 8X you're looking at people who need more CPU power but also must have the single system image (i.e. not networked) and don't want to go too far beyond 8x. Anything much further is going to require infiniband interconnects or something.

I'd love one, but I suspect there's a tiny segment since the use cases are somewhat obscure.

Re:SMP?

[Anonymous Coward]

It's awesome but it is just starting to hit the bottlenecks in the architecture.

And of course you're willing to point out what these "bottlenecks in the architecture" are...oh, wait. You aren't.

First to 64-bit

[shizzle]

x86 is holding us back, so much that servers are turning towards ARMv8 (inferior design to Itanium, but tons of momentum from mobile/embedded).

The x86 ISA is not holding us back. IMO, the only thing that motivates people to turn to ARM for servers is that AMD is not giving Intel sufficient competition in the server space. No one wants an Intel monopoly, and if AMD is not going to be an effective alternative, then people are forced to look beyond x86 for one. But that has nothing to do with the relative technical merits of x86 vs. ARM.

Also, no way is ARMv8 and inferior design to Itanium. I think the fate of Itanium should make it clear that there were very few things in this world inferior to Itanium ;-).

the first to dual core

[Anonymous Coward]

The pentium D(netburst) was the first x86 dual core.

Need faster HyperTransport / more buses in all cpu

[Anonymous Coward]

The desktop ones need some pci-e in cpu or at least 2 channel s of HyperTransport so you can have PCIe HyperTransport bridge s

Manual context switching

[tepples]

It won't let you multi-task since you can't reprogram it when ever you context switch.

Mobile devices have tended to work around this by implementing a window management policy of all maximized all the time. If you want to context switch, you'll do so by pressing a dedicated button or dedicated area of the touch screen to pull up a list of contexts to which you can switch.

Re:Manual context switching

[viperidaenz]

That went so well with non-mobile didn't it?

I heard everyone loved Metro.

Input is another key part of the problem

[tepples]

3D gaming is starting to look like just another "Gold plated speaker wire" guy hobby as everyone moves to mobile devices.

Let me know when a substantial number of people start buying MOGA clip-on gamepads for their mobile devices. Until then, even the smartphone or tablet with the strongest CPU and GPU will be limited by its touch input. Mega Man 2 and Castlevania ran comfortably on 1.8 MHz CPUs, yet not even a 1.8 GHz CPU can add buttons to a device that doesn't have them.

For the next next gen

[tepples]

Effectively all the GPU revisions between Xbox One and Xbox Two and between PlayStation 4 and PlayStation 5 will be for PCs. AMD needs to keep its PC GPUs and respective drivers competitive in order to continue to have a leading-edge product. Without a leading-edge product, AMD has no real way to research tech that will keep Xbox Two and PlayStation 5 on AMD instead of switching back to NVIDIA like the original Xbox and PlayStation 3 did.

Re:For the next next gen

[viperidaenz]

But they've come so far with mediocre drivers.
They beat nvidia for both Xbox and PS4

Live encoding

[tepples]

If your mp3 encode takes longer, so what? You're encoding from a batch, right?

Not if you're offering live streams at multiple bitrates. Then you need an encoder faster than real time.

Performance/Price: AMD always wins.

[ReekRend]

Every time I go to buy a CPU, AMD wins on the price-performance charts, so... AMD wins. I have to assume that Intel is for the subset of people for whom money is not a concern.

whoo whoo.. let the games begin

[Anonymous Coward]

Finally,, a new round of aggressive innovation,
I hope AMD kicks the Sh1t out of INTEL AGAIN!!!!!!!
I think Intel has gotten too smug again and their pricing model reflects that..
I have allways liked AMD's approaches to the issues @ hand..More meat to chew on.
a more agreeable "bang" for Buck ration..
Plus I have allways been a FAN of the ALPHA chip innovations..
I find it interesting, that 20+ years later their innovations are still seen in a significant portion of the market..
Is it, the engineer's have gotten to lazy to bring REAL innovation? Have we run out of innovation (steam)? Is the Alpha class chip so far ahead of its time that no current design's come close??
i mean really even GM has moved away from it (bread&butter) Bowtie platform. (the 350ci engine v8).
Moving past that, perhaps I am wrong. But seriously i still see people basing alot of stuff of the EV7 bus/backend, etc...

