Intel Releases Broadwell Desktop CPUs: Core i7-5775C and i5-5675C

edxwelch writes: Intel has finally released their Broadwell desktop processors. Featuring Iris Pro Graphics 6200, they take the integrated graphics crown from AMD (albeit costing three times as much). However, they are not as fast as current Haswell flagship processors and they will be soon superseded by Skylake, to be released later this year. Tom's Hardware and Anandtech have the first reviews of the Core i7-5775C and i5-5675C.

Comments

About *** time

[Carewolf]

I was afraid we would have skylake ultrabook chips before broadwell desktop. This was a close call.

Re:About *** time

[Anonymous Coward]

...not sure that we would've missed anything IF that had happened...

also WTF did you link the co.uk and not the .com?!

Behind the news much, Slashdot?

[pla]

Did I miss something here? I've run a Broadwell i5 in my NUC for about three months now.

Re:Behind the news much, Slashdot?

[nine-times]

I'm not sure if you're joking, but even though your NUC is technically a desktop machine, it runs a mobile chipset. That's how they get it into such a compact package, by using chips and parts designed for laptops.

Re:Behind the news much, Slashdot?

[pla]

Ah, "desktop processors" counts as the key phrase there.

No, not joking... Just an idiot who can't read. :)

Re:Behind the news much, Slashdot?

[LWATCDR]

Your NUC uses a bga mobile chip in a "desktop". This is a cpu that you can throw onto a motherboard and add the video card of your choice as opposed to you NUC.

What about AMD Godaveri?

[Freedom Bug]

Tom's didn't test against AMD Godaveri, which has a substantially faster GPU than the Kaveri chips Tom's tested against. Godaveri is about 20% faster than than Kaveri, so would be competitive with these chips, as well as being about 1/3rd of the price.

Re:What about AMD Godaveri?

[Anonymous Coward]

Are you on crack? Godaveri is used internally by AMD. They released a minor speed bumped APU that still manages to perform worse (at higher power mind you) in many case than the A10-7850k. The true difference being that it now has a price of about what it's worth, c. $120. IOW they released a binned a10-7850k as a10-7870k kaveri APU, not even remotely news.

Re:What about AMD Godaveri?

[Anonymous Coward]

Does not matter, they aren't enough for gaming and too much for headless servers.

Re:What about AMD Godaveri?

[edxwelch]

Tom's didn't compare it, but Anandtech did and Godaveri is actually slower than Kavari for more than half the games they tested. Check the benchmarks if you don't believe me. In Alien Isolation, Total War Attila and GRID autosport the 7870K slower than 7850K. That's 3 out 5 games where it's slower!

Re:What about AMD Godaveri?

[bongey]

The Intel CPU is on par with the Godaveri over here http://www.anandtech.com/show/... .
Once you add a R7-240 the AMD chip is faster with dual graphics and sometimes faster by itself. Still cheaper buying AMD and the 240 card than one intel cpu.
Somewhat bogus benchmark because they didn't enable dual graphics for the AMD chips in the intel test. There is reason that you would want a AMD chip to crossfire later if you don't have a lot of money.

But since nothing is CPU bound

[Anonymous Coward]

It really doesn't matter. Desktop PCs need better IO of all kinds - disk and network primarily. Eleventy GHz machines don't do anything special coupled to a 3 Mbps DSL line, or running an OS on a 7200 rpm spindle.

Re:But since nothing is CPU bound

[Overzeetop]

Depends on your application. Most of us that would appreciate faster CPU speeds have already moved to SSDs and Gb for local network storage. Transcoding is processor intensive with local SSD, and lots of media center applications - running on desktop hardware - are now transcoding for remote viewing devices. I appreciate the desire to reduce part count and beat costs down, but stealing from the CPU performance for on-board GPU sounds like low-end chip work, not high end i7 stuff.

Re:But since nothing is CPU bound

[Anonymous Coward]

I agree, but I think we could even do better with the basic designs (I/O included) that we already have.

Desktop PC's need about a shit-ton more RAM. The 7200 RPM spindle doesn't matter so much when the entire OS is in RAM and not swapped back out to the damned disk.

I upped my minimum spec to 16GB about 4-5 years ago, with the introduction of DDR3. Now, I'm looking beyond that. Perhaps 64-128GB is in order now, but I'm hard-pressed to find a non-server motherboard that will support that much. Most boards still have only 4 slots, and will only accept 8GB per slot. That's pathetic. RAM is cheap, and we should be swimming in it, metaphorically speaking.

There's absolutely no reason why I shouldn't be able to set aside a chunk of RAM and make a RAM swap disk (to get around the brain-dead limitations of OSes that assume there must be a swap partition and cripple themselves without one). Sure, SSD's are fast, but RAM is faster. And cheaper, since it's meant to take constant writes and rewrites and not need replacement. There's no need to wear-level RAM. And what's in swap should die with a power-down anyway.

And I/O throughput? Hell, buffer it and have background processes that do nothing but spool buffers into I/O. That means more cheap RAM. Sure, increase I/O speeds as much as possible, but don't break the bank doing it. A pile of extra RAM is cheaper than wringing the last 5% out of I/O.

Basically, PC hardware design has stagnated. It works well enough for most things, but nothing outstandingly "better" gets produced anymore. It's just more configurations of the same old thing.

Re:But since nothing is CPU bound

[MrFlibbs]

You're mis-understanding the conclusion. Intel did not steal from CPU performance to improve the GPU, and in fact the cores on Broadwell are slightly more efficient than Haswell. Here's a quote from the Tom's Hardware article:

"As host processors, Core i5-5675C and Core i7-5775C should be marginally faster than Haswell-based CPUs at similar clock rates. The issue, of course, is that they employ lower frequencies than a number of previous-gen chips. So, they'll actually post lower scores in workloads that emphasize host processing (like the Sandra Arithmetic benchmark, above)."

Re:But since nothing is CPU bound

[petermgreen]

Afaict Intel and their customers knows there are people who want a PC but don't want a big box. They also know that some of those people will have money for a high end product and don't want to call attention to the fact that people are making a performance sacrifice by bying such a box. So they put their top end brand on a chip that is designed for such boxes just like they put their top end brand on laptop chips. For those with a bit less money to burn they market a marginally less powerful versoin of the chip under their midrange brand.

They know that the people who really need lots of CPU and/or GPU power will see through this and realise that these chips are weaker than the combination of a regular desktop CPU and a seperate graphics card but those people aren't who the marketing is aimed at.

Re:But since nothing is CPU bound

[Anonymous Coward]

Hm, but my computing needs are mostly CPU bound.

Re:But since nothing is CPU bound

[tompaulco]

That is a lot of RAM. My system is generally running about 25 applications on the task bar, plus whatever is running in the background for the OS. The system is using just under 6 GB of memory. I have 8 GB on board. I have occasionally contemplated going to 16 GB, but more than half of it would be sitting idle most (all, really) of the time. The only time I have ever used all 8 GB it was due to svchost hogging all the memory.
There are probably some switches you can set to turn your leftover RAM into swap. I know there is on Windows Server.

Re:But since nothing is CPU bound

[Anonymous Coward]

I tend to have anywhere from 3-10 instances of various versions of devenv.exe (Visual Studio) running at any given moment, so 16GB is pretty easy to fill for me. Throw in Outlook, SSMS, Firefox, and Visio, and 16GB becomes the minimum.

That's why.

Re:But since nothing is CPU bound

[TsuruchiBrian]

Why would you want to turn your ram into swap? Isn't swapping what happens when you run out of available ram? By turning your available ram into swap, you are causing the need to swap.

Re: But since nothing is CPU bound

[Anonymous Coward]

Swap is always used. No matter what for various tasks. You can control where the swap file exists. And if you put it on a ramdisk that swap file is operating at the speed of the rAm.

Re:But since nothing is CPU bound

[MechaStreisand]

Clocking them down is not stealing from CPU performance? Your own quote contradicts what you're saying.

