Next Gen Intel CPUs Move To Yet Another Socket

mr_sifter writes "According to a leaked roadmap, next year we'll be saying hello to LGA1155. The socket is 1-pin different from the current LGA1156 socket Core i3, i5, and some i7s use. Sandy Bridge CPUs will be based on the current 32nm, second-generation High-k metal gate manufacturing process. All LGA1155 CPUs will have integrated graphics built into the core instead of a separate chip. This is an upgrade from the current IGP, PCI Express controller and memory controller in Clarkdale CPUs. which is manufactured on the older 45nm process in a separate die (but still slapped together the same package). This should improve performance, as all the controllers will be in one die, like existing LGA1366 CPUs."

Comments

This socket goes to 1155

[jayhawk88]

Well, it's one louder...err faster, isn't it?

Re:This socket goes to 1155

[Bakkster]

Nigel Tufnel: Look at this pin... still has the old tag on, never even used it.
Marty DiBergi: [points his finger] You've never used...?
Nigel Tufnel: Don't touch it!
Marty DiBergi: We'll I wasn't going to touch it, I was just pointing at it.
Nigel Tufnel: Well... don't point! It can't be used.
Marty DiBergi: Don't point, okay. Can I look at it?
Nigel Tufnel: No. no. That's it, you've seen enough of that one.

Integrated graphics in the CPU?

[Lord Lode]

I can see that integrated graphics in a CPU can be handy for some applications, like low-power mobile stuff and such.

But for a desktop PC, isn't this a disadvantage? If you're using a proper graphics card, couldn't that space in the CPU be used for better things than a redundant graphics circuit?

Re:Integrated graphics in the CPU?

[Farmer Tim]

...couldn't that space in the CPU be used for better things than a redundant graphics circuit?

At first I read that as "retarded graphics circuit". Still made perfect sense...

Re:Integrated graphics in the CPU?

[TheLink]

Basically they've run out of ideas on how to use those billions of transistors to make things faster or better.

It's either:
1) Another CPU core
2) Yet more cache.

And now GPUs...

Too bad Intel can't make great GPUs.

Re:Integrated graphics in the CPU?

[Thanshin]

Not if you convince "proper graphics card" to see it all as a CPU integrated in their graphics card.

I don't think it'd be very hard right now to convince an alienware buyer to uy a computer that's essentially a graphics card with all the rest integrated around it. Except, maybe, the hard drive. And even there you could argue "it has a SSD for you to install one or two games at a time. You can buy a standard HD for the rest."

The only thing to leave outside would have to be the mouse (some elite pro killer razer) and the keyboard (a pro ergonomic strafer pro elite pro fatal1ty). And the headset.

Re:Integrated graphics in the CPU?

[marcansoft]

Um, no. Cache is very important, especially with 64-bit code. In fact, x86 is a terribly die-area-inefficient architecture; we'd be a lot better off with a modern RISC, opening up space for more cache.

Re:Integrated graphics in the CPU?

[msgmonkey]

Your point would have been valid 10 years ago but the die area used for the CISC instruction decoder on a modern x86 processor is negligible. Infact the x86 instruction set is more compact than a pure RISC cpu so you can fit more instructions into the instruction cache (ARM processors have a THUMB mode with more compact 16bit instructions because of this).

Re:Integrated graphics in the CPU?

[Calinous]

RISC typically needs more RAM than CISC (and it seems less than 10% of the die area is devoted to x86 instruction decoding, at least in high-performance processors), so you'll trade the space for more cache for the need for more main memory.

Re:Integrated graphics in the CPU?

[Rogerborg]

Heck, I remember "integrated graphics" the first time round. It was called "using the CPU to do graphics", and it was good enough for us to render buggy whips in 2D, sometimes even 2.5D.

Also, what's with kids these days playing their hippety-hop music way too loud using integrated chips rather than a good old ISA SoundBlaster 16?

Re:Integrated graphics in the CPU?

[HarrySquatter]

In fact, x86 is a terribly die-area-inefficient architecture; we'd be a lot better off with a modern RISC, opening up space for more cache.

Is this ignoring the fact that most of Intel's chips for many years have basically been RISC processors with an x86 translation unit?

Re:Integrated graphics in the CPU?

[Targon]

Low end systems become even cheaper to produce when the chipset on the motherboard does not need to include graphics support. Also, if your add-in video card fails, you can always run off integrated until you can replace it. You are right about a 'proper' video card being a better choice overall, but if you look at those $400 to $500 computer towers being sold all over the place, not a single one has a dedicated video card.

Now, AMD is moving forward with their Fusion project, which will add a GPU to some processors, but since AMD has Hybrid CrossfireX, the GPU portion on the CPU COULD be made to work with an add-in Radeon video card for extra graphics processing power. Intel on the other hand does not make stand-alone graphics products, and they also have zero experience with multi-GPU technologies to do something similar. NVIDIA has SLI, but without a CPU, they are not a part of this discussion.

The big problem that Intel has is that their graphics technology is sub-standard, so at best, it is still about 'integrated graphics' for the performance levels. Intel also loves forcing people to upgrade motherboards CONTINUALLY, because Intel isn't just the CPU manufacturer, they are the chipset manufacturer as well. In general, Intel may have a performance lead at the high end, but due to the long-term costs, those looking for a mid-range system can still find AMD products worth using. The fact that you can start with a basic system with a dual core processor, and with just a BIOS update, you will be able to drop in a six code Phenom 2 without ANY other changes.

Re:Integrated graphics in the CPU?

[marcansoft]

The key is modern RISC, not RISC. x86 is horribly inefficient. I'm not talking about the instruction decoder, I'm talking about the instruction semantics. x86 was never designed for today's high-performance CPUs, and the result is that the instruction set basically allows the programmer to do anything they want, even if it goes against modern CPU design optimizations. This forces the CPU to devote a large amount of die area to workaround logic that detects the thousands of possible dirty tricks that a programmer might use which are allowed by the ISA. For example, every modern RISC requires that the programmer issue cache flush instructions when modifying executable code. This is common sense. x86 doesn't, which means there needs to be a large blob of logic checking for whether the data you just touched happens to be inside your code cache too. The fact that on x86 you can e.g. use one instruction to modify the next instruction in the pipeline is just so ridiculously horribly wrong it's not even funny. There are similar screw-ups related to e.g. the page tables. I can't even begin to imagine the pains that x86 CPU engineers have to go through.

You can make an x86 chip reasonably small and very slow, or very large and very fast. x86 doesn't let you have it both ways to any reasonable degree.

Re:Integrated graphics in the CPU?

[ryantmer]

Who needs a proper graphics card these days? Only people who need real-time high-def geometry rendering. They are free to chose a CPU without an on-die GPU. Everyone else is happy they won't need additional cooling infrastructure on their mainboard.

Have you ever tried playing anything but Farmville on integrated graphics? Apparently the Clarkdale CPU/GPUs are only about 1.5x more powerful than integrated at this point. While I realize the Sandy Bridge (who comes up with these names, anyway?) should improve on this, I'm guessing that even playing TF2 on one of these things would be rather unsatisfactory...

Re:Integrated graphics in the CPU?

[marcansoft]

I love how everyone jumped so quickly on the instruction decoding bandwagon. Of course instruction decoding is cheap these days, even for x86. The problem isn't decoding, it's the huge amount of dirty things that instructions can potentially do after being decoded. Things that go against modern high-performance CPU design principles.

Re:Integrated graphics in the CPU?

[Firethorn]

But for a desktop PC, isn't this a disadvantage? If you're using a proper graphics card, couldn't that space in the CPU be used for better things than a redundant graphics circuit?

Don't look at the PC enthusiast/gamer market. Look at the desktop PC for basic business use. Cost is much more king there, as long as performance is acceptable. You gotta cut a lot of costs if you want to be able to slap down a whole PC for less than $200.

I wouldn't be surprised if in a couple more generations we're looking back at 'system on a chip' designs. No northbridge, southbridge, video controller, etc... Just a central chip on a board with power and interface leads.

Re:Integrated graphics in the CPU?

[maxume]

It's (mostly) useless for gamers, but one thing Intel does do is make excellent drivers for their graphics chips (at least, that is my experience under XP).

Re:Integrated graphics in the CPU?

[MoldySpore]

Integrated graphics in the CPU are also worthless if you don't have a motherboard with native video outputs, unless I have read the wrong info about this tech.

Re:Integrated graphics in the CPU?

[amorsen]

The neat thing about the x86 architecture is that it has forced the chip designers to be really clever. E.g. the register limitations has forced them to find ways to make level 1 cache really fast; you'll be hard pressed to find non-x86 chips with faster level 1 cache. Similarly, the system call latency is fantastic. Most importantly the (quite) strong memory ordering provided by x86 means that x86 is pretty much unmatched when it comes to inter-CPU communication. Look at the hoops e.g. PA-RISC goes through to handle SMP in Linux, then compare with x86 where some of the memory barriers are even no-ops and just get turned into compiler barriers. RISC designers believe that cache coherency is expensive and that programmers should be aware of that cost. x86 has proven that it doesn't have to be expensive -- except in chip real estate.

Anyway, if you can live with somewhat sub-par performance, Larrabee shows that it's still possible to get decent performance with a smaller chip. Not quite as small as ARM or MIPS though, so a lot of the embedded space is closed to x86.

Re:Integrated graphics in the CPU?

[toastar]

You can make an x86 chip reasonably small and very slow, or very large and very fast. x86 doesn't let you have it both ways to any reasonable degree.

Isn't that what ARM is for? Besides I want a Very Fast Machine with a huge heatsink.

Re:Integrated graphics in the CPU?

[MBGMorden]

I hate to break it do you, but a huge portion of PC users DON'T play anything more intensive than Farmville on their systems - if they game at all.

Even for myself - I do play games on my PC, but only on 1 of them. I've got 5 systems (Windows desktop, Linux desktop, Linux laptop, Mac desktop, and Windows desktop at work) and ONLY my Windows desktop at home ever sees any gaming. In the other 4 I really don't care what chip is in them because Chromium, Visual Studio, Safari, etc simply don't need it.

The simple truth is that this move will lower the cost for systems with integrated graphics, and likely increase the cost of systems with dedicated graphics. That sucks for those of us that end up buying a dedicated graphics card machine, but when the majority of the market doesn't need that, then overall the decision is a good one.

Besides - integrated solutions are getting better and better all the time. I'd put it akin to sound cards and network cards. Once upon a time no motherboards included these. Then some started doing it but the components were pretty bad. I personally specifically looked for motherboards WITHOUT integrated sound when it first started showing up. These days though, most of the integrated sound cards aren't really that bad. They're good enough for all but the most hardcore users. Same with integrated networking. Most new motherboards have gigabit network chipsets that work plenty well enough for almost all users. I'm going to guess that in 5-6 years, the integrated graphics will be plenty good enough to play your average game at an acceptable level. Separate graphics cards will still be available, but they'll cost a bit more (smaller market), and they'll only be needed by those who absolutely have to play at the highest resolution with every detail slider set to max.

Re:Integrated graphics in the CPU?

[amorsen]

Is this ignoring the fact that most of Intel's chips for many years have basically been RISC processors with an x86 translation unit?

This doesn't really make sense. ALL CISC processors are pretty much RISC processors with a translation front end. This has been true since the 8086 and (especially) the 68000, when RISC wasn't invented. The whole point of RISC is that it was discovered that you can live without that front end. Look up microcode.

The original 8086 was a bit RISC-like in that some instructions were in hard-coded logic and didn't go through the microcode layer. Modern x86 is less RISC-like, because all instructions need to go through the instruction decoder.

Re:Integrated graphics in the CPU?

[ThreeGigs]

Not sure if it's possible, but I'm guessing that if one added a graphics card, then the processing power of the graphics portion of the CPU could be used for other things. Granted, I wouldn't expect CUDA type performance, but I'd think a few new instructions that allowed programmers to specifically target unused graphics units for processing SIMD instructions would be welcome. Same thinking goes for the AMD chips. Basically an either-or choice: all-in-one chip, or increased computational power... which are generally markets which don't overlap.

Has anyone seen any patent filings related to something like this?

Re:Integrated graphics in the CPU?

[tlhIngan]

I can see that integrated graphics in a CPU can be handy for some applications, like low-power mobile stuff and such.

But for a desktop PC, isn't this a disadvantage? If you're using a proper graphics card, couldn't that space in the CPU be used for better things than a redundant graphics circuit?

Except, most PCs don't have proper graphics cards. It's why the top video card manufacturer is not nVidia or AMD, it's Intel. Yes, Intel sells the most graphics chipsets.