I am also not omitting AMD in that picture as well they have sliced and diced soem older tech as well.. (ev7)..

bottom line is something NEW has to come from this..
The next horizon, the next big thing..

Lets look FAR FUTURE, what would you like to see in the next REAL USS enterprise???

many things to consider.

thanks for you time, you may remove the soap box...

Never again

[Anonymous Coward]

I have owned AMD processors exactly 3 times, and every single time I have heat issues. It doesn't matter how much Arctic Silver I use, or having a good or better or best heat sink, or having a better case with better airflow, they always get too hot and tank. My last one was the Phenom II 6-core... after less htan a year I was down to 1 core.

I have had ZERO problems with Intel in that regard, and their CPUs perform well under a variety of loads, with no problem.

Re:Never again

[Anonymous Coward]

I have owned AMD processors exactly 3 times, and every single time I have heat issues. It doesn't matter how much Arctic Silver I use.

Maybe your problem stems from not knowing that thermal compounds are actually thermal *insulators*. Putting too much Arctic Silver (or anything else) on a CPU will cause it to run hotter. The only reason it is used is to increase the contact surface area between the heat sink and the cpu by filling in voids.

Mantle

[Anonymous Coward]

A lot of devs in the game industry are also eyeing AMD's mantle, a GPU solution that tends to achieve 40%+ performance over DirectX/OpenGL:
http://en.wikipedia.org/wiki/M...

There was an impressive demo at GDC this year by Oxide:
http://www.maximumpc.com/AMD_M...

I like Intel

[Anonymous Coward]

I like Intel. I'm running their chipsets and chips right now, but competition is also good. Eric Clapton never played as well as when he was trading licks with Skydog (Duane Allman). Mind you, Allman was considered the better player, but in each others presence, all the other strummers just stood around dumbstruck watching two masters jam and going 'how the hell did they do that? and 'I didn't know you could get a guitar to sound like that!' So now we have AMD pressing to give Intel some CPU competition. Good. If we could get either Intel or AMD to give Nvidia competition, that would be even better. You know IBM isn't going to, they have given up on attempts to be innovative in the consumer space.

Is the ARM-hype ebbing already again ?

[Anonymous Coward]

AMD very recently released an embedded quadcore processor, which is available in a Zbox-Nano from Zotac and there is an Intel Mini-ITX board available from ASRock, at a 10W TDP, also with four CPU-cores. It is late of course, given that ARM-CPUs are available already as octacore versions following the big-little design.
Nonetheless embedded PCs have much more attractive prices than high-end ARM-machinery. So unless you are enthusiastic about ARM, or you are paid by Google or Apple, there is no need anymore to buy ARM, it may also cost more than 10x as much as a comparable Intel-compatible product.
ARM-CPUs can run at 2 GHz or even higher frequencies, but their energy-efficiency is not necessarily better than of PC-architecture chips.
Very recently I upgraded from the AM2+-platform to FM2 and now I am wondering, whether single-thread-performance really _is_ the holy grail of today's desktop-computing, respectively how long it is going to remain this way, if so?
Linux-Software is not properly parallelized yet, so 2x3 GHz will be faster than 3x2 GHz in more than 93% of all use-cases.
So I guess now, that either the software-producers have to do something in order to make better use of multicore-designs, or it is the hardware-vendor's turn to deliver the technology to us, that automatically makes proper use of multicore-CPUs, also with single-threaded software, most probably both.
The PC-crisis is clearly hitting Intel more than it hits AMD, the Wintel-alliance is obviously over, the marketing-powers have shifted to other regions of the world, this shit is not being 'Politically-Correct' either, see ??

Re:Performance/Price: AMD always wins.

[smash]

Please find me a competitive AMD CPU that wins against my core i5-4430 over 3 years with power consumption taken into account. Cheers.

Re:Performance/Price: AMD always wins.

[smash]

ps. my power rate here in Australia is roughly 26c (aussie) per kw/h, and my machine is turned on approximately 12+ hours per day.

R O T F L M A O

[Anonymous Coward]

Hahahahaha http://news.slashdot.org/comme...

Sardaukar86: It's not polite to talk

[Anonymous Coward]

With your mouth full (as you "eat your words" vs. apk) http://it.slashdot.org/comment...

Sardaukar86: Eating your words

[Anonymous Coward]

!= good nutrition & not polite to talk w/ yer mouth ful eating 'em http://it.slashdot.org/comment... hahahaha

Re:Sardaukar86: Eating your words

[Anonymous Coward]

I wonder if SourDildoInACarFrom1986 would be kind enough to tell us how his words tasted? LMAO!