Re:But since nothing is CPU bound

[Ducho_CWB]

i7 supports up to 32GB, so if you want to use Xeon it's better to use with a high-end motherboard.

Re:But since nothing is CPU bound

[MrFlibbs]

Clocking them down is not stealing from CPU performance? Your own quote contradicts what you're saying.

Sigh. If you'd read the article, you'd understand why your statement makes no sense. Tom's Hardware goes on to note that Broadwell is ~5% faster than Haswell at the same clock speed. The reason Broadwell shows slightly lower performance on some benchmarks is that it's capable of dropping down to lower clock speeds to conserve power. But when performance is called for, Broadwell quickly ramps up to the same clock speed as it's predecessor. So for a sustained workload, Broadwell will be faster. It's only for those loads that frequently ramp up and down that Haswell comes out ahead.

So nothing is "stolen" from the CPU. Most of the extra gates proved by the 14nm process in which Broadwell is fabricated are used to enhance the GPU, which as noted by the reviews is now the fastest integrated graphics unit on the market. But nothing was taken from the CPU -- in fact, the CPU is enhanced to be 5% faster on a per clock basis and to also drop to lower frequencies when *idle*. If saving power isn't your thing you can always disable power features through either the bios or the O/S to keep the CPU at higher frequencies most of the time.

Dropping the CPU frequency to lower values when cores are inactive is an important feature in all modern CPUs. Many server customers care more about performance per watt than they do about raw performance. For laptops and handhelds, efficiency is critical. And even for desktops it's a nice feature.

Re:But since nothing is CPU bound

[KingMotley]

Say what? This machine is an i7-3930K and has 64GB of RAM...

Re:But since nothing is CPU bound

[toddestan]

Depends on the model. Your i7 supports up to 64GB. The first ones IIRC were 24GB.

http://ark.intel.com/products/63697

Of course, that all depends on finding a motherboard that supports an i7 (not a Xeon which is a different socket) actually accepts more than 32GB of ram.

Re:But since nothing is CPU bound

[Agripa]

The demand for using unbuffered memory places limits on the number of DIMMs which each memory channel can support and each memory channel adds to the cost of the processor and motherboard. This limits DDR3 and DDR4 to two DIMMs per channel unless the memory clock frequency is lowered.

Server processors and motherboards often support buffered memory so more DIMMs per channel but there is not enough demand to support this on desktops in the face of market segmentation.

Broadwell is yesterday's news. Bring on Skylake!

[SchroedingersCat]

The world is waiting for HDMI 2.0.

Re:King Frosty, First poster, who is royal

[Anonymous Coward]

Sorry, but cyborg_monkey is the one true King.

albeit costing three times as much

[nitehawk214]

For the past 5 or 10 years this has been the story of me building new computers. I don't follow tech pages on architectures much any more, just when I go to build a new computer I go and see what the latest offerings from amd/intel/nvidia are.

For pretty much ever it is, "AMD is kill, Intel rules all!" Except the fine print is that in order to rule all, you must pay 2x to 3x as much. So all of my performance/gaming computers for 17 years have been AMD/Nvidia (and VIA chipsets before Nvidia). (I have tried ATI a few times and just never cared for them.) And I get 3+ years out of each computer before it needs to be replaced.

Now, from a heat dissipation and power usage perspective, no amount of price/performance can replace that. And this is why I have not seen an AMD laptop in quite some time.

So why is AMD constantly on the verge of bankruptcy? Is there some Apple effect on Intel that causes people to throw money at them for no better performance increase? Do people simply not care how much they spend on computers? Is the laptop/mobile market cutting into PC/Server that much? Or are they just poorly managed. Over 15 years and I simply don't get it.

Re:albeit costing three times as much

[beelsebob]

The reason is simple - the title in the headline is misleading you about needing to pay 2 to 3 times more. It's comparing the Intel chip to a relatively low end AMD chip that happens to have a GPU, not to the high end AMD chips that it actually competes against.

If you go look at the first review, you'll see that in the CPU speed tests, the i5-5675C turns out to be substantially (about 30%) faster than even the FX-9590 (AMD's fastest desktop chip). That and it has a decently fast GPU built in too.

The i5 costs $276 (list price, so likely higher than what you'll actually get it in the shops), the FX-9590 costs $249 (on newegg today). So that's a 10% markup for a 30% faster CPU with a very usable GPU on board. Most people see that as a pretty good deal.

Re:albeit costing three times as much

[DigiShaman]

So why is AMD constantly on the verge of bankruptcy?

Typically when you're at that level, management is usually the problem. Technically, AMD can take market share. Marketing-wise, they're not. R&D isn't the problem.

Re:albeit costing three times as much

[bongey]

For some reason anandtech didn't enable crossfire in any of the benchmarks. Who in their right mind would have a AMD APU and AMD G-Card and disable crossfire, unless you are trying to make intel to appear much faster.

Re:albeit costing three times as much

[nitehawk214]

True, this is all about integrated stuff that I do not care at all about.

But, I don't know, its been 2+ years since the last time I have looked at hardware (had to look at my newegg order history to figure out how long ago), so maybe things have changed; but I really tried to buy Intel last time out. But when you add the motherboards that cost $100 more for high end boards, I just couldn't find a price point where Intel was able to match.

The result was I paid $200 for an FX-8350, which probably wasn't AMD's fastest chip at the time, and $120 for an ASRock moberboard with onboard raid. I remember all of the benchmarks compared it to the i7, which of course trounced it. But compared to similarly priced i5 at the time the AMD was just better. (also considering motherboard prices)

Of course, I was already planning a large case with a large heatsink/fan combo, so thermal concerns were not part of my calculation. If I wanted a reasonably sized computer, I would almost have to buy Intel.

It is probably just the specific configuration I do (bargain medium-high end graphics and gaming) is not as popular as it once was.

That being said, I looked at Newegg's featured AMD processors and the most popular one was... still the same 4ghz FX-8350 with the same serial number for nearly the same price I ppid... after 2 years? What? The FX-9590 doesnt seem to be a significant step up in performance from the 8350. It is only slightly more expensive.

On boards... I guess I don't know how to price Intel boards. The most expensive ASRock AMD motherboard is $190 (though it seems to have the same features as the $140 ones, I did not look closely. The Intel ones all all over the map from $120 to $650, but it seems like the i5 boards are less than $160 so it looks like these are more competitive now.

And the FX-9590 is 220 Watts?? At this point I should be looking at price/W instead of price/$.

Re:albeit costing three times as much

[Anonymous Coward]

At the top end that may be true, but in the following very common scenario Intel beats AMD not only in price/performance, but overall performance:

GPU+Discrete graphics card

Top-end AMD cpus get beaten by low-to-middle grade i3s /that cost the same/

AMD's integrated graphics beats intel's low-end offerings, sure. "Iris pro" integrated graphics to beat AMD's integrated graphics at 3x the cost, but the CPU coupled with them is much much much faster than anything AMD offers. Iris pro products are aimed at premium markets where max performance, low power consumption, and small physical space is at a demand.

AMD's offerings also consume a LOT of power. That power becomes heat, which needs to be dissipated.

AMD, therefore, only competes in a niche space where:

Power is not a problem
Heat can be dissipated easily
Integrated graphics are desirable (usually for cost reasons)

That's a surprisingly narrow niche because once you scale up to the AMD SKUs that have vastly superior integrated graphics.. You'll likely want the benefit from a graphics card. Even entry level gaming cards run circles around integrated graphics, and don't cost that much. 9 time out of 10 you'll be better served by going intel+graphics card.

Re:albeit costing three times as much

[Moof123]

Also fab access. Intel has kept their fab enough ahead of their other faults to be ahead most of the time. AMD has arguably had a better architecture than intel, but is stuck a couple process nodes behind due to not having access to a comparable fab.

I am simply dumbstruck with how flat the performance has been for intel. The power reductions are impressive, but the speed has been nearly flat for the last 4-5 years.