And graphics is one of the last high-speed devices to be integrated into the CPU core - these modern CPUs are basically the CPU core plus north bridge in one, integrating the PCIe bridges and memory controllers (and now GPU). The other peripherals are much slower and can hang off the PCIe bus.

And one less chip on the motherboard == big savings for OEMs like Dell/HP/etc, where 99% of the PCs they sell by volume (not by model) probably ship with Intel graphics.

Re:Integrated graphics in the CPU?

[bberens]

It's only useless for gamers because this is the early generations of the technology. More than likely for the next several decades hard-core gamers will want a dedicated video card. But at some point soon you will wind up with "good enough" graphics on the main die to get by for the casual gamer on all but the most graphics intense games.

Re:Integrated graphics in the CPU?

[The Archon V2.0]

I can see that integrated graphics in a CPU can be handy for some applications, like low-power mobile stuff and such.

But for a desktop PC, isn't this a disadvantage? If you're using a proper graphics card, couldn't that space in the CPU be used for better things than a redundant graphics circuit?

One could make the same argument about motherboards right now. A lot of them come with onboard graphics that takes up space on the board better used for SATA ports or some such, and yet people still buy them and stick video cards on them.

I've seen a lot of systems. If it has a video card, odds are better than 50/50 it's also got onboard video.

Re:Integrated graphics in the CPU?

[timeOday]

More than likely for the next several decades hard-core gamers will want a dedicated video card.

The 3dfx Voodoo card was released less than 1.5 decades ago. Game console generations are getting longer and longer. My crystal ball says few people will be buying standalone graphics cards in 5 years.

Re:Integrated graphics in the CPU?

[Hal_Porter]

Yeah but the whole CPU core including those tricks is a small percentage of die area. Look at any CPU and it will be mostly cache. As someone puts it here

http://www.edn.com/blog/920000692/post/370042437.html
1. The claim that the instruction decoder on the Atom is huge is nothing more than a straw man. While this sort of claim might have been somewhat accurate 20 years ago, instruction decoders are tiny compared to the rest of a processor with today's designs. Atom's decoder is almost nothing compared to other parts of the processor, which mostly consists of caches by this point (like all modern processors used as CPUs). I have no special knowledge because I don't work for Intel, but there's a die photo of Intel’s Atom processor in the April 7, 2008 article on the Atom by Tom Halfhill ("Intel’s Tiny Atom") in The Microprocessor Report. The die photo outlines the processor’s main blocks, including the FEC (front-end cluster), which includes the instruction decoder, branch predictor, I cache, prefetch buffers, and instruction TLB. The entire FEC consumes about 10% of the die (that’s a quick eyeball estimate) and the instruction decoder is a small portion of that 10%. The very visible, regular features of the I cache, prefetch buffers, and TLB constitute most of the area devoted to the FEC. Put the instruction decoder down as consuming a mere 1-2% of the Atom's real estate.

You mention self modifying code. SMP machines need to bus snoop anyway - handling x86 code semantics means you need to snoop on data writes from the processor into its own instruction scheme.

Still it has to be correct but it doesn't need to be fast. Basically the CPU has to catch any hint of this and flush caches. So in practice self modifying code will run very slowly and thus be rather uncommon.

It's like all x86-isms. Yes some nasty stuff is legal and a CPU must support it. However it is not common and performance critical code won't do it. Which explains why only a few percent of the die is used. Most of the die is used for cache, and that is not wasted by any reasonable definition.

Re:Integrated graphics in the CPU?

[znerk]

I hate to break it do you, but a huge portion of PC users DON'T play anything more intensive than Farmville on their systems - if they game at all.

Ya know... I've seen Cafe World bring a dual-core with 4 gigs of RAM to its knees. 60% CPU usage, with half a gig of RAM in use for Firefox alone... with no other apps running, nor even any other browser windows/tabs open.
Hell, it's sluggish and choppy on the quad-core 2.8 in my living room. Farmville is a little better on most days, but still.
Browser-based Flash Game != low-powered app. Might not be graphics intensive, admittedly... but Zynga really needs a head check when it comes to resource usage.

Re:Integrated graphics in the CPU?

[Hal_Porter]

It's ignoring the fact that 90-99% of die area on any modern CPU is a cache of some sort - i.e. instruction or data, or a TLB, or an branch predictor.

Re:Integrated graphics in the CPU?

[ryantmer]

While you are absolutely correct on most points, including the fact that most people do not play anything more intensive than Farmville, people are becoming more obsessed with the pretty on their screens. It simply seems that it is a constant battle between graphics processing power and the number of pixels people want to see. At this point, playing a Blu-Ray film on a desktop/laptop with integrated graphics is viable, but not excellent. In addition, don't count PC gaming and PC enthusiasts as dead just yet. There are still plenty of people that want and need more power than integrated graphics can provide at this point.

Re:Integrated graphics in the CPU?

[ultranova]

Basically they've run out of ideas on how to use those billions of transistors to make things faster or better.

How about a specialized CPU? Lots and lots and lots of weak single-threaded cores with their local non-shared memory, all running their own small program and connected to a very fast bus, allowing them to pass messages to each other. It would be ideal for many emerging applications, such as image recognition and AI in general.

The thing is, a general-purpose serial CPU is already as fast as it's ever going to get; already manufacturers are seeking further growth from adding extra cores rather than making existing ones faster, and have for a long time (since Pentium) used behind-the-scenes parallelism for it. While this leads to nominally x86-compatible processors, in reality using them efficiently requires completely reworking most programs, and even then, things like shared memory and the synchronization it necessiates incurs a huge penalty. So why not bite the bullet and start the inevitable transition to a truly parallel processor architechture right now?

I'm talking about something like "network-on-a-chip" here. It shouldn't even be difficult to implement, since as I said, modern CPUs are already working more or less like this behind the schemes, and simply emulating a serial processor (which is what allows things like HyperThreading). Whoever opens direct access to this programming model first gets to tailor it to their architechture, so there should definitely be incentive.

Re:Integrated graphics in the CPU?

[Nadaka]

Not all integrated graphics are made the same. Intel integrated are utter shit.

I haven't tried nvidia integrated graphics.

But the ATI 3300 HD series of integrated chips?

It is as good as a top of the line GPU from about 3 or 4 years ago. 128Mb integrated memory, 32 stream processors, decent clock speed. There are a handful of high end games it won't do well with, but it will meet the needs of most people, including those playing TF2.

Re:Integrated graphics in the CPU?

[LWATCDR]

Simple.
There is a HUGE segment of the population that doesn't need anymore graphics capability than what even the crappy Intel integrated graphics offer.
The current offerings are much better.
Here is the maximum graphics requirement for about 80% of all Windows PCs.
Will it playback 1080p video.
And that is the maximum they require.

A lot of people never play any video game that is more graphically intensive than Plants vs Zombies.
A lot of people never play any video better than what is on Youtube.

Think all the PCs that are ill the offices in the world. Integrated graphics is good enough the vast majority of them.
Even on the highest end system in may office that we use for development system we can get by with any video card that can drive two good sized monitors. We do not need high end graphics cards for any of our systems and integrated graphics are just fine for most of them.

Intergrated graphics from ATI and nVidia are actually pretty good and can even do some light gaming. The i3 and i7s that I have tested seem to be much better than what Intel used to provide and seem to be good enough.
Enthusiast systems like what I have at home to play FSX are a small minority of the PC market.
Also the GPU on the CPU could also be used for things like Physics simulation in the game if you have a graphics card installed so even then it could be used.
Odds are that gamers will get the ultra high end quads and they will not have IG on the die anyway.

Re:Integrated graphics in the CPU?

[jvin248]

Agree. Build enough machines and you see that cache trumps raw clock cycle by a lot. That's also what Intel did with Pentium vs Celeron .. same basic clock speeds but Pentium had 2x the cache - and performed much better. A 2.0Ghz Pentium could easily beat a 2.8Ghz Celeron in one rig I was testing for an FEA workhorse a while back.

Re:Integrated graphics in the CPU?

[ryantmer]

The ATI Mobility Radeon HD 3000 Series of discrete graphics allow you to experience the power of HD with graphics processing designed for how you work and play

From here. Perhaps this is the wrong chipset, if so, please correct me. If not, emphasis on the word "discrete".

Re:Integrated graphics in the CPU?

[ryantmer]

From here. Perhaps this is the wrong chipset, if so, please correct me. If not, emphasis on the word "discrete".

That'll show me to do my research. Checked up on the Radeon HD 3300 (which is not part of the HD 3000 series...?). True enough, it does look pretty damned good. However, I will refer to my point above that although this may be enough for some, even most people, there are still those of us who either require or desire more power than even that particular set has to offer.

Re:Integrated graphics in the CPU?

[nmb3000]

It's either:
1) Another CPU core
2) Yet more cache.

More cache please.

Cache is a pig in terms of die real estate, but it has a huge impact on performance for memory-hungry applications (these days that means every application). That said, I'd assume they're using system RAM for the video memory of the on-chip graphics so the actual size of the GPU components is probably pretty small.

Re:Integrated graphics in the CPU?

[sjames]

It seems that x86-64 is going to be with us for a long while. I would like to see strategies to phase out the worst of those "features". Perhaps as part of new enhanced security modes. That way, marketing uses it as bullet points and software gets updated to deal with it. Then EOL the non-enhanced mode.

The superiority of the post-EOL chips should argue against the inevitable moaning about the missing "features".

Re:Integrated graphics in the CPU?

[antientropic]

The fact that on x86 you can e.g. use one instruction to modify the next instruction in the pipeline is just so ridiculously horribly wrong it's not even funny.

This hasn't been the case for a long time (if ever). I remember writing some some self-modifying code on a 386SX that worked fine when single-stepping through it in a debugger but not when executed normally, due to instruction prefetch. It's required to do a JMP to flush the instruction prefetch queue. (See also here.) It is true however (AFAIK) that the CPU requires logic to detect when a data write implicitly invalidates the instruction cache.

Re:Integrated graphics in the CPU?

[Svartalf]

However...it's not GPU intensive...which would be where you might need something more than an IGP.

Re:Integrated graphics in the CPU?

[Svartalf]

Which is why I'm waiting to see what ARM does with their stuff. :-D

Re:Integrated graphics in the CPU?

[imgod2u]

Not even close. Look at the die photo of Nehalem, for instance. About ~50% is visible cache and special purpose caches may make up ~10% more.

Re:Integrated graphics in the CPU?

[imgod2u]

Adding more cache isn't free. Ignoring power and area, indexing more cache also increases access times. This is why Intel has always opted to use small sizes for their L1 cache.

You can add more L3 on-die, but there are vast diminishing returns to that due to the high-latency access times of L3 cache.

Re:Integrated graphics in the CPU?

[imgod2u]

That really depends. The instruction binary itself is more compact compared to say, ARM, but the actually implementation is far from.

For instance, almost every x86 processor these days pre-decode and store that data into cache. These are invisible to the programming model but it also means that you really aren't getting that many more instructions-per-bit of cache compared to say, ARM.

If you look at how x86 is implemented, and how most compilers use it, about 80% of the ISA isn't even used and the implementation is almost all about accelerating the RISC-like instructions (load, store, execute, permute) rather than the CISC-ish instructions (execute with a memory operand).

That's not to say it's that much worse than its ARM counterparts when it comes to area. But it is a lot worse when it comes to power.

Re:Integrated graphics in the CPU?

[marcansoft]

You clearly haven't bothered to look at any die photos. Check out the Atom. Less than half of the main area is L2. You can see the smaller L1 blocks and other caches, but the vast majority of the rest is logic. That's at least 40-50% logic, more if you count the FSB IO areas. larger CPUs aren't much different.

Then you're also missing the fact that the caches themselves are more complicated (and larger per size) on x86, due to the aforementioned extra snooping required.

Re:Integrated graphics in the CPU?

[colinnwn]

Linux users that want decent graphics performance. Unfortunately right now closed source Nvidia drivers set the standard for Linux performance.

Re:Integrated graphics in the CPU?

[msgmonkey]

Self modifying code on x86 whilst supported, is there just for backward compatibility and has been very slow since the pentium pro and although it is messy it is handled at the x86 to micro-op conversion stage. It may make the decoder twice as large but as my original point stated (and still stands) the die area used is nothing compared to the space used by caches on modern CPUs so decreasing its complexity for a few extra K of cache would give you no net gain.

There is no difference in concept between a modern RISC and a classical RISC processor in terms of instructions being fixed size and hence tend to use more space compared to x86 code generated by compilers. What you can and can not do to memory varies on RISC processors too for example ARM processors do not support snooping so in order to a DMA you have to make sure that your caches are flushed to memory whilst many MIPS implementations support it.