Re:Sardaukar86: Eating your words

[Anonymous Coward]

Hahahahahahaha ( @ "SourDildoInACarFrom1986" )

Always a bridesmaid never a bride

[Anonymous Coward]

I have read stories about AMD making a come back for so many years its not even funny. I know many consumers who still don't even know AMD exists.
For me AMD as a processor company has at least managed to tread water all these years. But to say they have ever really attracted mainstream in a significant way since desktops began their slow demise is a reach. AMD simply dropped the ball for too long on mobile to ever gain much confidence from PC makers. Other then HP and a couple others who offered some hope of competition with Intel but always fell short of what consumers wanted. Which was processing power and good battery life. Sure ATI purchase was a big help in the graphics department. But the end results have been mixed and far from setting the world on fire against Intel. I hope AMD holds together and continues its quests competing against Intel. But they may have to resolve themselves to the fact that being number two is not so bad.

Tantrums, Sardaukar86? Hypocrite

[Anonymous Coward]

Sardaukar86 profanity riddled "FoaMiNg-@-The-MouTh" priceless "ReAcTioN" #1 -> http://news.slashdot.org/comme...

Sardaukar86 profanity riddled "FoaMiNg-@-The-MouTh" priceless "ReAcTioN" #2 -> http://news.slashdot.org/comme...

Why? He had to "eat his words" vs. APK http://it.slashdot.org/comment...

Re:Tantrums, Sardaukar86? Hypocrite

[Sardaukar86]

Oh, you poor child. I might have known an appeal to your consideration for other humans would be pointless. You're just too selfish.

Keep an eye on my posting history APK, keep desperately trying to prove yourself. I know you only do so because you haven't the chops to hold your own in a debate, much less win a point. Do your attacks help you? Do you think that someday we'll all 'come around' to your 'thinking'? Yeah, those sock-puppet 'supporters' of yours are all the friends you have here buddy.

You have no debating skills at your disposal beyond shrieking and fling poo, and, oh yeah, links to old comments that do more to prove my point than yours, pity you don't understand. BTW thanks for trotting out your list again, can we please see that in every response? It always gives me a chuckle.

K. S. Kyosuke = "Run, Forrest: RUN!"

[Anonymous Coward]

From a fair challenge like a chickenshit blowhard http://slashdot.org/comments.p...

K. S. Kyosuke gets called out & ran

[Anonymous Coward]

From a fair challenge like a chickenshit blowhard http://slashdot.org/comments.p...

I love summaries written by 20-somethings

[Anonymous Coward]

It's funny to read summaries written by 20-somethings who obviously weren't around in the 1990s when AMD was king and outsold Intel 3 to 1 in the desktop market, all starting with the 40Mhz 386 DX, and again trouncing the Pentium with the K5.

Because fabs can run $10 billion each

[Nova Express]

Actually running your own fab can give you tremendous economies of scale if you know you'll be running you part (or its die shrink successors) 24/7/365. The per chip costs are going to be lower.

But to build a state-of-the-art, 300mm, 14-nm fab with all the latest process technology can run you $10 billion. AMD doesn't have enough mnoney to make those bets anymore, and few companies do.

Going with a foundry means you earn less profit per chip sold, but it also let's you avoid that $10 billion up-front investment.

Rationale behind definition of mobile

[tepples]

Should laptops be considered "mobile"? If not, why not?

Go back to "eating your words" goof

[Anonymous Coward]

Sardaukar86 "FoaMiNg-@-The-MouTh" ReAcTioN #1 http://news.slashdot.org/comme...

Sardaukar86 "FoaMiNg-@-The-MouTh" ReAcTioN #2 http://news.slashdot.org/comme...

Why?

He had to "eat his words" vs. APK http://it.slashdot.org/comment...