This latest round reeks of being an Apple specific processor. Anyone wanting a good machine would pair a halfway decent GPU into the system, making the 6200 Iris graphics meaningless. Apple sticks out as a likely buyer wanting "good enough" graphics with low power dissipation for a desktop build. Anyone else spending $230-330 on a CPU won't mind spending another $100 for a bottom end GPU that would still blow away the 6200 performance.

Re:albeit costing three times as much

[mister_playboy]

Yes, the AMD FX line has not had any updates since you last went shopping and it has gotten less and less competitive versus Intel's offerings, especially on single-threaded tasks and in work done per Watt. It was widely believed they were actually going to abandon that market segment entirely, but the new Zen architecture is now planned to to first appear as a revamped FX line.

Re:albeit costing three times as much

[Kjella]

I've never understood what market wants a powerful CPU paired with a meddling and power crippled yet still expensive GPU though, except in a laptop where it's all you got. Pretty much every benchmark shows that if you want gaming performance, put almost all your money in the graphics card. I mean the high end processor is $366, you can get a $64 Intel G3260 and pair it with a $299 Radeon 290X for less that'll be a much, much better gaming machine though it'll use 200W more when you're playing.

Now if you really want that powerful CPU for non-gaming purposes that's fine, but then you can buy an i7-4790K and save the rest towards buying a real graphics card. I mean seriously, you're spending $300+ and the benchmarks are if you can play at 720p low quality between your number crunching? It does not compute. And it's a total waste if you decide that 720p is not enough, the integrated graphics will then be dead weight, which seems more likely to happen if 60-80% of your budget went to buying the CPU as opposed to buying an APU where you spent 60-80% on the GPU in the first place.

Re:albeit costing three times as much

[FlyHelicopters]

The result was I paid $200 for an FX-8350, which probably wasn't AMD's fastest chip at the time

That same $200 would have bought you a Core i5, which is faster in most respects to the AMD chip while using less power.

Yes, there are edge cases where the AMD chip is faster. Are you one of those edge cases?

$120 for an ASRock moberboard with onboard raid.

You can get nice Intel boards for about the same money, the $190 boards are overkill.

Of course, I was already planning a large case with a large heatsink/fan combo, so thermal concerns were not part of my calculation. If I wanted a reasonably sized computer, I would almost have to buy Intel.

Thermal may not matter, but how about your power bill?

The Intel chip will use less power, over 3 years of owning it, the power bill difference can easily wipe out any up front price difference.

And the FX-9590 is 220 Watts?? At this point I should be looking at price/W instead of price/$.

Insane, isn't it? These new Intel chips max out at 65w, and use less when the GPU isn't in heavy use.

However much time your computer is actually in use, times 150w of power, times three years, is how much in your power bill?

---

I'll be frank, a few years ago I didn't much consider the power consumption either, until I replaced my HVAC system with something from this century and then replaced all my incandescent bulbs with LED bulbs. I've started to do the math on how much of my monthly power bill is due to electronics, and the percentage is growing.

So I do now consider the typical lifetime power cost of something before I buy it, something I never used to do.

Re:albeit costing three times as much

[FlyHelicopters]

AMD looks better in benchmarks than they actually perform for a lot of applications.

Let me be clear. If you're doing something like image processing, compression, video work, an 8 core AMD chip is likely to be faster than a 4 core Intel chip. Except... the 4 core, 4 HT i7 chip is likely to be faster than anything AMD makes and if you're REALLY doing that kind of work, another $100 or so in computer cost is nothing compared to time saved.

If you're doing basic Internet surfing, e-mail, Angry Birds, etc. Then frankly ANYTHING modern is likely fine, get an Intel Pentium Anniversary chip for $70 and rock on while sipping power.

Where AMD falls apart is, frankly, everything in the middle. A $100 AMD chip appears to be faster than a $100 Intel chip, so long as you're either running an intensive benchmark that doesn't sit and wait for human input, or running massively threaded applications that need 4 real cores. Except, see above, people doing that shouldn't be buying $100 CPUs anyway.

Some games appear to run well on AMD chips, others do not due to their poor single core performance. It generally takes a 4 core AMD chip to equal a 2 core Intel chip. Which is fine if you're running something that will use 4 cores. If you're not, the Intel chip runs away from the AMD chip, all while using just over half the power.

---

I used to be a big AMD fan, Athlon XP was wonderful, Thunderbird was wonderful, but those days are long in the past now. Since Core2Duo came out, AMD has simply been behind.

I honestly wish AMD could do better, I think the past few years have been slow due to a lack of real competition.

Re:albeit costing three times as much

[DigiShaman]

Frequency speeds have been pretty stagnate due to the laws of physics. Asynchronous clocking is one way of dealing with signal propagation and timings as I understand. But the focus these years has been on adding features as the die shrinks further. Not many programs take advantage of multi-threaded operations, hence why the home and office market doesn't see much in the way of increase core count. For raw multitasking, it's my understanding that the Xeon line sacrifices multi-media features for extra cores and cache. Die space is limited as such gets allocated to whatever that market it's optimized for; in that case server and workstation.

Re:albeit costing three times as much

[Kartu]

FX-9590 is the fastest AMD CPU you can buy and it is on par with i7 4770k. (as usual, behind in single threaded test, ahead in multi-threaded tests)
http://cpuboss.com/cpu/AMD-FX-...
It IS faster than i5, if you are after multi-threaded load.
On top of it, it has margin corresponding to "my fastest processor".
It has no GPU.

Your choice of CPUs to compare reviewed i5 to is questionable, to say the least.
A10 APUs that were reviewed by Anandtech, cost half/third of Intel's, yet are within 20% performance wise.

Re:albeit costing three times as much

[TsuruchiBrian]

Onboard raid is for suckers. Just do a software raid. Processors are really fast now. You don't need dedicated raid hardware anymore, especially the cheap chips they put on motherboards just to say they do raid. Also, with a software raid, you are not relying on a particular chips implementation of raid. You can take the drives out, stick them on a different controller in a different computer and still have access to your data as long as you are running the same raid software.

Re:albeit costing three times as much

[dinfinity]

5.5 years ago, I bought an Intel i7 860 and accompanying mid-range motherboard for 350 EUR. That means I've paid ~65 EUR/year, ~5 EUR/month for that combination, which is _still_ serving me ridiculously well (so much so that I really really really need to convince myself that I want to upgrade it -- it's far from necessary, but it 'feels' like it is time).

Taking into account that I work from home, for me it is pretty simple: I just can't be bothered to skimp by going AMD and shave off maybe 3 EUR/month on what is hands down the most important device in my life.

Re:albeit costing three times as much

[nitehawk214]

Remember that the 125W is only the max usage. This is a computer that gets turned off when I am not using it, so the power usage is minuscule compared to a refrigerator. (or the second beer fridge in the basement...)

However, if the new chips are going to double power usage, for very little gain in performance, well perhaps it is time for a change. Eventually... I don't see myself needing a new computer for a couple years. I have no illusions that AMD will actually start to care about power usage anytime soon, however.

Also in my previous post I meant to say "performance/Watt instead of performance/$".

Re:albeit costing three times as much

[nitehawk214]

Well I have an AMD computer, so I am obviously a sucker.

But this is interesting. I am using RAID1 mirroring only, as giant drives are so cheap and plentiful. So RAID performance really isn't an issue to me at all. Maybe there is no need for hardware RAID. The super high performance stuff I do all goes on an SSD anyhow.

I know the setup I have works as I did recently replace a drive. I have said many times before that I am done buying spinning disks, the next machine will be all SSD.

Re:albeit costing three times as much

[FlyHelicopters]

Regarding "max power pull", you're right of course, when idle they all use less.

I will say, take a look at the "idle power" of AMD's chips and the "idle power" of the new Intel chips.

One of the reasons I upgraded from Sandy Bridge to Haswell was not speed (part of it, but not all of it), but power consumption.