Infact whilst the ARM instruction set is nice and clean devices vary so much most applications support V4 as the lowest common dominator. Floating point also a complete mess, VFP, VFPv2, VFPv3, NEON etc. So it is no different in real terms compared to x86 where you have to support MMX, SSE, etc.

Yes the decoder needed to handle x86 does n't allow you to make small CPUs that perform well but that was n't my point or your point that I replied to.

Re:Integrated graphics in the CPU?

[TheEldest]

That's why Intel is coming out with 2 sockets. 1155 and 2011. The 2011 is the non-IGP socket, 1155 is for IGP use.

The situation isn't changing. Intel will make an IGP and a non-IGP solution. Same as now. Same as it's been for quite a while.

If you don't want integrated, then don't use it.

Re:Integrated graphics in the CPU?

[nabsltd]

My crystal ball says few people will be buying standalone graphics cards in 5 years.

Based on the fact that there are games being released today that won't run smoothly (i.e., at least 30fps at all times) with highest levels of detail when using the current fastest standalone card, I suspect your crystal ball is cloudy.

Since these standalone cards can generate upwards of 200 watts of heat, there is no way to put that much computing power into a package that already generates 80-100 watts to do CPU things. As the chips continue to use smaller processes, there will be advances that allow more computing with less heat, but those graphics cards are already at 40nm tech, so they're only about half a generation behind.

Based on the progress over the last few years, I suspect that 5 years from now we will see about 2-3 times the GPU power with about 50-75% of the heat. Even if the CPU portion does as well in technological advance, you'd end up with a 140 watt TDP (at best) for a combined CPU/GPU that doesn't require a standalone card to run most games.

Re:Integrated graphics in the CPU?

[pszilard]

But for a desktop PC, isn't this a disadvantage? If you're using a proper graphics card, couldn't that space in the CPU be used for better things than a redundant graphics circuit?

The simple answer is: no!

Think about OpenCL and in general stream computing - which I believe is a the future of computing through a slow convergence of multicore and manycore. Having a more tightly coupled CPU and GPU (i.e. no PCI express bottleneck) would have the huge advantage of being able to do more streaming tasks on the GPU while not having to worry that communication will kill the performance.

There are already quite many GPU accelerated applications in various HPC research and industrial applications as well as desktop applications (just a few examples: Adobe CS5, OS X Snow Leopard, Badaboom media converter,...)

Re:Integrated graphics in the CPU?

[Bearhouse]

I can see that integrated graphics in a CPU can be handy for some applications, like low-power mobile stuff and such.

But for a desktop PC, isn't this a disadvantage? If you're using a proper graphics card, couldn't that space in the CPU be used for better things than a redundant graphics circuit?

You're forgetting that many OEMs can save a bundle by using integrated graphics, offering stand-alone cards only on 'high-end' models or as upgrades, (but using the same chassis & mobo).

Re:Integrated graphics in the CPU?

[Calinous]

Not all applications benefit from more cache - some benefit from faster cache, and more cache and faster cache are usually exclusive.

Re:Integrated graphics in the CPU?

[yuhong]

For example, every modern RISC requires that the programmer issue cache flush instructions when modifying executable code.

Even Motorola decided to require this for later 68K processors too, which in the short term did cause compatibility issues (remember when the MC68040-based Mac Quadra was released and the programs incompatible with it?).

Re:Integrated graphics in the CPU?

[dfghjk]

There's a large body of proof that says that you should stop displaying your ignorance. Modern x86 designs are approached the same way as other processors. RISC, in original justification, is dead. x86 is not "die inefficient" any more than RISC is why you consider the greater demands RISC places on memory bandwidth and subsequent cache sizes.

Re:Integrated graphics in the CPU?

[dfghjk]

and yet the doom predicted by that has failed to materialize over the 20+ years we've been hearing about it. Huge amounts of "dirty things" don't necessarily translate to large amounts if die size or anything else other than engineer's tears.

Re:Integrated graphics in the CPU?

[Baloo Uriza]

I haven't seen too many games that don't perform adequately on GPUs from 2-5 years ago these days.

Re:Integrated graphics in the CPU?

[mrawhimskell]

Hmm, reminds me how Apple improved on Nvidia's Optimus in their latest macbook pro ranges. Turn off the on-chip graphics when not in use and use the discreet when there's cause to do so. Optimus, from what I gather, just switches to the discrete graphics whilst leaving the on-chip graphics running. Clever huh? From a non-apple fanboi

Re:Integrated graphics in the CPU?

[petermgreen]

But for a desktop PC, isn't this a disadvantage? If you're using a proper graphics card, couldn't that space in the CPU be used for better things than a redundant graphics circuit?
I doubt it's worth intels while to produce a specific chip for those who want a generally low-midrange config (2-4 cores, 2 channel memory, limited PCIe etc) and yet don't want integrated graphics. Especially as even with the integrated graphics disabled these will probablly be among the most power efficiant chips in their class.

The high end stuff (which will be on other sockets that support more stuff) will probably remain free of integrated graphics.

Re:Integrated graphics in the CPU?

[amorsen]

Microcode didnt come until the Pentium Pro

This is simply not true. UPDATABLE microcode appeared in the Pentium Pro, before that it was burned directly into the chip. Microcode has been used since the 8086, possibly since the 4004.

Re:Integrated graphics in the CPU?

[Vectormatic]

To be honest i dont quite understand the gripe people have with nvidia's linux drivers being closed source.

My main machine is a linux box, it has a kick ass nvidia card (GTX260), and everything works just fine. If i chose to game (which i hardly do these days), performance will be good, and it will work.

If you need to be sure that 100% of code is trustworthy on your machine, you shouldnt be running any ol game anyway (most open source games hardly require top performance), and it isnt like you are going to troubleshoot that one graphics bug by wading through megabytes of driver source code anyway.. (i wouldnt anyway, even though i code for a living).

Sure, open source is more in-sync with the whole linux/gpl/gnu philosophy, but practically speaking, i dont see any reason to bitch about it, as long as it works...

(hell, you could even set up a different linux partition soully dedicated to games if you want to be 100% sure your main system is fully open source)

Re:Integrated graphics in the CPU?

[Vectormatic]

this is totally architecture dependent.

Intel needed their large caches for netburst because pipeline stalls where extremely costly, and the memory had a very large latency, they solved both issues in core 2, with very advanced prefetchers to hide the latency, and a shorter pipeline to make stalls less costly.

AMD on the other hand, moved the memory controller on die with the advent of the Athlon64, which made for extremely low memory latency. Back in the day i read up on cache sizes for AMD chips (because i was going to buy a chip, and wanted to know if going for the large l2 was worth it). As it turns out, in not a single benchmark did a chip with 1 mb l2 beat a chip with only 128k l2 (an 1/8th) at the same clock by more then 10% (and the 10% was an extreme).

The only reason todays chips have such large amounts of L3 is because we dont just have one core sucking down data, we have four (or six, or in case of nehalem, 4+4 virtuals). If intel still had the old FSB, i3/5/7 would need even bigger caches

Re:Integrated graphics in the CPU?

[colinnwn]

I don't have a big issue with it, and it isn't a philosophical thing with me. I don't game at all so that's not a problem. I know there are probably trade secrets in the code they'd like to keep under wraps.

I hope more companies feel comfortable creating closed source software on Linux, and that they believe if it is a good product, people will buy it. But a graphics card driver isn't just a piece of non-critical software that can have minor bugs. It is a critical piece of the computer subsystem required to be very high quality for a good user experience.

I appreciate Nvidia is putting some effort into supplying half decent Linux drivers. And by far Nvidia offers the best performance on Linux. But the quality of the drivers isn't up to the quality of the Windows drivers for the same card. There are bugs, I've been bit by them. Sometimes features in the Windows drivers don't make it to Linux, or they are delayed by 1 to 2 years, an example is VDPAU. I've been needing that on my Myth box for a long time. I just got that on my last dist upgrade.

Some people have said they'd like to try fixing the bugs or adding features. I know it is beyond my skills, but I would appreciate it if those people had access to the source and tools to try.

What GMA stands for

[tepples]

All LGA1155 CPUs will have integrated graphics built into the core

Will the new integrated GPU have performance even on par with a Wii's GPU, or is it the same GMA (i.e. "Graphics My Ass") that's been built into Intel boards for years?

Re:What GMA stands for

[Calinous]

If it's the graphic chip from the i5-661, then it's competitive with the AMD's IGP (AMD might have better drivers though)

Re:What GMA stands for

[bdenton42]

It will probably be junk like usual. If they released on board graphics on par with something like a 9800 GT it would crush NVidia and AMD/ATI as there probably isn't enough of a market above that to keep them operating.

Then there will be Federal investigations and anti-trust lawsuits... they just don't need that kind of trouble.

Re:What GMA stands for

[LinuxIsGarbage]

Will the new integrated GPU have performance even on par with a Wii's GPU, or is it the same GMA (i.e. "Graphics My Ass") that's been built into Intel boards for years?

I always thought it was Garbage Media Adapter.

Re:What GMA stands for

[NewWorldDan]

I have an i3 cpu. Given the pricing, I don't expect great things from the integrated graphics, but it's certainly been capable for light to medium gaming, and as an office desktop (we're standardizing on it at work), it's fantastic. If you want to run Crysis or Dragon Age, go buy a $150 gaming card. Otherwise, as an integrated graphics package, it's all I need and much better than I'm accustomed to.

Re:What GMA stands for

[durdur]

Some of us don't care. Emacs is my main UI and it doesn't need graphics. If I had a good GPU, I'd probably be just abusing it to get more compute cycles.

Re:What GMA stands for

[izomiac]

The newer ones are more powerful than dedicated cards from just a few years ago. The latest line can do 1080p at levels that ATI only recently has been capable of (still not officially IIRC). 3D performance is fine for everything but the newest games at higher detail levels. With open source drivers, minimal power consumption and price, what more do you really want from an integrated card?

(Yeah, yeah, we'd all love for an open source Intel card to out benchmark nVidia and ATI's best, while costing less than $10 and using only a couple milliwatts of power under load. I hear that model comes with it's own Pony.)

Nah, great

[vikingpower]

I just bought a computer with an 1156 socket on its motherboard, which means that THAT computer will be locked to i5 / i7 in a few years. Hmmm.

Re:Nah, great

[IBBoard]

Damnit, I just upgraded my old Athlon 64 3500+ with a nice new Core i5 750 as well. £320 for processor, mobo and 4GB memory. Good job I was hoping for it to last for a while, cos it sure looks like Intel don't want me to just upgrade my processor when it gets to be lacking.

Figures...

[Captain Centropyge]

"Yes, let's force users to upgrade all their hardware when they want a new CPU! Show me the money!"

Re:Figures...

[TheKidWho]

You upgrade the CPU/Motherboard/RAM. Big woop.

You would need a new motherboard regardless if they changed the socket or not. You would also need new RAM since the RAM requires lower operating voltages.

They probably did this so you don't try to plug in the new CPU on your old motherboard thinking it was a straight upgrade when it requires different circuitry.

Re:Figures...

[zach_the_lizard]

You've had to do this for a while. Don't you remember having to get a new motherboard to use newer CPUs, even though they had the same socket? Yeah, I do. That was very confusing at times, and at least with a new socket, you will have a better chance of knowing what will / will not work.

Re:Figures...

[oldspewey]

This is pretty much what I do these days anyhow. I used to get excited about the idea of an "upgradeable" MB, but for the last bunch of years I have found that I just replace the whole machine (maybe minus the case and the stuff the plugs into the back) when the need/whim hits me.

Re:Figures...

[sznupi]

One would think somebody at Intel noticed by now that it's good to write motherboard specs with bigger headroom for lower voltages...

Of course, they simply don't want it; Intel chipsets bring quite a lot money, too.

Re:Figures...

[Captain Centropyge]

So, there's no way to do this using the current socket/motherboard? My guess is that they do this purposely (at least some of the time) so that users need new hardware for their upgrades. It generates more revenue. I work in the software resale industry and the software vendors pull this crap all the time. (e.g. no backward compatibility forces more users to upgrade so that they can all work together)

Re:Figures...

[ircmaxell]

You would need a new motherboard regardless if they changed the socket or not.

Ummm, why? You can upgrade the CPU on an AM2/AM2+ motherboard with at most a flash of the bios. And the AM2/+ CPUs are typically backwards compatible (a AM2+ will run on an AM2, but with reduced functionality). So that AM2 board you purchased 4 years ago is still compatible with the latest processors (but not with DDR3). Given AMD's track record with sockets, I'd be surprised if the AM3 gets "phased out" within the next 5 years (meaning that they stop releasing new processors for it)... So can you explain to me why Intel comes out with a new socket every year (seemingly at least)?