Re:Go back to "eating your words" goof

[Anonymous Coward]

ROTFLMAO ( "goof" ) hahahahaha

Itanic's ORIGINAL goals

[unixisc]

No, the original goal of that CPU was to replace x86 altogether and give Intel an architecture that they didn't legally have to share w/ anybody - unlike x86, where Intel had not just AMD but other competitors like Cyrix and Centaur. They saw the introduction of new CPUs like Alpha & MIPS running NT, and the possibility that PowerPC would run OS/2, and thought they'd be left behind in the CPU sweepstakes. They also thought that others were already ahead of them in RISC, which is why they decided to risk things w/ VLIW. Since there was the AIM alliance behind PowerPC, MS-DEC-MIPS w/ NT and Sun running w/ Solaris/SPARC, the only one left was HP. Hence, Intel went w/ this alliance, and HP too was enthusiastic, since Intel's fabs would give them the manufacturing edge.

AMD64 was exactly the iceberg that sunk the Itanic - not b'cos it was AMD, but b'cos it was x86 compatible. And to add insult to injury, when Intel came up w/ their own extensions, Microsoft told them that they weren't re-spinning 64-bit Windows for them, forcing Intel to be AMD compatible for a change. Also, x64 - initially the Opterons and later the Xeons - started being used everywhere that Itanium was supposed to have been used. Before long, the Itanium workstation market was dead, and later, every server manufacturer that used it was gone as well. Today, it's only used by some Chinese vendors in a few places, other than HP's Integrity servers. Also, now even HP/UX is being ported to Xeon, completing the irony. Previously, when Windows NT dropped support for PowerPC, MIPS and finally Alpha, it looked like a nail in the coffin for RISC. But when everybody - Windows, Monterrey UNIX, Solaris, and even Linux dropped support for Itanic, it was really funny. You know you're bad when even Debian ultimately drops you. Right now, it's just HP/UX and FreeBSD left on Itanic

Changed markets

[unixisc]

You're missing the point. Itanium's target market changed b'cos AMD came up w/ a compatible 64-bit x86, and Intel was forced to join. While a major portion of the market may not have switched to AMD, they did switch to Xeon, Core and other 64-bit x86s from Intel. Which was supposed to have been dominated by the Itanium in the first place - Intel didn't want there to be a 64-bit CISC

NOT very scientific of you

[Anonymous Coward]

Especially seeing this http://slashdot.org/comments.p... after your utter line of crap above it here http://slashdot.org/comments.p... so how stupid do you feel asshole? APK fried you like an egg yet again, lol!

Re:NOT very scientific of you

[Sardaukar86]

Oh dear, you really are rather butt-hurt, aren't you poor baby? My commentary of your myriad issues must have been a little close for comfort, eh? My, look at all the love-letters you sent me! You must really be upset with me - I shall cherish them always!

Oh and I'm very sorry you've gotten sand in your inflamed vagina over me, APK, no really I am.

2 yrs. later you're still butthurt?

[Anonymous Coward]

Since apk made you "eat your words" (lmao)? Please http://slashdot.org/comments.p...

Re:2 yrs. later you're still butthurt?

[Anonymous Coward]

It's obvious Sardaukar's a troll who stalked apk since those posts where apk made him "eat his words" by downmodding apk's posts because it's the best he has after apk made him totally "eat his words" like the bigmouthed erroneous fool and troll that's butthurt that Sardaukar86 obviously is all butthurt and what-not.

No, no: Tell us about "rational discourse"

[Anonymous Coward]

Sardaukar86 "FoaMiNg-@-The-MouTh" ReAcTioN #1 http://news.slashdot.org/comme...

Sardaukar86 "FoaMiNg-@-The-MouTh" ReAcTioN #2 http://news.slashdot.org/comme...

Why?

He had to "eat his words" vs. APK http://it.slashdot.org/comment...

Then, Sardaukar86 the reprehensible little lying jackass did a post on "rational discourse" pointing out APK directly in it too AFTER the above? LMAO - please: Give me a break! See here http://slashdot.org/comments.p...

Re:No, no: Tell us about "rational discourse"

[Anonymous Coward]

R O T F L M A O @ Sardaukar86 the butthurt little clown wannabe computer guy!

Re:Performance/Price: AMD always wins.

[ReekRend]

Concerning power consumption, technically probably not, but I suppose I'm referring to home users. Even if you work out all the math for difference in electricity costs, it's not going to make a difference unless you're putting heavy loads on your computer every single day (busy servers, your whole life is gaming, etc). Realistically for home users the difference would only be a few $ per year. Of course it's obvious for data centers that they will save money by paying more for power saving CPU's. For home users the computer is idle most of the time anyway. Reference for cost/CPU wattage: http://www.tomshardware.com/fo... http://www.bit-tech.net/blog/2...