Re:albeit costing three times as much

[TsuruchiBrian]

I am not even sure that hardware raid is any faster than software raid at this point. The bottleneck is surely the spinning disks. If anything a CPU can probably do a better job at RAID and have enough spare processing power that you wouldn't notice a performance hit.

I don't think it really matters that much for a simple mirror. I think the only real benefit of a software raid in this case would be better reporting of statistics related to performance and integrity. You can use any software you want to do this, while with a hardware raid you are relying on the driver from the manufacturer to do this.

An advantage of software on larger RAID arrays is that often only some of the ports on a motherboard support raid, with software RAIDs, all the ports can be used in a RAID.

Re:albeit costing three times as much

[rdnetto]

The result was I paid $200 for an FX-8350, which probably wasn't AMD's fastest chip at the time

Maybe not, but close - the FX-8370 is just a slightly better binning of the same part.

I remember all of the benchmarks compared it to the i7, which of course trounced it.

Funny thing about that - there were some pretty major discrepancies at the time between benchmarks done using Intel's compiler and those done using GCC. When using GCC, the FX smoked the i7 - it wasn't until the next generation (or possibly the one after that) that the FX started to lag behind. Even today it's reasonably competitive (if not faster than) against Haswell i5s.

The FX-9590 doesnt seem to be a significant step up in performance from the 8350.

The FX-9590 isn't even a step-up - it's the exact same part, overclocked ridiculously (hence the 220W TDP). It's a stop-gap measure by AMD to try and disguise the fact that the FX line-up hasn't been updated in years.

I'm planning an upgrade myself atm, and as much as I prefer AMD, they're literally a non-option to me (even with the sticker shock from i7 prices). I hope Zen goes well for them, because otherwise Intel has a monopoly on high-end CPUs...

Re:albeit costing three times as much

[Shirley Marquez]

Why is AMD on the verge of bankruptcy? The lower performance of their chips means they have to sell them at lower price points to get any business at all. The fabrication technology they have access to is a couple of generations behind Intel's, so those lower performance chips are probably actually costing them more to make than Intel's faster chips but they can't get nearly as much money for them.

For many years, AMD owned its chip fabs. They didn't have the money to make the necessary investments to keep them at the cutting edge so they were always playing catch up. Then they spun off the fab business as GlobalFoundries, with the hope being that the new company could attract enough outside business (as well as AMD's own) to put it in a financial situation to catch up with the big players. That has not happened; GlobalFoundries has not caught up, and AMD can't get enough fab capacity at better equipped TSMC to make its CPUs there.

To really get back in the game, AMD needs a new CPU architecture that is more competitive with Intel so they can start making some higher margin chips again. They are known to be working on one that will be ready in 2016 or 2017; we will see whether it is good enough to compete. In addition, they need access to better fabrication technology to manufacture that new design. If they can make both of those things happen, AMD might finally claw its way back to health.

Re:albeit costing three times as much

[Shirley Marquez]

Here is a link with info about AMD's planned next generation: http://www.anandtech.com/show/...

When do we get a real boost over 2013 speeds?

[rbrander]

I've got a machine over two years old now - I do some pretty heavy number-crunching with GIS map programs and always tell the counter guy I want the nearest thing he's got to a machine that finishes infinite loops. After conceding that the next model up from the i7-3930K was $500 more for another 15% of horsepower, I picked that one.
I'm sure there have been a few percent of gains with two years of subsequent chips, but basically, it's same cores, same GHz. Is this 'skylake' in several more months going to be more than a 10%-15% upgrade over my early 2013 chip? (Actually, it's older, probably came out in 2012?)
I really need to be buying a second machine in just a few months, but I'll endure some inconvenience if we're just a few months after that from a significant upgrade. But frankly, anything under 25-30% speedup in math operations will not be worth the wait.
They say Moore's Law is still going, and in low-power circles, I'd agree. But for the market segment of people who don't mind the computer doubling as a room heater if it'll just crunch numbers on a few million rows of geodatabase table a few minutes faster, it sure feels like Moore's is over for us.

Re:When do we get a real boost over 2013 speeds?

[Anonymous Coward]

Only so much juice you can squeeze from an orange, dude. I'm still using a setup from 2008 since nothing yet guarantees the 100% improvement that would make me upgrade.

Re:When do we get a real boost over 2013 speeds?

[mrchaotica]

Aren't GIS problems embarrassingly-parallel enough that you should be worried about finding a faster GPU (or maybe switching to a multi-socket Xeon system) rather than about having the fastest single-threaded performance?

Re:When do we get a real boost over 2013 speeds?

[David_Hart]

Only so much juice you can squeeze from an orange, dude. I'm still using a setup from 2008 since nothing yet guarantees the 100% improvement that would make me upgrade.

Personally, any upgrade would be for a motherboard with USB 3.1, PCIe 4.0, and DDR4. Basically, faster I/O. I wouldn't be upgrading for more processing power as the I7-3770 works perfectly fine for just about everything that I throw at.

Re:When do we get a real boost over 2013 speeds?

[Bob the Super Hamste]

My experience would indicate this as I have an i7 3770k (I think that is what is is) and I can load all 8 virtual cores at 100% for extended periods. The biggest bottleneck I had previously was memory but having 32GB has solved that for now. I could have gotten a multi-socket Xeon but for amateur work that gets into the silly price range. As far as GPU acceleration I don't know if Esri supports it as I stick with open source tools as I can't afford the phenomenal cost and the open source tools have always gotten me what I needed, even if they don't support GPU acceleration. This may have changed as I am probably a couple of version back now.

Re:When do we get a real boost over 2013 speeds?

[wbo]

Since most GIS systems are heavily multithreaded, wouldn't you be better off with more CPU cores rather than a few fast cores? The Xeon series of processors are designed for this type of thing much more so than the Core i7 series.

The i7-3930K is a 6 core CPU. If you want to stick with a single-CPU system, you could go with an Xeon E5-1691 v3 which gives you 14 cores.

If you are willing to go with a dual-CPU system, you could go with something like the Xeon E5-2698 which 16 cores per processor (for a total of 32 cores in a dual-CPU workstation.)

Of course the Xeon CPUs can get a bit expensive and many require different motherboard chipsets (although low-end motherboards compatible with Xeon CPUs cost about the same as a good motherboard for an i5 or i7 now.) However, the CPUs I referenced are pretty close to top of the line. You can save quite a bit and still far more cores than you can on an i7 by going down a model or two.

Re:When do we get a real boost over 2013 speeds?

[Moof123]

Same here, SATA3 would be nice over my SATA1 for my SSD, but my i7-920 is not very far behind the current crop of CPU's considering it is 6 years old. I at least had hope for 6-8 cores by now, but 4 is still considered plenty so that is what can be gotten without getting too gouged.

The die pictures are pretty disappointing too, the core's are pretty small, surely adding a couple more would not increase the die size much on a percentage basis, but intel seems to want to more than double the price for their 6 core chips compared to a 4 core (and drop the clock speed to erase a good part of the expected gains).

I am planning on pulling the trigger when Skylake comes out, but Moore's law looks pretty much saturated at this point. Even DDR4 looks to be a very minor improvement over DDR3 from the benchmarks thus far.

Re:When do we get a real boost over 2013 speeds?

[MetricT]

I do high-performance computing for a living, and Moore's Law has been on its last gasps for a while now.

Until around 2006, the smaller you made a transistor, the faster it could work. This was called Dennard scaling. But once transistors reach a certain size, current leakage and thermal issues prevent you from making the transistors faster.

While they can't drive transistors any faster, smaller processes still allow them to put *more* transistors on a chip. This is why we've gone from single-core to multi-core to multi-core with GPU compute on a die.

Despite all the complaints about "CPU's haven't gotten much faster since Nehalem", they *have* gotten quite a bit faster. You just have to rewrite/optimize/recompile your program to take advantage of multi-core, GPU compute, and SIMD instructions like AVX2.