And you only need difference circuitry when the base implementation is flawed (Point to point protocols should be flexible for anything that you may want to be doing... Oh, and AMD has had integrated point to point for what, 9 years now? And Intel JUST introduced one maybe 6 months ago?)...

Re:Figures...

[TheKidWho]

Not without a new chipset.

I don't get why socket compatibility matters when you need a new chipset anyways.

I believe the new processors are using PCI Express 3.0 and require more lanes/copper as well.

Re:Figures...

[sznupi]

But upgrading CPUs has become much more attractive lately - you can go, say, from cheap singlecore (AMD still has some singlecore Semprons; plus singlecore Athlon64 AM2 was quite popular for some time) in original, cheap machine to...also cheap now quadcore. Getting huge boost for very little money (you might also upgrade memory while ddr2 is still cheap)

Of course Intel simply wants you to buy more; chipsets are also quite lucrative after all (maybe pointing out it's a horrible waste would work with current enviromental sentiments?)

Re:Figures...

[bdenton42]

I just upgraded a Northwood 2.8 to a Core 2 Duo 3.0. I used the same graphics card (8800 GT) in the new machine.

Night and day difference.

Re:Figures...

[sznupi]

Only because Intel chooses to obsolete old chipsets (or, more preciselly, arbitrarily changes bus specs on new motherboards - I've seen an ASRock one for C2D with i865). AMD somehow manages to keep latest versions of their CPU interconnect backwards compatible...you really want to say Intel isn't capable of doing so? (especially if Intel simply uses PCI Express for those chips, which is explicitly backwards compatible)

Re:Figures...

[Captain Centropyge]

I guess this tends to highlight my lack of knowledge around hardware and chipsets. I'll concede this round to you. But it would actually save Intel (or their partners) some cash by not having to come up with a new socket every time they develop new CPUs. They need new molds, tooling, etc. to manufacture even just the socket itself. I guess this line of thought doesn't work so well when dealing with this industry.

Re:Figures...

[ircmaxell]

Don't concede. They wouldn't need a new CPU if there was intelligence in the design. CPUs should not be talking directly to anything but memory. All other communication (other processors, PCIe, South Bridge, etc) should be done via a point to point protocol. So then the only thing tying a specific CPU to a specific mobo or socket would be the memory technology. The graphics communication could talk to a PCIe device for the display driving (for the actual conversion of signal to DVI). So a legacy mobo could simply use a dead simple PCIe device to do the translation. A newer one could build that device in. But electrically there should be no need for a new motherboard (since it's communicating out via the point to point protocol). The only thing that's really tying the CPU to a specific mobo/socket would be the memory technology (since the memory controller is on CPU, if a CPU has a DDR2 controller it can't be used in a DDR3 mobo and vise verse)... Intel does this because it can make more money on it. That's it... There's no other reason with today's technology to bind a particular CPU generation so tightly to a motherboard or socket...

Re:Figures...

[oldspewey]

But I'm not likely to buy the cheapest single core CPU in the first place - I typically look for the fattest part of the performance/value curve (usually somewhere in the $150-$200 range).

As for environmental sentiments, it's worth noting that one approach means I end up with an old CPU and some sticks of RAM sitting in a cardboard box somewhere, while the other approach means I have a functioning MB plus all the subsystems ready to be repurposed or donated someplace where it will be of some value to somebody.

Trust me, I've spent a lot of years and money doing incremental upgrades to machines. I guess I've just arrived at a point where I can't be bothered hunting for that elusive 10% performance delta on an existing platform. I'll still add memory or hard drive capacity if necessary, but even that doesn't seem to be particularly frequent anymore.

Re:Figures...

[228e2]

QFT

Im still running my 2.2 Ghz Dual Core on a mobo i bought 6+ years ago . . i'll upgrade when that dies. Its perfectly fast enough for all my games and everything i need to do.

Re:Figures...

[maxume]

The set of users that upgrade anything is pretty small. The set of users that upgrade CPUs is even smaller.

(so maybe they are doing it to force more sales, but they are pissing in the wind if they are)

Re:Figures...

[Targon]

There are different things to consider. On the AMD side of things, which everyone is using for comparison, you can often drop a new CPU into pretty much any AM2+ or AM3 motherboard with just a firmware update. You don't need to replace the RAM or motherboard, and you get the benefits of the new CPU. Going to a new MEMORY type would require a new motherboard, but with all of the new AMD processors, they support BOTH DDR2 and DDR3 memory.

There really is no good excuse for needing an all new chipset for each new generation of processor UNLESS there is a very fundamental change going on. The move from DDR1 to DDR2 to DDR3 for example might be required if the CPU does not support the older memory types(meaning you WANT to prevent users from using a chip in a system that will NEVER support it). Moving to an integrated memory controller, or adding additional pins for more banks of memory MIGHT be an excuse, though these days, extra "reserved" pins should have been put into the socket specification for this, with backwards compatibility so you could drop it into an older system with a degradation in performance(you lose the extra memory controller functionality). Adding graphics to the processor SHOULD work the same way, where the graphics on the processor would not be used if you plug the processor into a system without support for it. Again, looking forward at future needs when designing a new socket would make sense, so you just have a bunch of pins on the CPU that are "reserved" for future use, then, a new CPU would just switch off features the motherboard would not support.

AMD will be moving to a new socket type in the next year or so, due to things like adding the extra pins on the CPU for graphics, a 3, 4, 6, or whatever channel memory controller, or other functions being a part of their plans for the future. But, that next socket should be good to go for the next few generations after that, and for all we know, it may even support current DDR3 processors(DDR2 would probably be dropped since new motherboards would probably not have DDR2 memory support with the new socket).

So, if AMD can do it, people would expect that 'the leader' in the industry SHOULD be able to do it as well.

Re:Figures...

[LinuxIsGarbage]

I just upgraded a Northwood 2.8 to a Core 2 Duo 3.0. I used the same graphics card (8800 GT) in the new machine.

Night and day difference.

Power bill dropped that much?

Re:Figures...

[sznupi]

But you were likely to buy it 3 years ago, in the form of AM2 Athlon64. Now, and still for quite some tinme to come, you can slap in a quadcore. Or if buying now some cheap CPU you would still be able to upgrade to significantly faster one later on...

You can also donate just the old CPU, btw...somebody will need it (old sticks of RAM? You just buy reasonable amount in two sticks at first and have two free slots for later expansion...yes, you have take note to get a motherboard with 4 RAM slots, but that won't really cost you any extra)

"Incremental" is not a good word in this case btw. Just two-step (plus you can get quite a bit more than 10%...)

Re:Figures...

[sznupi]

That's to be expected if for many systems sensible upgrades are blocked. Anyway, if the group willing to upgrade is so small...why Intel blocks it?

Re:Figures...

[cynyr]

see AMD's AM2+/3 how many CPUS fit in those? i bet you can even drop a newer 6 core AM3 chips into a AM2+ and get at least 4 of the cores, even thin all that may be needed is a bios update.

Re:Figures...

[bdenton42]

Power bill dropped that much?

It probably did... it certainly runs a *lot* cooler. I'll have to see if it is actually noticable on the next electric bill.

Mostly I was amazed at how much higher the framerate is on the 8800GT. The same video card went from ~15 FPS with medium settings on Northwood to ~60 FPS on high settings on Core2.

Re:Figures...

[maxume]

All I'm saying is that it might be incidental. If they don't make physical and electronic compatibility a priority, then some other small design decision may trigger the change.

If you haven't noticed yet, 'consumers' are pretty much wholly on laptops and, maybe, integrated desktops. Very few people want to mess around with the insides (and most tinkering is going to be swapping in larger drives or more ram, and the ram thing is sort of going out the window).

Re:Figures...

[voidptr]

Because it's not worth the extra engineering effort to cater to the 3 people on the planet who would base a purchasing decision on it.

Re:Figures...

[sznupi]

What about extra engineering effort to block upgrades, eh?

Re:Figures...

[PitaBred]

With my current AM3 socket, I can upgrade to a 6-core AMD chip with just a BIOS update. Why can't Intel do that?

Re:Figures...

[sznupi]

I can agree it might be oversight...which doesn't speak very good of Intel if AMD manages to do it usually. Heck, I've seen a very late i865 ASRock motherboard with Core 2 Duo support. And latest Intel CPUs (as well as those upcoming in 2011) basically use just PCI Express and some interface to output video...

Yes, very few people want and expect to mess around with the insides...so why Intel, seemingly, does some effort to outright block such possibility?

Re:Figures...

[TheKidWho]

I can do that with my 1366 socket too.

Re:Figures...

[0123456]

With my current AM3 socket, I can upgrade to a 6-core AMD chip with just a BIOS update. Why can't Intel do that?

They could. But since only a tiny fraction of people ever upgrade CPUs, there's no reason to cripple your CPUs with support for old chipsets when you can just release a new one; every current AMD CPU has to support DDR2 RAM as well as DDR3, for example, and there's some evidence that requirement is significantly affecting AMD's memory performance with DDR3.

It would be different if AMD had better CPUs than Intel, but since Intel's are the fastest right now you can either buy the fastest CPU with the appropriate motherboard from Intel, or you can cheap out sticking your fancy new 8-core AMD chip into a dual-channel DDR2 system while Intel can release their 8-core chip with a quad-channel DDR3 motherboard.

Heck, the last two motherboards I bought from Intel came with the CPU soldered in.

Re:Figures...

[BrentH]

Even then, AMD manages DDR2 and 3 with the same socket, even the same chip.

Re:Figures...

[geekoid]

But doesn't get the performance or power savings.

It would be cheaper for Intel to not change the socket.

Re:Figures...

[drsmithy]

"Yes, let's force users to upgrade all their hardware when they want a new CPU! Show me the money!"

I'd be astounded if the percentage of people who have ever upgraded only their CPU was a meaningful fraction of 1%.

Re:Figures...

[drsmithy]

That's to be expected if for many systems sensible upgrades are blocked. Anyway, if the group willing to upgrade is so small...why Intel blocks it?

They don't block it, they just don't cater to it.

Re:Figures...

[Captain Centropyge]

Then you don't have a choice but to upgrade everything, now do you? It'd be cheaper/easier to just put it in a socket and lock the socket, like replaceable CPUs. Soldering it in forces people to upgrade the mobo and CPU, boosting revenue.

Re:Figures...

[sznupi]

It would be fun to determine where the line differentiating those lies.

But remember we're talking about a company which tried to block later s370 from working in slotkeys, activelly blocked possibility of dual cpu operation, used illegal tricks to ensure maintaining upper hand, goes far with castrating cheap CPUs from features, tried hard and succeeded in killing "alien" chipsets for their CPUs (hence ensuing profit for their own...now, how to ensure more will be sold?...)

Re:Figures...

[avandesande]

Well maybe they didn't want to spend money validating these configurations, so instead of dealing with the headache of customers creating unsupported configurations they make them physically incompatible.

Re:Figures...

[Ephemeriis]

And yet, somehow, AMD manages to keep their CPUs backwards compatible.

You upgrade the CPU/Motherboard/RAM. Big woop.

That, right there, would be enough to keep me from upgrading at all. Well, if I had an Intel CPU at least.

I don't have a whole lot of cash to spend on computer upgrades. I'll spend some on a RAM upgrade when I can afford it... And then some more on a new CPU when I can afford it... And later I'll buy a new motherboard when I can afford it...

That's worked great for the last several generations of AMD chips that I've run. A little research, some careful selections, and I've been able to upgrade everything bit by bit. I haven't had to go out and purchase all three at once in years.

Re:Figures...

[Ephemeriis]

You've had to do this for a while.

No I haven't.

I've been upgrading my AMD stuff piece by piece for the last several years... New CPU works fine in the old motherboard, new RAM works fine in the old motherboard, and then the new motherboard plays fine with the old CPU and RAM.

Obviously this won't keep working forever... Eventually AMD will roll out a new chip that's just plain completely different and I'll have to but it all in one fell swoop... But, knowing AMD, I'll then be able to keep upgrading piece by piece again.

Don't you remember having to get a new motherboard to use newer CPUs, even though they had the same socket?

Yup.

Then I stopped buying Intel chips.

No, I'm not claiming AMD has never done this... And I'm not claiming that Intel is solely to blame for it... But the fact remains that AMD has been far better at backwards compatibility than Intel has. Maybe it has something to do with all the motherboards that Intel sells...