This is the primary reason programs aren't running much faster than before. Silicon isn't getting any faster, and rewriting programs to scale isn't easy and sometimes isn't worth it so many people don't. Moore's Law no longer results in "free", "easy" speed-ups.

CPU's for the next few years are looking pretty incremental. I'd expect a one-off moderate increase in single-core performance once Intel moves off silicon onto III-V semiconductors (10 or 7 nm?), but past that you will likely be waiting several years for your graphene/nanotube/topological insulator/spintronics overloads to deliver something substantially faster.

Re:When do we get a real boost over 2013 speeds?

[mrchaotica]

I didn't think to mention it in my previous post, but a dual-socket AMD Opteron 63xx system might be a reasonable alternative. Opterons appear to be significantly cheaper than Xeons (and in some cases, cheaper than I7s), to the point that you could get a 16-24 core AMD system for close to the same price as a high-end i7 (let alone a Xeon). I have no idea which would win on the benchmarks, though.

By the way, what sort of open-source GIS tools do you use? I occasionally find myself wanting to do GIS-related stuff, but I don't know where to start.

Re:When do we get a real boost over 2013 speeds?

[FlyHelicopters]

Same here, SATA3 would be nice over my SATA1 for my SSD, but my i7-920 is not very far behind the current crop of CPU's considering it is 6 years old. I at least had hope for 6-8 cores by now, but 4 is still considered plenty so that is what can be gotten without getting too gouged.

If you're wondering where all the "improvements" went over the past 6 years...

Your i7-920 is a 130W CPU running at 2.66 GHz. This new Broadwell chip is a 65W CPU running at 3.3 GHz turboed to 3.7 GHz, and it is about 20% faster per clock cycle, making it about as fast as your current system if it were running at 4 GHz, while pulling half the power and having a nearly 10 times better iGPU.

You may not think much has changed, but that is actually a huge change, and it is where the improvements have gone.

If you want a 130W chip, get the Haswell-E, it isn't much more than the i7-4790k and it gives you 6 cores, 12 threads, and all those nice things you want on your motherboard.

Re:When do we get a real boost over 2013 speeds?

[FlyHelicopters]

2008 would be a Core2Duo or Core2Quad, probably running near 3 GHz.

It depends of course on what you're doing with it, and if you care about power consumption, but a Core i7-4790k will kick the pants off the Core2 line all day long for CPU intensive work.

The 100% speed boost is there, if you need it.

Re:When do we get a real boost over 2013 speeds?

[FlyHelicopters]

've got a machine over two years old now - I do some pretty heavy number-crunching with GIS map programs and always tell the counter guy I want the nearest thing he's got to a machine that finishes infinite loops. After conceding that the next model up from the i7-3930K was $500 more for another 15% of horsepower, I picked that one.

It is very rare that 2 years will provide a huge jump in performance.

The exceptions are when major new developments come out.

The Core2Duo was one such development, it was so much faster than Netburst, it was obvious and major.

I remember the old Athlon Thunderbird chips, those were good and a nice upgrade over a middle range Pentium II.

Other times, CPU speed tends to just sit around. The jump from a 486DX/2-66 to a Pentium 75 was very ho-hum back in the day, at least until Windows 95 showed up, then it became more clear. On Windows 3.1? Meh.

Go back further, In 1991 I had a 386DX-25, wonderful machine that was fast for its day, until 2 years later in 1993 I spent a decent chunk of money and got a 486DX2-66, that was a noticable upgrade.

---

Go back further, I remember when the 80286-12 chips came out, compared to the older XT that we had at the time, lord that was a rocketship. Anyone else remember the "turbo button"? :)

Re:When do we get a real boost over 2013 speeds?

[FlyHelicopters]

My experience would indicate this as I have an i7 3770k (I think that is what is is) and I can load all 8 virtual cores at 100% for extended periods.

If that is the case, then Haswell-E is probably what you need. Another 4 core chip won't help you nearly as much as having 8 real cores will.

Not the cheapest thing in the world, but if you're actually running for "extended periods" at 100% CPU usage, then maybe it is time.

Re:When do we get a real boost over 2013 speeds?

[Dadoo]

The jump from a 486DX/2-66 to a Pentium 75 was very ho-hum back in the day

Maybe for some workloads. For others, a Pentium was a significant upgrade. I used to play Quake, a lot, and upgrading from a 486DX4/133 to a Pentium 133 was like night and day.

Re:When do we get a real boost over 2013 speeds?

[FlyHelicopters]

Keep in mind that Intel has focused much of its efforts the past 6 years on power reduction, not speed.

The speed can come later, at the moment they don't need it due to lack of competition.

---

Consider, 2009, i7-920 running at 2.66 GHz was a 130W CPU.

Today, these new CPUs are running at 3.3 GHz with a turbo to 3.7 GHz, with a FAR superior iGPU, along with better IPC, while pulling half the power, 65W.

To get similar performance out of the i7-920 chip you'd need to run it at 4 GHz, perhaps a bit more, to counter the IPC improvements.

How much power would that chip consume, assuming it would overclock to 4 GHz? 200W? More?

So we've tripled the power per watt that can be provided in 6 years.

That is an amazing improvement. The only trick is that we need multi-threaded applications to see that translate into more performance.

Re:When do we get a real boost over 2013 speeds?

[FlyHelicopters]

I would expect that it would be, given the 486 being limited to a 33 MHz bus speed compared to the Pentium 133's 66 MHz bus.

Like I said, the jump from a 486DX2-66 to a Pentium 75 was rather ho hum, at the time those two chips were mainstream. The 486 was still selling strong in 1994 when the Pentium 75 came out. By the end of 1995 when the Pentium 133 was released, the 486 was no longer mainstream, being really slow for Windows 95 at that point. Keep in mind that the Pentium 75 had a 50 MHz bus, compared to the Pentium 133 having a 66 MHz bus. In those case, with almost no on-die cache (they were very small), bus speed meant a lot. A 486DX-50 often could outperform a 486DX2-66, depending on the task at hand.

Also, there was no 486DX4-133. Intel stopped at 100 MHz but called it the DX4 anyway. AMD did make a 133 MHz version, but called it "Am5x86-P75"

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

If you are comparing an Intel 486DX4-100 chip to a Pentium 133, well then yes, the Pentium is probably double the speed.

Even the AMD 133 chip was not really any faster than a Pentium 75, the 33 MHz bus speed was too much of a limiter to see the gains the high clock speed would indicate. Once the Pentium got to 100 MHz and a 66 MHz bus, the 486 line was more or less finished for mainstream computing.

I remember those days well, I owned a computer networking and support business back then, I spent many a weekend ripping out 486 computers that we had installed in 1993 and 1994 with Pentium 133 and 166 machines in 1995 and 1996 due to the 486's inability to run Windows 95 at acceptable speeds (at least for our business users). A P75 would do it, but the difference in performance between a P75 and a P133 were noticeable to the average person.

Re:When do we get a real boost over 2013 speeds?

[Bob the Super Hamste]

At the time when I looked at a machine for what I am doing the Intel option was the cheaper one and I could have saved $20 by going with a mid range i5 which I was planning on doing but the unlocked higher end i7 was only $20 more that what I was planning on getting so it was a why not purchase.

As far as tools go, if you want to start out and play around first without getting buried under your own ignorance (I speak from experience here) try something like uDig GIS first. It isn't the most powerful, fastest, or most feature complete tool but it is easy to use and gives you good point to develop enough knowledge to know what to ask and look for. There are a number of reasonable tutorials for it as well. If you already have some GIS background just jump in and go whole hog with QGis or Grass GIS. These are compleatly different monsters and between the 2 of them you get some real power. I have friends who do GIS stuff professionally for some of the counties in MN and they use these tools. If you are on windows there is the OSGeo4Win project that will give you a bunch of stuff that just works that includes these tools as well as a bunch of other ones. I also do make use of tools like GPSBable and the MN DNR's Garmin Application for converting formats as well as GPSBable for pulling data off of some of my GPS units. Add in a smattering of RTKlib for some of my more recent dabbling with high accuracy trail mapping and some custom built GPS devices. If you need some data to play with the MN DNR has their Data Deli where you can get all sorts of stuff.

Re:When do we get a real boost over 2013 speeds?

[Bob the Super Hamste]

By extended periods I meant 10-15 minutes, and that is with some very large data sets that will consume 24GB of the 32GB of physical RAM in the machine. I can wait that long without issue as this replaced an Athalon 64x2 with 4GB ram and I tried running some of those tasks and they would take upwards of 5 days with the CPU pegged and the disk spinning like mad paging data in and out.

Re:When do we get a real boost over 2013 speeds?

[FlyHelicopters]

Fair enough... I infered from your post that you wanted more performance, that you wanted to know how to get a 25%+ speed jump.

i7-5960X would do it...

Not cheap, but have you considered such an 8 core, 16 thread system with 64GB of RAM, then using 32GB as a RAM disk and doing your work there rather than off a SSD?

You might find your 5-10 min becomes sub-5 min...

Keep in mind that your jump from an Athlon 64x2 to an Ivy Bridge was about a 10 year leap in technology. Even the Core2Duo was faster nearly 10 years ago than the 64x2. It will be hard to do that again without either spending megabucks, or waiting another 10 years.

---

As a side note, if I'm not mistaken, you're doing this on an amateur basis, which is fine of course... If this was professional work, it screams for a multi-CPU Xeon system. It also seems like it would be idea for a GPU application as well, given the professional cards that nVidia puts out, that might be even faster.

Re:When do we get a real boost over 2013 speeds?

[DigiShaman]

I don't think there's a single Core i7 that will support more than 32GB RAM. I believe that's hard-coded to the die, and not motherboard. If you need to address more than that, you're looking at a Xeon chip from the Intel offerings.

Re:When do we get a real boost over 2013 speeds?

[Anonymous Coward]

FWIW, I find that with my own image processing tasks my i7-4700MQ laptop is almost twice the speed of my Xeon E5530 workstation even though both are 4-core/8-hyperthread nominal 2.4 GHz CPUs with turboboost and pretty fat multi-channel RAM setups. This is comparing parallel OpenCL routines (using the Intel CPU-based OpenCL drivers). Sequential tasks using Numpy also speed up quite a bit, but I don't spend time doing that since it is so much slower than going in parallel.

Re:When do we get a real boost over 2013 speeds?

[WhoBeDaPlaya]

Wait another half a decade. A Haswell 4770K maxed out at ~4.5GHz (typical max if you do not delid) works out to be ~= 5.6GHz Nehalem i7. This is absolute crap if you consider that it was pretty much 5 years from Nehalem to Haswell, but I needed to upgrade some machines (i7 920 @ 4.1GHz) so went with Haswell and got only a 37% boost in absolute OCed performance for my trouble.

Re:When do we get a real boost over 2013 speeds?

[Moof123]

So a 2x speedup in 6.5 years (unless you take into account the nearly 1.5x overclocking of my i7-920). Sure, lower dissipation is worth something, but sorry if my socks aren't blown off by a speed doubling every 6.5 years.

The Haswell-E drops the clock 10%, so you net a 1.35x speedup at best.

And I am painfully aware of the clock speed drop. We plunked down a serious chunk of change for a dual Xeon box (16 total cores) to run our electromagnetic simulations on, only to have it run about 10% slower than our 6 core desktop workstations. It turns out the wild claims of the software vendor about scalability with cores was crap (shocker!). It seems that at some point with the clock speeds getting stagnant that we would see more significant movement to have programs use larger core count better.

Re:When do we get a real boost over 2013 speeds?

[FlyHelicopters]

So a 2x speedup in 6.5 years (unless you take into account the nearly 1.5x overclocking of my i7-920). Sure, lower dissipation is worth something, but sorry if my socks aren't blown off by a speed doubling every 6.5 years.

I understand your point... I was simply saying where the development went... Rather than speed, less power consumption was the goal...

Do you not think that a doubling of speed while halving power consumption is impressive? I do in any case... what it also tells me is that for the same power, they could have quadrupled performance (give or take), if that had been their goal, but it clearly wasn't.

And you're right, software hasn't done a great job of adapting to many cores, which is a shame.

Re:When do we get a real boost over 2013 speeds?

[KingMotley]

There are plenty of i7's that support more than 32GB of RAM. I'm using a i7--3930K with 64GB right now, and there are others that support 128GB as well.

Re:When do we get a real boost over 2013 speeds?

[Bob the Super Hamste]

I tend to make major jumps in technology, my current machine is robust enough that if I want I can crank up the clock speed and ram speed a fair amount an not have to worry for some incremental gains. I buy a machine that meets my needs and has room to continue to meet my needs, use it until it becomes painfully slow or starts having hardware problems, and then replace to get the close to 10x performance improvement across the board. Hence the jump from the Athalon 64x2 to the i7. As far a GPU acceleration I don't know if Esri ArcGIS makes use of it because there is no way I could afford that but I probably should check if updated versions of the open source GIS tools can make use of it as a lot of the data is just single precision floats or ints.

The next machine I get will probably have between 256GB and 512GB ram given how things have typically progressed and I will probably pay $1200 for it as that seems to be the price point I have been stuck at for a while. The progression of my computers has been 486dx4 75 with 36MB ram -> k6 2 500 with 512MB ram -> Athalon 64x2 with 4GB ram -> i7 3770k with 32GB ram so it really has been massive jumps in performance at every stage. I don't really chase technology

Re:When do we get a real boost over 2013 speeds?

[Anonymous Coward]

...hasn't really been worth it for the enthusiast upgrades, but if the hypetrain is right the skylake enthusiast should finally be worth it. Intel is back to pushing performance over power saving this time around, so adding up all the incremental ivy bridge, haswell, broadwell improvements plus skylake should make it worth doing the whole upgrade,but I'll be watching the benchmarks like a hawk.

AMD: they're sunk unless their new arch has VERY GOOD IPC efficiency WITH a SUBSTANTIAL reduction in power. Same goes for their GPUs. OTOH a complete AMD box makes an awfully good space heater...

Re:When do we get a real boost over 2013 speeds?

[FlyHelicopters]

Wow, you really are dedicated to major jumps.

Fair enough, more power to you and all. :)

The 486/75 to K6-2 500 is a HUGE jump, I personally couldn't have waited that long. :)

Re:When do we get a real boost over 2013 speeds?

[rbrander]

The doubling is impressive (except for it taking 6.5 years, if you computed through the 90s) but the power consumption is irrelevant. I'm not running a data centre, and I can afford a $2500 computer, so paying 130/1000 X 9 cents per hour = 1.17 cents per hour to run the chip is not worth mentioning.

Re:When do we get a real boost over 2013 speeds?

[rbrander]

I'm using PostGIS (PostgreSQL with a plug-in) for the back-end, and QGIS for the client. With the above-mentioned system and 32GB it utterly blows away the performance I get at work from a big ESRI server and Oracle with ESRIs SDE plug-in. That's probably because the corporate server is throttled per-user, but still, it means that home-user GIS for zero software costs is really here in convenient form. I've developed a "mapping system" that loads in about 50 layers for my city with one script (works in Mac, Win, Linux, slight script changes) and co-workers are starting to ask me for it on their laptops. It actually runs fairly well on an i5 and 16GB, so it's possible I won't even get that much speedup from a faster chip...but I'm game to spend money trying!
The suggestion above that I move the PostGIS db file from SSD to ramdisk is a very worthy one; I'll be trying various experiments, but since I date back to 6502 chips, "get the latest" has always been step 1.
If you want to contact about GIS, my home page is at the Calgary Unix User's Group: http://www.cuug.ab.ca/branderr...