Re:Figures...

[sznupi]

Almost any lack of interoperability could fall under that explanation. But does it really seem complete in the case of company which refused to license their CPU interconnect to 3rd party chipset makers? (compliance would be responsibility of those 3rd parties) Or sold Atom for higher price than Atom + its chipset...

Re:Figures...

[drsmithy]

But remember we're talking about a company which tried to block later s370 from working in slotkeys, activelly blocked possibility of dual cpu operation, used illegal tricks to ensure maintaining upper hand, goes far with castrating cheap CPUs from features, tried hard and succeeded in killing "alien" chipsets for their CPUs (hence ensuing profit for their own...now, how to ensure more will be sold?...)

All of which have clear and obvious benefits for segmenting the market and significantly increasing revenue.

Actively trying to block the fractional percentage of people who want to upgrade only their CPUs, however, is a completely different matter. There's simply not enough money involved for any but the paranoid to try and argue it would happen. It would probably cost more just in the people time of developing a strategy (let alone actually implementing it) than they would ever make back in increased sales.

It's like trying to argue that multipliers are disabled to stop people overclocking and "force" them to buy faster CPUs. It doesn't even pass a few seconds worth of cost/benefit analysis.

Intel have been well aware of the miniscule interest in CPU upgrades since the mid-'90s, with their "Overdrive" processors. Nearly everyone just gets a new PC, and of the tiny percentage left, most of them will use upgrading as an excuse to get a new motherboard and/or video card and/or RAM, etc.

Re:Figures...

[sznupi]

"Ovedrive" were a bit of a different case though - not only resulting in somewhat substandard performance, but also ridiculously overpriced (it was cheaper to just get the new mobo & ram...). Now the situation is different, with official Intel "tick tock" approach (it would be easy to anticipate future compatibility for process shrink...), quite stable situation with RAM & interfaces (yes, USB3...still hardly here, and the last version lasted a decade) and processors with 2, 4, and soon more cores.

Of course hardly anyone expects upgrading...but to what extent that's a reason, and to what an effect? Plus if there was hardly any money in chipset business (remember, using obsolete fabs), Intel wouldn't be so agressive in driving out any competition (which already had very little share of Intel platform)

Re:Figures...

[NervousNerd]

Actually, you must have forgotten about the LGA775 3.46GHz Pentium Extreme edition. Basically it's a Northwood (as it's 130nm, unlike Prescott which is 90nm and it only supports up to SSE2, unlike Prescott which supports SSE3) with 2MB L3.

Re:Figures...

[ThePhilips]

But since only a tiny fraction of people ever upgrade CPUs [...]

I know about at least one company which uses AMD CPUs in their workstations precisely for the reasons that few years later they can drop a new CPU (and more RAM) into it to accommodate advances (read "bloat") of the engineering software.

On Intel side of things, generally when you feel that you need to upgrade, you are forced to upgrade the PC completely: CPUs for the socket/slot you have are not produced anymore.

And since Intel is the largest CPU producer and most have Intel CPUs the stated "tiny fraction" is actually heavily biased and is the direct result of the Intel's strategy.

It would be different if AMD had better CPUs than Intel, but since Intel's are the fastest right now ...

Who the hell nowadays cares how fast the CPU is? I thin the "fraction of people" is even tinier than that of people upgrading CPUs.

My 5yo Athlon 64 X2 4200 (2x2.2GHz) runs pretty much everything even today. I have money now but I can't even find a justification to upgrade.

Re:Figures...

[Omestes]

I was in the same boat when I upgraded. All I really wanted, or could afford, was the processor and the mobo. I had my list down to a PhenomII x4 (955 Black), or an i7. The AMD setup cost around $650, where the Intel set-up would have been around a grand to replace all my RAM (8Gb) on top of the CPU and mobo. I personally didn't see any reason to switch to DDR3, even if I could stomach the cost, I've never noticed a problem with DDR2.

The i7 was drool worthy, but it wasn't that much better than the Phenom II, I'm not going to pay $400 for an almost unnoticeable performance boost.

But then again I generally find that the bleeding edge is nothing but an utter waste of money. If $100 can get you 95% of the way, is the extra 5% really worth another $100 on top? Perhaps if I was still young, and somehow thought that having the top specs at the time was somehow directly connected to my virility.

Though, to be honest, I hardly notice the difference between the Phenom II x4 3.2Ghz and the Intel Core 2 Duo 2.8Ghz chip it replaced.

Re:Figures...

[drsmithy]

Now the situation is different, with official Intel "tick tock" approach (it would be easy to anticipate future compatibility for process shrink...), quite stable situation with RAM & interfaces (yes, USB3...still hardly here, and the last version lasted a decade) and processors with 2, 4, and soon more cores.

There's been 3 different memory type changes (DDR1-3) and two bus changes (PCIe 1 and 2), to say nothing of the fundamental interconnect change between Core 2 and Core i7 & co. Chipset and socket changes don't seem unreasonable along with any of them.

Of course hardly anyone expects upgrading...but to what extent that's a reason, and to what an effect?

Few people upgrade because to most of them a computer is an appliance, not a bunch of individual components. "Upgrading the CPU" to most people would make about as much sense as "upgrading the motor" in their vacuum cleaner.

Even amongst the few people I do know who upgrade their PCs, I can't think of single one who has ever upgraded only the CPU, even when that was possible and would have been a significant raw performance improvement in and of itself.

Plus if there was hardly any money in chipset business (remember, using obsolete fabs), Intel wouldn't be so agressive in driving out any competition (which already had very little share of Intel platform)

Pretty much all non-intel chipsets have been crap - and that's a phenomenon dating back to the days of the original Pentium. There are very, very good reasons there's not a lot of competition in that market, and 90% of them involve bad (sometimes incredibly bad) products. Low quality chipsets are also one of the major reasons AMD does not have a larger slice of the pie (VIA did more to harm AMD than Intel could ever have hoped to in its wildest dreams).

Re:Figures...

[sznupi]

Realistically, in the timescale involved, only one type of memory change (which is still ongoing), one which doesn't matter much in context (since even upcoming CPUs from AMD will be able to use old one, and since Intel made clean break together with other important changes). But most importantly nowhere near as problematic as in the old days, with incompatibilities frequently introduced (heck, even modules of hypothetically correct type would often refuse to work...). Both PCIe are fully compatible, and without much difference to the user really (not across one generation). Core 2 & Core i7 - sure, that change was reasonable. But the one from TFA not even close.

Yes, few people upgrade...again, we can't really know what is fully the cause and what is the effect (they don't have to upgrade by themselves, you know; "making computer faster cheaply" is easy to understand). But you have to agree there has been, over the years, quite a lot of waste with thrashing perfectly good machines... (quite recently mostly revolving about low RAM, which was often easily corrected; changing CPU could be often icing on the cake)
And where have I said you have to upgrade only the CPU? Sure, throw in more RAM and new HDD, all the better. If you can, of course...

Is "pretty much all non-intel chipsets have been crap" the reason why, until quite recently, Intel has produced motherboards with SiS chipset? Is that why many recent Macs are on non-Intel chipsets and doing fine? (accidentelly, chipsets from a manufacturer which didn't get license extension...)

Re:Figures...

[indi0144]

faster chip =! quality chip

So, you think that a car that can't go from 0 to 60 in less than 6 seconds it's crap?

>> There are very, very good reasons there's not a lot of competition in that market, and 90% of them involve bad (sometimes incredibly bad) business practices by Intel bribing OEMs to not use AMD chips.

FTFY

I really can't see how AMD chips since Athlon fail in that abused marketspeach term "quality" I haven't read news about faulty AMD CPUs overheating and causing fires neither massive bugs that corrupt data (Which intel had) because thats what I'd call a "quality" failure. Maybe you can fool the average consumer with that arguments, that doesn't make them real facts.

By your logic then we can be positive about the fact that Microsoft Windows is truly the best OS in the world by lack of quality competition? Or is the other way around, both Linux and Mac OSX are WAY better than Windows just that Microsoft have a monopoly in there, this is the same as intel.

Both AMD and Intel make good quality chips, both chips are used by servers and high performance systems. Intel can make them run faster and AMD can make them consumer/pocket friendly, which target market are you?

Re:Figures...

[petermgreen]

Afaict the reason desktop CPUs aren't soldered to the motherboard (many laptop CPUs are because thickness is at a premium) is because it would mean OEMs would have to decide what CPU was going in a machine much earlier. This would both require the OEM to hold more stock (bad for the OEM) and make the OEM less likely to offer higher end CPUs as build to order options (bad for intel).

With socketed CPUs an OEM can design and build most of a system and only decide late on in the process (sometimes even only when the customer actually buys it) what CPU to drop in and hence where in the lineup it will come.

Re:Figures...

[drsmithy]

Realistically, in the timescale involved, only one type of memory change (which is still ongoing), one which doesn't matter much in context (since even upcoming CPUs from AMD will be able to use old one, and since Intel made clean break together with other important changes). But most importantly nowhere near as problematic as in the old days, with incompatibilities frequently introduced (heck, even modules of hypothetically correct type would often refuse to work...). Both PCIe are fully compatible, and without much difference to the user really (not across one generation). Core 2 & Core i7 - sure, that change was reasonable. But the one from TFA not even close.

You're missing the point. All of these changes could quite reasonably result in a new socket design, if no particular effort was being made to get them working with existing sockets. Making no effort to retain backwards compatibility (because the cost/benefit analysis says it's not worth it) is a significantly different intent to actively breaking backwards compatibility. There is much evidence and reasoning to support the former situation, and very little to support the latter.

Yes, few people upgrade...again, we can't really know what is fully the cause and what is the effect (they don't have to upgrade by themselves, you know; "making computer faster cheaply" is easy to understand).

We can quite easily infer the cause and effect simply by observing that outside of computer enthusiasts, virtually no-one upgrades *any component* of their PC at all - when they want a more capable machine they buy a new one. There is zero evidence to suggest waves of people eager to upgrade their PC's CPUs, if only Intel made a more backwards-compatible socket.

But you have to agree there has been, over the years, quite a lot of waste with thrashing perfectly good machines... (quite recently mostly revolving about low RAM, which was often easily corrected; changing CPU could be often icing on the cake)

That has zero to do with whether or not Intel happens to support multiple CPU families on a single socket.

And where have I said you have to upgrade only the CPU? Sure, throw in more RAM and new HDD, all the better. If you can, of course...

The whole discussion is about Intel not keeping backwards compatibility in their CPU sockets, thus meaning upgrades can't be only the CPU. If you're [also] going to upgrade the motherboard/RAM type/video card/whatever, then the criticism that you can't replace just the CPU is moot.

Re:Figures...

[sznupi]

You're overextending the point. New (and nowadays quite fully backwards compatible) peripheral buses resulting in new cpu socket? Really? RAM is less clear of course, and any "forced" change can be understandable here; but again AMD shows it can easily be done. Remember this isn't black and white, about trying to keep absolute compatibility for as long as possible. Only where it's reasonable. It would be certainly reasonable to expect in the case of TFA. A socket wchich was touted as neccessary for the market (not long after introducing another one), that will live barely more than a year - in light of many murky practices of Intel, does scream that it's intentional obsolescence (why it had to be leaked? Why Intel isn't forthcoming with this information? Could it be that consumers now would prefer something with a little more longevity?...)

Haven't it crossed your mind that what you observe is mostly a peculiarity of your local market? Quite wastefull, all things considered, when comparing with many other places where PCs are bought (ah, but I have to concede it has nothing to do with the issue, just becuase you want it so...)

And I didn't mention motherboard there. It was just mentioning that, apart from swapping the CPU, literal addition of those two parts to empty slots and bay will indeed contribute even more to the new life of any machine (if going from 2 to 4 or 6 cores at higher clockspeed wasn't enough...). But for some reason you needed to twist it in a way that's basically just a total swap comparable to the least ideal situation...

I'm dissapointed you haven't dragged on with "3rd party chipsets = bad". After all, Apple and Intel (sic) certainly didn't know what they're doing...

Re:Figures...

[nxtw]

The set of users that upgrade anything is pretty small. The set of users that upgrade CPUs is even smaller.

And the set of users who upgrade systems with integrated graphics is even smaller - that's what this socket is primarily used for.

Re:Figures...

[drsmithy]

You're overextending the point. New (and nowadays quite fully backwards compatible) peripheral buses resulting in new cpu socket? Really?

If they use a new chipset and no effort is being made to retain backwards compatibility (eg: by increasing the bandwidth from chipset to CPU in the new models), then sure, it's quite reasonable.