Re:When do we get a real boost over 2013 speeds?

[FlyHelicopters]

Perhaps not, but at 8 hours a day, 5 days a week, 40 hours a year, that is more than $50 over 3 years, which more or less wipes out any savings from having purchased the AMD chip over the Intel chip.

It increases your carbon footprint (if you care about that sort of thing), and it is twice as bad if you have to AC the space, since that same heat must now be cooled, nearly doubling the cost to $100 over 3 years. That is somewhat removed if you otherwise would have heated the space of course.

Re:When do we get a real boost over 2013 speeds?

[Bob the Super Hamste]

I had to as I was poor and couldn't afford to replace things. The 486 worked well enough for what I was doing with it at the time but it did get flaky. I am one of those really odd people who will use something until it is worn out and the repair costs exceed what it would cost to replace it. I am really not dedicated to the huge jumps it is just I grew up poor and am too cheap to replace a functional device until it becomes non functional.

Re:When do we get a real boost over 2013 speeds?

[Anonymous Coward]

This actually hurts my brain a bit... If you get 64GB of RAM, instead of using 32GB for a swap file, just turn swapping off and you'll get better performance and dynamic allocation of memory. A swap file is used to take stuff from RAM and put it in the HDD; so making a virtual HDD in your RAM to take the stuff that's supposed to be written in your RAM and put it back in the RAM again is just convoluted and adds overhead. Also, for nearly every single use case, with 32GB of RAM, you can turn swapping off and never notice a performance hit except for ridiculous workloads that force you to swap more or less everything other than the application running... and even then, 64GB over 32 isn't likely to make a difference in that behavior.

Re:When do we get a real boost over 2013 speeds?

[Dadoo]

AMD did make a 133 MHz version, but called it "Am5x86-P75"

That may have been the official name, but many people who sold them called them DX4/133s. I've certainly never heard them called "Am5x86-P75" before, and I've been in the business since the early 80s.

Re:When do we get a real boost over 2013 speeds?

[FlyHelicopters]

I owned a small computer networking business back in the 90s, we sold a LOT of those chips, and we of course called them P75s... :)

What am I missing??

[Billly Gates]

I don't get it. These are slower and a downgrade from other 2013 cpus for socket 1150. So why would I upgrade from what I have? Did I miss something special? If you can't afford an x99 it would make more sense to get a 2 year old i7 4770

Re:What am I missing??

[halivar]

These are budget CPU's for low power consumption. The initial Broadwell offering was for mobile, and this is their first desktop offering, targeting business and office with pretty decent integrated graphics, but nothing you'd want for gaming. You would stick with your i7 4770 until the 14nm line begins targeting performance computing.

Re:What am I missing??

[TechyImmigrant]

I don't get it. These are slower and a downgrade from other 2013 cpus for socket 1150. So why would I upgrade from what I have? Did I miss something special? If you can't afford an x99 it would make more sense to get a 2 year old i7 4770

The slowest [new gen] cpu is slower than the fastest [last gen] cpu. This is normal. The bazillion core global warming Broadwell will presumably come out later. It's faster per core and lower power.

Re:What am I missing??

[Moof123]

They are not priced like budget CPU's, in fact they look like a 5% price increase from what I could dig up. So you can pay more for a top of the line CPU with slight less awful integrated graphics, great...

Intel is really not blowing my skirt up.

L4 cache!

[Anonymous Coward]

I can see not far in the future DRAM will be replaced with flash. Not some fancy dram+flash combo, just a simple flash storage, since the cpu has enough fast on die already.

Re:L4 cache!

[sribe]

I can see not far in the future DRAM will be replaced with flash. Not some fancy dram+flash combo, just a simple flash storage, since the cpu has enough fast on die already.

That's a horrible idea--flash memory write characteristics just don't match that use at all.

Re:L4 cache!

[DigiShaman]

Actually, if NAND memory become fast enough, that could replace RAM in future smart phones. It would be like having the concept of Core Memory all over again. And what, NAND storage holds data 6 to 12 months powered off? So still volatile compared with flash or magnetic storage.

Re:King Frosty, First poster, who is royal

[Anonymous Coward]

How dare you let these words of blasphemy pass your lips! I'm King Frosty, The First King! I'm King! I'm Royal!

Intel AMT/VPRO/VT

[Anonymous Coward]

Intel AMT/VPRO/VT

On chipset VNC server. Ram Upload. etc.

AKA Backdoor.

FUCK YOU AMERICUCKS AND YOUR FEMINIST POLICE STATE.

Re:Intel AMT/VPRO/VT

[ledow]

If someone is able to talk directly to your computer's network ports without your permission, you have bigger problems than a VNC-like option in your BIOS that you can turn off.

Seriously.

Similarly for uploads, etc. There's a reason that we have things like authenticating proxies, firewalls and all that other jazz that people moan about.

If you're vaguely techy and think this means it's beyond your control, you really shouldn't be on this website.

Re:Intel AMT/VPRO/VT

[Anonymous Coward]

3G cuck shill.

Re:Intel AMT/VPRO/VT

[Anonymous Coward]

So basically you're saying it's ok that it is impossible to lockdown an Intel based computer.

Yep you are a government payed shill and you should be killed in a painful manner publicly.

Re:Intel AMT/VPRO/VT

[Anonymous Coward]

Fuck your Queen...

In a hockey stick situation like this...

[sshir]

... AMD has a chance for survival. I think.
Basically every few additional percentage points in performance cost untold billions in investment. Thus it is possible to tailgate market leader by producing something only a little bit slower by spending half the money. As long as performance/watt, performance/rack are not outrageously bad (so data centers will not shun you) AND Intel does not engage in monopolistic tactics (big question mark here) it's possible to make a decent living.

Re:In a hockey stick situation like this...

[Anonymous Coward]

But performance/watt is exactly the metric which Intel has been improving for the past 7 years or so, while all but ignoring max performance. They more than double AMD's efficiency (and that's being generous to AMD), and have all but kicked them out of data centers and laptops because of it. AMD has some competitive desktop offerings because efficiency isn't nearly as big as performance/dollar there, and they can compete with Intel's low end there.

Why Intel generally thumps AMD in business

[sjbe]

So why is AMD constantly on the verge of bankruptcy?

Because AMD has historically made their business model making a product that is compatible with another company's product and that other company (Intel) has a cost advantage in making the product and generally controls the architecture. Intel is actually quite the manufacturing juggernaut in microprocessors whereas AMD has basically no manufacturing of their own. Intel also has a lead in die size as well so AMD is typically playing catch up. Intel basically can make a smaller, faster processor cheaper and sell it for less any time they want to. Hard to compete effectively with that. AMD has to be smarter than Intel and they haven't shown themselves to be capable of doing that on a consistent basis. Even when their designs have been better, Intel has been able to leverage their die size advantage to overcome design deficiencies. Furthermore they've made some pretty bad tactical business errors (the acquisition of ATI hasn't been the smoothest) and Intel has been known to engage in some arguably shady business dealings with their customers.

Basically probably the only reason AMD is still with us is that Intel doesn't want the anti-trust scrutiny that would come with killing them off. Having AMD around gives Intel a "credible" competitor, albeit one that hasn't shown any meaningful ability to compete consistently. AMD has been trying to diversify away from just PC microprocessors for a while now with mixed success.

Re:Why Intel generally thumps AMD in business

[nitehawk214]

Actually for a while it was the other way around. AMD pioneered x86-64 and Intel was the one playing compatible catch-up when they tried to bank on IA-64 and it tanked badly.

However AMD managed to squander any gains they had their and have fallen to the distant #2 once again.

Re:Why Intel generally thumps AMD in business

[Anonymous Coward]

Not to mention that for a while AMD did have Intel trounced and Intel cheated by rigging the compilers so they deactivated all the optimizations whenever a non-Intel processor was used so the benchmarks would all lie and say Intel won when they really didn't.