RAM is less clear of course, and any "forced" change can be understandable here; but again AMD shows it can easily be done.

I don't doubt that it's doable if the requirement exists. My point is that within Intel, the requirement obviously *doesn't* exist.

Remember this isn't black and white, about trying to keep absolute compatibility for as long as possible. Only where it's reasonable. It would be certainly reasonable to expect in the case of TFA. A socket wchich was touted as neccessary for the market (not long after introducing another one), that will live barely more than a year - in light of many murky practices of Intel, does scream that it's intentional obsolescence (why it had to be leaked? Why Intel isn't forthcoming with this information? Could it be that consumers now would prefer something with a little more longevity?...)

In what way is it "intentional obsolescence" in the context of there being no market requirement defined for compatibility to be kept ? On what basis do you make that claim that someone inside Intel redesigned the socket with an explicit objective of making it incompatible with existing hardware ?

Haven't it crossed your mind that what you observe is mostly a peculiarity of your local market?

Over the last 15 years, my "local market" has crossed half a dozen different countries - and that's not including the people I interact with online. I'm pretty confident the sample is representative (if anything, given the field I work in, I expect the proportion of people I see upgrading is vastly *over* represented within my peer group).

Quite wastefull, all things considered, when comparing with many other places where PCs are bought (ah, but I have to concede it has nothing to do with the issue, just becuase you want it so...)

Quite wasteful in what sense ? All those old PCs usually don't get tossed, they get recycled to friends and family, sold on ebay, or donated to charity.

And I didn't mention motherboard there. It was just mentioning that, apart from swapping the CPU, literal addition of those two parts to empty slots and bay will indeed contribute even more to the new life of any machine (if going from 2 to 4 or 6 cores at higher clockspeed wasn't enough...). But for some reason you needed to twist it in a way that's basically just a total swap comparable to the least ideal situation...

I didn't "twist" anything. I pointed out that of the tiny minority of people who do upgrade their PCs by replacing components, they typically do multiple components at a time.

I'm dissapointed you haven't dragged on with "3rd party chipsets = bad". After all, Apple and Intel (sic) certainly didn't know what they're doing...

I never said "3rd party chipsets = bad". I said *most* non-Intel chipsets have been awful, buggy atrocities. This was particularly true in the '95 - '05 timeframe, and while the situation was improved markedly by nVidia and ATI's entrance into the market, both of them had some screwups as well.

The fact that Intel have sold motherboards with SiS chipsets is irrelevant. They sell motherboards with those godawful Realtek NICs on them, too (and they should also be avoided like the plague). The only thing I commented on was Intel's *chipsets*. Apple, of course, have a lot more control over their platform and are able to implement workarounds for problems and push them to a large proportion of their customers, something that is much more difficult in the PC world.

Re:Figures...

[sznupi]

Those new Intel CPUs use PCIe to connect to "southbridge". PCIe is designed to be explicitly backwards compatible across versions (yeah, there will be less bandwith to the "southbridge" available...irrelevant in nearly all scenarios).

Obviously the requirement doesn't exist if its something quite opposite is followed...the question is why Intel chooses so. Based on their past actions, I have my doubts that what they they do is all around optimal. This is a good reason for suspecting "intentional obsolescence"...Intel has already done so quite a few times (no, don't limit yourself only to cpu sockets, if you really want to have vert clear examples...)

And please, stop with portraying like it's a solid fact that there's no demand for greater (this all I'm talking about - greater, not absolute) upgradeability. Places I visited largely disagree with you. So there...
You're also kidding yourself if you think a lot of perfectly fine machines isn't trashed.

Even multiple components doesn't have to include so many major ones as you want to believe...

If you want to paint 3rd party chipsets that way now, please mention also major screwups of Intel in chipset area...
(plus - Intel sold motherboards with SiS chipsets because they were damn good, I can assure you that; and Apple isn't the only choosing Nvidia chipsets)

Re:Figures...

[drsmithy]

Those new Intel CPUs use PCIe to connect to "southbridge". PCIe is designed to be explicitly backwards compatible across versions (yeah, there will be less bandwith to the "southbridge" available...irrelevant in nearly all scenarios).

How about adding more lanes than are in the original spec ?

Obviously the requirement doesn't exist if its something quite opposite is followed...the question is why Intel chooses so.

Because not doing so reduces costs ?

Based on their past actions, I have my doubts that what they they do is all around optimal.

What do you mean by "optimal" ? "Optimal" for whom ?

This is a good reason for suspecting "intentional obsolescence"...Intel has already done so quite a few times (no, don't limit yourself only to cpu sockets, if you really want to have vert clear examples...)

For example ?

And please, stop with portraying like it's a solid fact that there's no demand for greater (this all I'm talking about - greater, not absolute) upgradeability.

I'm not. I'm observing that it is a "solid fact" very few people upgrade systems at all, and even fewer are only interested in upgrading the CPU.

Even multiple components doesn't have to include so many major ones as you want to believe...

It's got nothing to do with what I believe, it's got to do with what I've been observing computer enthusiasts actually doing over the last 15 years. There is very little evidence to suggest a general desire across the PC market for greater upgradeability.

Look, seriously. The whole argument here is that there's a group of people in Intel who specifically set out with each new CPU design to make it incompatible with existing hardware, because by doing this they hope to squeeze more profits out of... the fraction of 1% of users who are interested in CPU upgrades. If that doesn't make you go "WTF ?" straight off the bat, then there's absolutely nothing I'm going to be able to say that convinces you otherwise.

Re:Sigh

[MrNaz]

There's always AMD's Fusion on the horizon. If they can execute well on that they have a chance to do what they did with the Athlon. Intel has yet to demonstrate that they actually have GPU tech that can compete with nVidia and ATI in this space. I really hope they do, Intel has had too long at the top of the market and they're getting all monopolistic again.

Re:Sigh

[MrNaz]

As in, I hope AMD can execute, not I hope Intel have tech that can compete with nVidia and ATI. The former would lead to better competition, the latter would give the monopolist more power.

That'll teach me to not preview.

Re:Sigh

[MrNaz]

Gah! I meant "that'll teach me to preview".
Someone pass me a mallet. My head seems to need a little percussive maintenance.

And yet,...

[Pojut]

...the AM2+/AM3 socket on my AMD board continues to be useful for new AMD CPUs literally years after I originally purchased it.

Re:And yet,...

[Kjella]

Same socket, but can it run all the newer processors? That at least happened to be with a Shuttle I had that I thought about upgrading - for various reasons with the board it couldn't even with a BIOS upgrade. And there always seemed to be some sort of shift like AGP to PCIe, PATA to SATA, DDR2 to DDR3, USB 1.0 to 2.0 or some various other good reasons to upgrade anyway. Expansion cards are just silly expensive compared to motherboards, I'm guessing due to volume.

To take one example, any decent mobo today comes with 6+ SATA-II ports. As expansion cards the cheapest 4-port SATA-II controller is 439,- NOK. I can get a full new P43 motherboard with 6 channels for 499,- NOK. I guess it works nice if the CPU is the one and only thing you would like to upgrade. By the way, Intel really should put their 6-disk controller on an expansion card and kill the competition, don't understand why they don't.

Maybe for once things are finally stable enough to be worth it... but for my part, I've been burned many enough times that I don't even think about future compatibility - I only look at what I want here and now and if that hardware will last.

Re:And yet,...

[MadKeithV]

Pointing out the obvious, it was obviously not necessary to point that out.

Re:And yet,...

[Pojut]

Yup, it can do it all...the ONLy thing i can't do is run DDR3 (it has four DDR2 slots), but other than that I can take care of all the new stuff (exceptin' USB3.0 and Sata6, of course...but not much on the market can do that yet either)

Re:And yet,...

[Jazz-Masta]

...the AM2+/AM3 socket on my AMD board continues to be useful for new AMD CPUs literally years after I originally purchased it.

Intel had a long run with the Socket 775 boards, and AMD pulled this stunt back with their Socket 939 to AM2 upgrade. AM2 is a 940 pin socket.

I do agree AMD did something right with their AM2, AM2+, AM3 sockets being interchangeable for many CPUs. Just some of the more interesting features get disabled when running an AM3 cpu on an AM2 socket.

Re:And yet,...

[Rudeboy777]

Almost every board on the market does USB3 and SATA6 except at the lowest end. The chipsets don't support it, but motherboard manufacturers put additional chips on the motherboards for both

http://giga-byte.ca/Products/Motherboard/Products_Overview.aspx?ProductID=3284

I fully agree however that a 3 year old AMD motherboard with a new CPU gives you just about the same experience as brand new system as long as the motherboard OEM provides ongoing support through BIOS updates. I'm a loyal Gigabyte customer for this reason.

Re:And yet,...

[geekoid]

sooooooo... no.

Re:And yet,...

[gilboad]

I can't vouch for other manufacturers, but I've got a number of Gigabyte GA-M55S-S3 and GA-M56S-S3 boards that came with single core Athlon64's or low-end Athlon64X2's.
Most of the GA-M55S-S3's have been upgraded to high-end Athlon64 X2's (AM2) or 95w Phenom X4 CPUs (AM2+, Agena core).
The somewhat younger GA-M56S-S3's are far more capable (latest BIOS added AM3 support), I plan to upgrade all of them to Phenom II X4 CPUs (AM2+ or AM3 Deneb), or better yet, if Gigabyte continues to release new BIOS' for these old boards (Some of them are close to 3 year old!), I may even upgrade one or two of them to Phenom II X6 CPUs.

- Gilboa

Re:And yet,...

[RzUpAnmsCwrds]

Intel had a long run with the Socket 775 boards, and AMD pulled this stunt back with their Socket 939 to AM2 upgrade. AM2 is a 940 pin socket.

939 uses DDR, AM2 uses DDR2.

The only reason that AM2+/AM3 are compatible is that newer AMD CPUs support both DDR2 and DDR3.

Re:And yet,...

[nxtw]

...the AM2+/AM3 socket on my AMD board continues to be useful for new AMD CPUs literally years after I originally purchased it.

Sure. And you can upgrade to a CPU about as fast as what you can install in a literally years old LGA 775 system...

A win for AMD

[Albanach]

I can't understand why they would force another socket design on customers. I am using a four year old motherboard and recently replaced my AMD CPU with a current model. It was a drop in replacement. Sure I could get some benefits from a newer MB, but I can make the upgrade at a time of my choosing. I can spread the cost, get the big boost from the CPU now and get a smaller boost from a new MB in a year's time.

Board manufacturers have to spend money implementing the new socket. Retailers are stuck with old stock that no-one wants because a new socket is around the corner.

It raises prices and hurts the end user. Why are we still seeing this behavior?

Re:A win for AMD

[CrimsonAvenger]

It raises prices and hurts the end user. Why are we still seeing this behavior?

I think you answered your own question in the first three words of the question....

Re:A win for AMD

[NevarMore]

It raises prices and hurts the end user. Why are we still seeing this behavior?

It raises prices and helps Intel.

Re:A win for AMD

[Rogerborg]

Uh, perhaps because renegades like me and thee - heck, we're probably filthy hackers, and we may even have links to organised crime - who upgrade our systems are an insignificantly small market, and Intel are happy to cede it to AMD in order to squeeze more profit out of the other 98% of their customers?

Re:A win for AMD

[PhrstBrn]

Because Intel sells motherboards and chipsets too. They don't want to sell you just a new processor, they want to sell you a new processor and a motherboard.

If Intel thought they could make more money by keeping their stuff backwards compatible, they would, but I'm sure the bean counters figured the amount of sales lost to AMD would be less than the profits they could make by forcing you to buy new motherboards too, and I would tend to agree with that.

I don't like it, I don't think it's good for consumers, but it makes sense from Intel's perspective.

Re:A win for AMD

[SharpFang]

Board manufacturers get to push a new board model for people who want to upgrade the CPU.

I upgraded a CPU once. The CPU required a new motherboard. The new motherboard required new RAM and new gfx card. And the new components combined required a new PSU.

Pure business.

Re:A win for AMD

[TOGSolid]

Intel is riding high right now and thinks that everyone will fall right in step no matter what Intel does. They're getting greedy, pure and simple and it's about to bite them in the ass.
I was going to write something completely different after this, and then I read the last line of the article.
"Oh, one last thing: one of our sources states LGA2011 will launch with quad-and six-core CPUs (with Hyper-Threading so eight and 12 execution units) although another source has stated eight-core CPUs are also on their way"
Really? They couldn't do anything else with Socket 1366? Good job Intel, you're going to manage to piss off a lot of people all at once. Oh well, it was time for the playing fields to reverse again. Sure Intel's processors may be the biggest and baddest on the block, but if I have to switch out my motherboard every time they issue a new generation of processors, then I think I'm going to switch back to the AMD camp (I always did have a soft spot for my old Athlon 2800+ rig. That thing was unbreakable.)
Funny how both GPU and CPU leaders seem to swap at the same time. Nvidia just recently shot itself in the foot with their 400 series cards, and Intel is about to do the same with yet another generation of sockets. Meanwhile Radeon is riding high with their latest offerings from the 5000 series and if they play their cards right, AMD is going to end up stomping Intel if only due to the fact that consumers are going to get tired of Intel's shit.

Re:A win for AMD

[avandesande]

I can understand- 99% of cpus are never upgraded, so why should INTEL waste engineering resources on the 1% that do?

Re:A win for AMD

[Jeff DeMaagd]

I think it cuts both ways though, if I have to upgrade most of the computer just to upgrade one part, I'll just put it off.

Oh, I prefer my toast in bites!

[angiasaa]

Honestly, this would work for just the corporate users, and perhaps the oldies who don't go FPS'ing around the place. It would also do for mobile apps at a cinch. But for all practical purposes, it would seem that Intel is taking a step back into the netherrealm.

The additional surface area offered by a separate GFX chip allows it to cool faster. Frankly, I'd rather slap on a separate GFX card altogether, than waste transistors in my main processor for physics and pixel processing.

Keep the space for cache or add some more muscle to the chip, but don't go stuffing graphics or audio processing in there. The BUS speeds today are good enough to handle the stuff we throw at them with separate chips, so there!

If my Graphics chip blows, I can always replace it. If I somehow manage to fry my processor, I can replace that. Why replace everything if one goes kaput?

Re:Oh, I prefer my toast in bites!

[tepples]

Frankly, I'd rather slap on a separate GFX card altogether, than waste transistors in my main processor for physics and pixel processing.

Or in a desktop PC, you could have a GMA running one monitor and a GeForce running the other.

Re:Oh, I prefer my toast in bites!

[bluefoxlucid]

It'll never catch on. Well, it'll catch on fire.

Re:Oh, I prefer my toast in bites!

[Calinous]

If your sound card blows, you could replace it. If your mainboard blows, you could replace it. Why replace everything when one goes kaput?
Just as discrete CDROM controllers (ISA) went the way of the dodo, just like IDE controller boards did the same, just like the network cards got integrated...

to bad it's the same gma crap that amd has a bette

[Joe The Dragon]

to bad it's the same gma crap that amd has a better on board chip and plane to work on getting in the cpu + letting it boast a add in ati card as well. what will intel card do just shut down when a better card in installed?

Real-time high-def geometry rendering

[tepples]

Who needs a proper graphics card these days?
Only people who need real-time high-def geometry rendering.

More people will need this than you might think. Let's look at each piece of your claim:

• Real-time: Graphical user interfaces must respond instantly to the user's commands. Newer systems have added window entry/exit effects to clarify the relationship among various items on the screen.
• High-def: PC GUIs have met the definition of high-definition video since the 1990s, when XGA resolution (1024x768px) became common.
• Geometry: I'll admit that most office and web applications currently use much simpler geometry than a typical Xbox 360-class 3D game. But as IE 9 becomes popular, SVG and <canvas> will finally become viable, and even font rasterization will move to the GPU.

Re:Real-time high-def geometry rendering

[T-Bone-T]

I'm using a desktop that I recently built with a Core i3-530 and the built-in graphics are quite acceptable, even at 1600x900(the monitor was free so I won't complain about the odd resolution). The only place they suffer is in high-performance areas like games. The IGP is designed to process Full HD video.

Re:Real-time high-def geometry rendering

[Firethorn]

More people will need this than you might think. Let's look at each piece of your claim:

I think that the issue here is where you place the line on a 'proper' graphics card.

By that I mean that today even integrated video cards are easily able to keep up with GUIs, play even blue-ray movies, etc...

I'm not sure SVG/Canvas, rasterization will really bog down modern integrated graphic engines. Or if it doesn't support it, it'll fall back to the CPU, and assuming you're not doing anything too CPU intensive at that moment, it won't matter. You don't need a 5870 to run Office or IE.

Re:Real-time high-def geometry rendering

[timeOday]

Geometry: I'll admit that most office and web applications currently use much simpler geometry than a typical Xbox 360-class 3D game.

I would be shocked if the GPU integrated into Intel's next-gen CPU doesn't blow away what's in the XBox 360, which after all is a medium-high end card from 2005. And yet, you'll notice there seems to be little push for next-gen consoles beyond the XBox 360 and PS3. Integrated video will eventually be good enough for most applications including games, the question is not if but when.

Granted, top GPUs have transistor counts comparable to top CPUs. On the other hand, the 8-Core Xeon Nehalem-EX has as many transistors as an NVidia GT200 PLUS a Six-Core Opteron 2400, which sounds to me like a perfectly fine gaming rig.

Re:Real-time high-def geometry rendering

[Bakkster]

Don't forget that modern Operating Systems are able to store the contents of windows in the Graphics memory, leaving more room in system RAM for other items. This also speeds up the loading of less recently viewed windows (by having them cached less often).

One freaking pin?!

[Hatta]

How about you design the next socket with twice as many pins as you think you'll need? Then we won't run out and have to buy a whole new motherboard when we just want a faster CPU.

Re:One freaking pin?!

[GungaDan]

The new one has one FEWER pin than the current socket. So obviously next time they should either design one with a single removable pin, or no pins at all.

Re:One freaking pin?!

[geekoid]

How about you study ship design before opening your mouth so we won't have to hear your nonsense?

Re:One freaking pin?!

[angiasaa]

No pins on a processor? They already have that! It's called a Land Grid Array Socket.. or LGA for short.

http://en.wikipedia.org/wiki/Land_grid_array

Re:One freaking pin?!

[GungaDan]

Hmmmmm... would that be the same "LGA" referenced in the LGA1155 socket with its 1155 pins, or the LGA1156 socket it's replacing, with its 1156 pins? Or is that a different LGA entirely, possibly associated with women's golf or a lesbian and gay alliance?

Re:One freaking pin?!

[nedlohs]

Yeah, because the old motherboards will magically do the right thing with the pins you weren't using when they were built. And with the pins that now serve a completely different purpose.

Re:One freaking pin?!

[angiasaa]

Yup, the very same unfortunately. :(

Here're a couple of high-res images of the sockets:
http://techreport.com/r.x/clarkdale/lga1156-socket.jpg
http://www.tomshw.it/guides/hardware/cpu/20090905/images/Asus%20P55%20LGA%201156.png

Frankly, I think it's just an attempt at making MoBo manufacturers bear the brunt of the expenses instead of Intel having to dig into its buffer stock of gold bullion. :| It costs a whole ton more to build an LGA socket than it does, to build a PGA socket.

Having said all that, I tend to lean toward the theory that it's got more to do Women's Golf than a Lesbian and Gay Alliance. :P

Duhhh!

[spammeister]

Of course we all saw this coming. It's what Intel does time immemorial.

I'm sure all the intel and AMD fanboys will do what they do. They chose their camp, now they gotta take their lumps.

Re:Duhhh!

[geekoid]

Because the more advantages technical reasons don't mean anything?

Re:to bad it's the same gma crap that amd has a be

[Calinous]

http://www.anandtech.com/show/2972/the-rest-of-clarkdale-intel-s-pentium-g6950-core-i5-650-660-670-reviewed/2

i5-661 (with the fastest on-package graphics) is performance-competitive with AMD's latest integrated graphics. The slower on-package GPU from Intel are behind, but not by much. Nothing Intel can't solve in its next processor (especially as AMD did not increase its IGP performance)

Re:Sigh

[MoldySpore]

I really hope that AMD gets back on top and can compete with Intel on the top-level CPUs again. I am tired of the Intel fanboy's crapping all over AMD for the last few years, and really the industry NEEDS AMD to get back on top and help drive the price of these Intel chips down. The price gap is so huge between AMD and Intel that it makes building a top of the line Intel machine very daunting for us working-class enthusiasts and system builders.

Thankfully AMD's new hexacores will work in AM3 sockets so a motherboard upgrade isn't necessary at least for the Phenom II X6's. To me that is a big deal. I think it will be for a lot of others as well.

Get a Mac

[Anonymous Coward]

No more CPU upgrades problems! /duck

(only half-kidding, though)

This simplifies cooling design so much!

[Kazymyr]

I'm sure that combining the two biggest heat sources in a computer on the same die is a very well thought move. Especially for mobile versions. Yay.

Re:This simplifies cooling design so much!

[Nerdfest]

Only one thing to cool now ...

also amd HT is in all CPU's unlike Intel that only

[Joe The Dragon]

also amd HT is in all CPU's unlike Intel that only has there in high end cpus.

so intel low end cpu are stuck with low pci-e lanes to the point where usb 3.0 can get in the way of x16 video cards make some boards use a pci-e switchs. and foreing apple to use core 2 in there 13" laptop just to get good video with needing to add full video chip + chipset.

Intel also uses this to lock out NVidia. They should put there new bus in the i3 i5 i7 (low end) and not crap GMA video + 16 pci-e lanes.

This why form day 1 apple should of used amd as 1st mac pro had less pci-e then the g5 had. If apple had used amd back then they could of had a system with a lot pci-e + maybe even a nvida sli chipset.

This is why apple is thinking about useing AMD.

Re:Sigh

[caerwyn]

I don't think the GP is upset at *Intel* in this regard; I think it's more a perfectly realistic consumer complaint: "I wish there were more competition in this space because that would be better for me as the consumer." AMD dropped the ball pretty badly after a very strong run with earlier Athalons. It'd be great to see them get back into the game and really help push things along again.

The processor is only one part of performance

[CFD339]

A large part of the performance gain in new generation processors is actually the combination of the processor and chipset. The core i5, core i7, etc. processors did away with a a separate memory controller -- that itself has been a huge power and speed advantage. Without upgrading the stuff supporting the chip, you don't get much benefit from an upgrade.

Re:The processor is only one part of performance

[cynyr]

sort of like AMD did a number of years ago?

16 pci-e lanes to low when the chipset lacks usb 3

[Joe The Dragon]

16 pci-e lanes to low when the chipset lacks usb 3, and other things like sata 3.0 and other new buses fores MB makes to use switchs and other stuff to fit in video + sata 3.0 + usb 3.0 or cut down the video card to x8.

Re:Sigh

[sznupi]

Intel, through illegal practises, prevented AMD from benefiting fully from their lead with K7 and early K8 Athlons. This illegally rerouted money weakened AMD R&D and fabs, while strenghtening Intel ones at the same time.

Re:Sigh

[sznupi]

AMD dropped the ball pretty badly after a very strong run with earlier Athlons. It'd be great to see them get back into the game

(corrected one product name...)

That's not so simple. How much of "AMD dropping the ball" was because if illegal, anticompetitive practices of Intel? Practices which, essentially, robbed AMD from money needed for aggresive R&D and fab expansion.

Re:Sigh

[bberens]

Maybe it's changed in the last couple of years but my understanding was that AMD was still the place to go for database type machines because of the bus speed but Intel was the way to go for number crunching app machines. At the time I did the research and tried to explain that to my manager but it didn't matter because the Core 2 Duo was kicking the pants off AMD on the personal computer so Intel was faster and that's what we were getting for all the machines.

Changing sockets sounds bad, but

[obarthelemy]

I've never really upgraded CPUs. By the time my CPU is outdated (2-3 years), my motherboards usually is, too: newer RAMS (SDR - > DDR -> DDR2 -> DDR3), faster HD interfaces ( PATA -> SATA -> SATA2 -> SATA3) and others (USB -> USB2 -> USB3; PCI -> PCIE -> PCIE2), bigger/faster HDs... In the end, I usually rotate entire PCs, they go My Main PC - > My Backup PC -> My parents / Niece.

My gripe with Intel is more about the price of their MBs, especially compared to AMD's. The cheapest AMD MB within an AMD IGP is listed at 54 euros at my favorite retailer ( Asus AMD2+, not 3, but perfs are broadly the same), while Intel's cheapest MB is 84 euros (Gigabyte). Their low-end CPUs are also kinda expensive. And their IGPs also still kinda suck, even for playing video, and definitely for even light gaming.

The interesting thing these days is smaller size. Mini-ITX mainboards are becoming common, there's cheapish ones with AMD2/3 or 1156 sockets, good cases (Silverstone...), huge HDs. Unless you really need a graphics card, you can build a very small and quiet PC.