Intel just recently paid a settlement over that, but still actually won overall due to it due to all the lost sales and PR AMD would have received.

Re:Why Intel generally thumps AMD in business

[Anonymous Coward]

It is not arguable.

Bottom feeding

[Anonymous Coward]

I've been bottom feeding and enjoying my 6 core AMD FX-6300 for $110 for a couple of years. But I'm just a l33t haxor not a gamer. The performance gains for Intel are so non-linear to price it isn't a difficult decision.

We can all be generals now.

[some old guy]

Like David Petreus, we can all now have a Broadwell under our desks.

Built-in chip level backdoor (VPro/VT/AMT), no sys

[Anonymous Coward]

Built-in chip level backdoor (VPro/VT/AMT), no systemd required.

So answer me this...

[WSOGMM]

Will i5 prices go down this month?

I'm building a gaming/mini-simulation computer, and I have a mix of poor student syndrome along with excessive computer drooling disease (much more debilitating). Basically all I want is a fast GPU (nvid gtx 960), but I can't help but want a fast processor too, so I've been comparing the i5's, i3's and pentium G3450's.

Should I wait to buy an i5? or should I stick with the cheap processor? or maybe you know of a motherboard/cpu combo deal that will cut the cost of an i5 just enough?

*napkins mouth*

Re:So answer me this...

[halivar]

For gaming purposes, the processor is definitely not your bottleneck, as long as it supports enough PCI Express slots to fulfill the needs of your video cards.

Re:So answer me this...

[halivar]

*err PCI lanes

Re:So answer me this...

[Anonymous Coward]

Will i5 prices go down this month?

I'm building a gaming/mini-simulation computer, and I have a mix of poor student syndrome along with excessive computer drooling disease (much more debilitating). Basically all I want is a fast GPU (nvid gtx 960), but I can't help but want a fast processor too, so I've been comparing the i5's, i3's and pentium G3450's.

Should I wait to buy an i5? or should I stick with the cheap processor? or maybe you know of a motherboard/cpu combo deal that will cut the cost of an i5 just enough?

*napkins mouth*

Find a trustworthy fellow gamer and buy their old rig. Plenty of us are latent hoarders. If you were in near me, I could unload an x6 3.4 Ghz 16GB box for $250 with quality case+PS+2TB drive. Since you are not, and you don't know me, it would be foolish to broker that deal. For more cash, I'd have a 8x 4Ghz 32gb box. I get lazy and don't want to bother swapping motherboard/cases so I just buy a shiny new case and have 2 systems... (9 really, but I digress). The time to carefully dismount and remount the parts and reinstall everything is more expensive than any reasonable/nice case so I'd rather have an extra than burn the price in time and have 1.

dual socket X 16 core = 32 cores

[raymorris]

A dual socket mobo with 16-core AMD CPUs in each socket will probably spank your current Intel system. That's one area AMD excels at, they sell 8 and 12 core CPUs cheap, 16-core if you're serious.

oops, they left the chipset behind

[slashmydots]

I doubt they're actually faster at all graphics operations than AMD. A Kaveri APU has a memory controller that runs natively at 2400MHz. These new i5 and i7 CPUs run at an utterly pathetic 1600MHz. When I changed a trinity APU system's memory from 1600 to 2133, the graphics rating in Windows 7 went up 0.7 points. Memory bandwidth is twice as important when you're sharing it with the rest of the system for normal operations as well as graphics operations. They reeeeeally need to upgrade their CPUs and chipsets to handle higher memory frequencies.

Re:oops, they left the chipset behind

[Anonymous Coward]

http://www.anandtech.com/show/7364/memory-scaling-on-haswell/6

Not really with intel. their memory controller is way more efficient than AMD's

Re:Broadwell is yesterday's news. Bring on Skylake

[sexconker]

Idiots are waiting for HDMI 2.0.
People with brains are waiting for DisplayPort 1.3.

64 Bit x86

[sjbe]

Actually for a while it was the other way around. AMD pioneered x86-64 and Intel was the one playing compatible catch-up when they tried to bank on IA-64 and it tanked badly./quote.

That situation lasted for all of about 1-2 years and even then AMD never really were able to capitalize on it because Intel was better capitalized, and had cost advantages and 64 bit didn't matter enough at the time. While it was a misstep by Intel it wasn't one they couldn't recover from. Intel putting a 64 bit version of the x86 wasn't exactly a huge technical challenge for them. Intel has made a number of mistakes over the years but AMD simply has never been smart enough or well funded enough to make Intel pay for them.

Re:64 Bit x86

[Kremmy]

I was bummed as hell that they weren't able to capitalize on it. I watched Apple jump from the 64bit PowerPC G5 CPUs to 32bit Intel CPUs, which I considered a HUGE step backwards. AMD was the only company producing a 64bit desktop CPU that wasn't the G5 at the time, and they got glossed over. I think we'd be seeing a very different field right now if AMD had won the Apple contract. Hoping they can capitalize on their gains being the supplier for the console chips and start giving Intel some real competition again.

Good. Older parts will come down in price.

[msobkow]

It has been a very long time since I bought a "latest and greatest" chip when building a new computer, because 2-3 revisions old still has many times the performance of the machine it's replacing and far more "snort" than I could ever use on day-to-day activities.

With any luck, this announcement and release will bring the price down on the chips I want by another $100 or so by January-February, when I hope to actually be building a new machine.

The bleeding edge is fine for gamers and hard-core video encoders and number crunchers, but for the rest of us folks, it is just an insane waste of money to buy the "latest and greatest." It's been a lot of years since anyone needed to do that for anything even vaguely resembling sane home or business use.

Re:Broadwell is yesterday's news. Bring on Skylake

[Shirley Marquez]

Both HDMI and DisplayPort are worth having. HDMI is what TV sets have; if you want to attach an affordable big UHD screen to your computer, HDMI 2.0 (the new version that supports 4K) is what you need. (You can use a adapter but a native HDMI port is more convenient.) Computers displays have a variety of things: DisplayPort, Thunderbolt, HDMI, and DVI-D (plus those legacy displays with analog VGA connections), so you're going to need adapters or adapter cables as often as not.

In the future all the computer stuff is likely to move to the just-announced Thunderbolt 3, the one port to rule them all. It uses the new USB-C connector and is backward-compatible with both Thunderbolt 1 and 2, with DisplayPort, and with USB 3.1 and earlier versions. Most of those will need adapter cables.

Re:Broadwell is yesterday's news. Bring on Skylake

[sexconker]

DP is superior to HDMI. Yes, trash TVs have HDMI, but that doesn't change the fact that DP is the better choice every single time.

As for Thunderbolt 3 taking over everything? Intel can't let a spec sit still for more than 6 months. It would take 6 years minimum for OEMs to adopt Thunderbolt 3 on hosts and peripherals to the point that they feel safe using it as the primary connection for everything. And by then we'll have Thunderbolt 9 (still over copper instead of optical).
And of course, there's no incentive for OEMs to do this - USB 3 is fine, DisplayPort 1.2a and 1.3 are fine, and Ethernet is fine. Thunderbolt simply costs way too much to implement (especially if you want to make a cable longer than a few feet). Daisy chaining shit is a novelty, and it's a novelty that's supported by USB, DisplayPort, etc. The mass market doesn't need 40 Gbps. The mass market hardly needs 10 Gbps of USB 3.1 for data or 32.whatever Gbps of video+audio+whatever of DP 1.3.
Thunderbolt is FireWire all over again. Niche and expensive.

Re:Broadwell is yesterday's news. Bring on Skylake

[Shirley Marquez]

Good luck if your DP adapter goes missing when you have a midnight panel or movie showing to do. You probably can't just pop down to a local store and buy one, but you can get HDMI cables anywhere. Around here you can buy them 24/7 because they have them at CVS, and I'm sure Wally World also sells them. That's why I'd rather have both ports on my system, and if I have to have just one I'd rather have HDMI.