Re:Changing sockets sounds bad, but

[Anonymous Showered]

That is exactly how I upgrade my computers. I usually buy the CPU/RAM/Motherboard at the same time, as chipset technology also involves. Video cards are every 2 years: I plan to keep my 8800GT for my i7 upgrade. Later this year, I'll pick up two new nVidia GTX 470 or 480s.

However, at this point, I am more interested in the future of LGA 1366...

Re:Changing sockets sounds bad, but

[Mashiki]

Did you miss the days of socket7 when everyone used one socket design, and you could upgrade inline for 2-3 years with no issues. And then there was the slot design. Both AMD and Intel used the same design, except reversed 180. Then there was RAM, you could use various types of 30pin in later designs(as long as it met the voltage requirements) and 72pin again the same deal. The real change when DIMM's first came available but even then they lasted for ages, the only limiting factors were: Had to be same voltage(unless the MB supported variable voltage by pin/DIP switches), and had to be "near or close" to the same NS rating.

Wrong title

[unity100]

Title should have been "Intel wants to sell you motherboards and shit, along with their new cpu line" ...

i really got tired of this old trick.

Re:Sigh

[DJRumpy]

Someone help me here. Although I understand the basic need for new sockets sometimes, I dont understand the drive to implement them so often and frequently. If there truly is only a single pin difference, wouldn't it make sense to at least attempt to design the chip to meet existing sockets? It also seems like it would speed adoption of a new processor if it is socket-compatible with existing motherboards. This is the piece that confuses me. It seems like any time a new socket is required, it's bad for business, unless you happen to be in the motherboard business.

What gives?

because design includes the interface

[YesIAmAScript]

The design of a CPU includes the way it interfaces to the motherboard. If you make a new CPU on the same interface (bus), you don't get full performance. And you can't optimize power either. And it buys you very little to not pair the two up. Very few people upgrade their CPU, they usually buy a CPU with the motherboard and don't change it until they get a new motherboard.

And heck, few people even buy their own motherboard anyway! People who build their own systems don't realize how few people do so now. It can't be over 20% of the market worldwide, and a lot less than that in the US.

Re:Sigh

[interval1066]

I dont understand the drive to implement them so often and frequently.

The cynic in me says 'Yeah, its a money grab; new cpu = new socket, and the user shells out for new hw, Intel wins big until everyone else reverse engineers the whole set up and sells clones."
But then the rationalist in me knows that socket aside, the new cpu is going to require a new chip set, especially if the package contains a gpu as well. So its kind of moot. Plus, anything they can do to remove pins is generally a good thing. Generally. One less pin sounds like they've made something more efficient.

Re:Sigh

[HarrySquatter]

So basically AMD's failures are always Intel's fault and not their own, right?

Re:Sigh

[jvin248]

Won't matter for either Intel or AMD what they do, until they rework their marketing and naming programs. Average consumers don't know what i5 or i7 or athlon or duron or celeron means. Back in the days that clock-speed was it, people could grade their purchase "I'll pay 20% more for a machine that has 20% more cpu speed/performance". Too hard to tell. And cpu performance is beyond what most people need (email, checkbook spreadsheet, surfing) - so they shop (and are sold) pc's based on price, OEM computer brand name, and HDD storage size.

It wouldn't have to be clockspeed - but it has to be something related to performance. Then they have to educate the retail outlets on how to market the technology cleanly and clearly - they will sell more.

Re:Sigh

[WaroDaBeast]

Dude. Intel v. AMD. There's a clear reason why it happened. 'Nuff said.

Re:16 pci-e lanes to low when the chipset lacks us

[TheKidWho]

It's PCI-e 3.0 which is 2x faster then 2.0

Re:Linux

[TheEldest]

AMD/ATI has started a new open source driver project for their video cards (http://www.anandtech.com/show/2338). AMD is at least working for good linux support though it may not necessarily be up to snuff today (I haven't looked at benchmarks between win/linux on AMD/ATI's cards).

At least they're trying?

Re:Sigh

[Antiocheian]

It has been tried before. Cyrix did it with their "+" marking. You would get an 166+ Cyrix processor which would supposedly perform better than an Pentium 166. But when people started playing software rendered 3D games on their systems they noticed performance was less than Pentium 133. AMD did it a few years later. But then Intel started producing better chips (abandoning the Netburst joke) and the + marking became obsolete once again.

It's impossible to tell from a single number anymore because of differences in architecture. Some processors offer a better score in generic benchmarks, other get a better score in games, others in encoding. Intel chips generally do very well in encoding. AMD are quite competitive in games. The only way to figure out what you are getting is to find a benchmark for your applications. That's why hardware sites are so useful and have generally gone through the dot com bubble unharmed.

Re:Sigh

[sznupi]

The only major "failure" of AMD is that Intel lately managed to be noticeably ahead in both R&D and fab technology; there is nothing particuarly wrong with AMD tech, nothing that says "failure" - they are just behind. Behind in areas which are crucial for CPU manufacturer and vastly expensive to pursue; to the point that hardly anybody does so.

Money which Intel did illegally channel from AMD (you haven't heard about quite recent EU rulling?) would help greatly with pursuing them.

Re:Sigh

[psycho12345]

Yes that changed. AMD had integrated memory controller (HyperTransport), which gave them a huge memory bandwidth advantage over Intel. That ended with the i7, Intel moved memory controller on die as well. Intel has been beating up AMD offerings in both cores and raw speed ever since. Hence the first hexcore being an extreme edition, along the $1k price tag. Right now AMD is catching up on cores, but still behind on manufacturing tech, they still on 45nm, where Intel is already on 32nm. The new socket/board is no surprise, its all part of Intel's Tick/Tock strategy.

Re:Linux

[faragon]

At least they're trying?

Trying it is not enough. It's 2010, and AMD bought ATI almost 4 years ago (1), so there are no excuses. I would be glad of buying AMD+ATI integrated graphics instead of Intel, but it is a no-no until drivers for Linux reach its Windows counterparts performance-wise, and of course, I will not buy anything from AMD+ATI until then, not before. I buy products based on facts, not promises (I already made a mistake 3 years ago buying a AMD/ATI integrated graphics, still today without proper driver for Linux WTF!!!).

Re:Sigh

[Wolfgang8741]

The only major "failure" of AMD is that Intel lately managed to be noticeably ahead in both R&D and fab technology; there is nothing particuarly wrong with AMD tech, nothing that says "failure" - they are just behind.

I would not blame all of it on Intel, the manufacturing has not been yielding too well on the new techs. Which is contributing to the lower revenue for AMD. Someone needs to talk with the supply chain people about multiple sourcing of manufacturing.

I wonder

[sheph]

If that one pin is for the evil bit. It would be about time.

1 pin difference

[slick7]

How does a one pin difference in socket design correlate with motherboard manufacturing and compatibility with other systems.
Why is a 1 pin difference incrementally superior to present form factors.
It seems to me that not everything is aboveboard about this issue.
First you give the natives blankets.
Then you give them blankets with smallpox.
This is basically a 1 pin difference(?).

I guess 20/20 hindsight will show the truth of the matter.

...except when the same thing happens with AMD.

[InvisiBill]

I happened to buy an AM2 board around the time AM2+ was released (late 2007). In February 2009, I upgraded to an i7 system because my AM2 board wasn't updated to support AM2+. If my board had supported AM2+/AM3, I could've simply dropped in a new CPU for a significant upgrade. However, it didn't, so a faster AMD processor also meant a new motherboard and most likely new RAM (the exact same components needed for an Intel upgrade). The simple fact of the matter is that depending on what you buy and when, you'll sometimes get lucky and be able to upgrade for quite a while and sometimes you won't. Previous to my A64 X2, I had a SocketA system. At the time I bought it, it was about the best you could buy. I was able to drop in a faster CPU, but the board didn't support the additional multiplier bit of the unlocked CPU, so I was never actually able to run it at the full rated speed.

LGA1366 has been out for over a year and a half now and most X58 boards support the new hex-core i7 with just a BIOS update. At this point, I'd say that my new i7 system is more upgradeable than my last AMD system, and probably just as upgradeable as the previous AMD system. That's due partially to the coincidental timing of my purchases, but it doesn't change the fact that my last two AMD systems simply weren't as upgradeable as some people would have you believe all AMD products are.

Re:...except when the same thing happens with AMD.

[BikeHelmet]

What were the manufacturers and motherboard models, so I can avoid them?

My Asus AM2+ board got updated with Phenom II support.

Mac Upgrade just a dropin replacement...

[AmigaHeretic]

>>No more CPU upgrades problems! /duck

Upgrading the CPU in Mac: (Also, applies to HD, Video card, Monitor, etc...)

Step 1. Throw old Mac in the trash.
Step 2. Buy a new Mac.

(wish I was only half-kidding ;-) )

Re:Mac Upgrade just a dropin replacement...

[Omestes]

Not quite true. I just upgraded a crappy MacMini, changing out it Core Duo 1.8Ghz, for a Core 2 Duo 2.6Ghz chip. It was a pain in the butt, and required creative use of paint scrapers, and the avoidance of cheap peices of flying plastic (holding the heat sink to the socket), but it was doable in the matter of 15 minutes or so. The biggest pain was actually finding a chip for the socket, laptop chips are expensive and relatively hard to find. It also made me realize that Phoenix no longer really has decent Mom&Pop PC stores. You used to be able to upgrade older iMacs too (before the Intel switch). As for MacBooks, I'm guessing they are made to be as disposable as my HP laptop, or any other laptop in the world.

GMA 950 vs. newer GMAs

[tepples]

3D performance is fine for everything but the newest games at higher detail levels.

Really? I seem to remember that the GMA 950 is more like a Voodoo3, lacking even hardware vertex processing. Or has GMA 950 been pretty much discontinued in favor of more powerful GMAs?

Re:GMA 950 vs. newer GMAs

[Baloo Uriza]

You remember wrong (or Second Life would die horribly enabling VBO on Linux).

Re:GMA 950 vs. newer GMAs

[izomiac]

I've never used the GMA 950, so you could be right. My current laptop has a GMA 4500 MHD, which is considerably better than my old laptop's midrange nVidia. There's also the GMA 4700 MHD, which should be better still. The 4500 series was a pretty big jump from the 3100, which was a big jump from its predecessor, which (from what I hear) was fairly bad.

Re:GMA 950 vs. newer GMAs

[Vectormatic]

the 950 chip has pretty much been discontinued ages ago. It resides on the 945gm chipset, which these days is only used in some netbooks (and even there is is getting replaced rapidly by the nm10+n450 combo).

The mATX intel board i bought two years ago already has a G35 chipset with a much better graphics chip, granted it was a higher end matx board, but still, even if you only want intel chips, the gma950 has been replaced ages ago.

But yeah, a gma 950 is pretty awefull for anything but old games, it wont even run unreal tournament 2003 acceptably, we had to go for UT99 for the office lanparty..

Re:GMA 950 vs. newer GMAs

[nxtw]

Really? I seem to remember that the GMA 950 is more like a Voodoo3, lacking even hardware vertex processing. Or has GMA 950 been pretty much discontinued in favor of more powerful GMAs?

The GMA 950 was obsolete in 2006 (when the G965 and GM965 were released) for most applications - the exception being low-powered systems such as Atom. The Atom N450 integrated a newer GPU core (GMA 3150) into the CPU, and was released in December.

Re:Motherboards are not your "upgrade platform"

[breeze95]

Sockets are created to be as cheap as possible while supporting the cpu in question. If it wasn't for lack of configuration options and replacement of faulty parts, we would be buying our cpus soldered to motherboards.

Actually at one time CPUs came soldered onto the motherboards and if you needed to upgrade your CPU you had to know how to desolder the CPU from the board and solder the new CPU back on. Ahh, the good old days.

Yeah, only it worked out better...

[CFD339]

...kind of like what AMD did, only Intel has been better able to capitalize on it apparently.

I was the AMD stuff was more competitive, but in terms of performance for the work I'm doing, I have identically configured linux boxes and what should be the lower spec on the intel machine produces better results for me than the "on paper" higher spec AMD based one.

Re:Sigh

[petermgreen]

One less pin sounds like they wanted a different name so tweaked the number of power/ground pins marginally and/or removed some debug/test pin.

There isn't much of a chipset with the LGA1156 stuff anyway. All the chipset does is deal with the slower stuff (SATA, low lane count PCIe, traditional PCI) and provide the physical layer stuff for the video. The video core (where present), fast PCIe and memory controller are already on the processor package anyway.

P.S. According to TFA, it's actually TWO new sockets, LGA1366 is apparently going to get replaced as well (YAY quad channel memory, lets hope there is a dual socket version of it to get 8 channels total).