AMD CPU speculation... and expert conjecture

[juanrga]

@noob2222, your questions are otff-topic. I did send you PM with answers.

@blackkstar, one cannot compare two chips on two completely different laptops from two companies: changes in the chassis, on the distribution of components on the mobo, on the quality of the cooler, etc. can do that one chip throttles whereas other doesn't.

The NUC review is more accurate to compare Broadwell vs Haswell.

 

[juanrga]

We have some actual "CPU" news guys. I don't know about you guys, but if this is real, I'm excited!

http://www.kitguru.net/components/cpu/anton-shilov/amds-zen-could-sport-intels-skylake-technologies-new-512-bit-fpu-rumours/

First, WCCFTECH said nothing really. Nowhere WCCFTECH said anything about AVX-512.

Second, even if Zen supports AVX-512 it doesn't not imply that Zen has a 512-bit FPU.

In short, we are at a new iteration of hype once again. When Jaguar announced 256-bit AVX support many people believed that Jaguar was incorporating bigger FPU. Wrong. Jaguar has one 128-bit FPU per core.

When AVX2 support was leaked for Excavator, again the hype that Excavator core would bring a new twice bigger powerful FPU was published

http://www.xbitlabs.com/news/cpu/display/20131018224745_AMD_Excavator_Core_May_Dramatic_Performance_Increases.html

What I said then? That Excavator used the same FPU than Steamroller. Well Excavator is here and... I was right: the FPU has not changed.

The chances of Zen supporting AVX-512 are 0.0000001% and even if it does, this doesn't imply a 512-bit FPU.

I continue maintaining my prediction/guess that Zen is a 2x 128-bit core, because I have no data suggesting otherwise.

 

[blackkstar]

@faye, 5.1ghz continuous in Windows, I do benching at 5.3ghz or so though. In Gentoo I only run 4.8ghz or so because I need it to be 100% stable since I compile the whole OS myself and if there's some instability in any of the chip (including AVX stuff), the whole OS gets really unstable and has problems until I recompile what was broken with a table system. I have a custom loop so I generally I see pretty cool temps, well under 80c. Sometimes the radiator gets dirty and temps get hot before I clean.

@juan, so the chassis design is more important than the CPU design, since Ivy Bridge with better chassis can be faster than Broadwell? Fascinating. Perhaps Intel should just stop designing new CPUs and start designing chassis then :x

Even being serious, I don't quite get your point. Macbooks are notorious for overheating and throttling because of chassis that prioritize being thin and light over properly cooled, and even Ivy Bridge Mac Book didn't throttle before Broadwell.

Compare the results for "macbook overheating" with "thinkpad x1 carbon overheating". Looks to me like one of these models has problems with temps and the other doesn't. But it's not the one you want to believe.

This is one of the most thorough reviews I've seen of Broadwell and it looks absolutely terrible. The only thing it looks like it has going for it is longer battery life under load because it throttles itself.

I would think that if Intel's smaller nodes were improving, we'd be seeing cooler, more efficient chips come out that clock better. Instead, we've seen peak overclock frequency diminish since Sandy Bridge on 32nm. I'm looking forward to seeing how poorly Broadwell on desktop overclocks. I am guessing it's going to be abysmal, low 4ghz peak range on average. And I know you're telling yourself there's no Broadwell desktop chips because there's no demand, but you're just going to completely ignore the fact that it might be because Intel is having problems scaling 14nm to higher frequencies so they release low clocked mobile parts.

We will see what happens when we do get Broadwell-K. I have a really good feeling it's going to be a lot of disappointment. When a company releases a chip that clocks well, it's usually met with a bunch of PR BS about world record overclocks. Instead, Intel 14nm is some mobile low TDP parts that over thermal throttling when the traditional models don't.

And by the way, just for bonus. Thinkpad X1 Carbon has been using the same (or extremely similar) chassis for some time with older Intel CPUs and not having terrible thermal throttling issues like the Broadwell is having.

 

[jdwii]

"AMD is hiring a CPU performance engineer to optimize its GPU drivers"

http://www.tweaktown.com/news/43680/amd-hiring-cpu-performance-engineer-optimize-gpu-drivers/index.html

 

[gamerk316]

Did you know that even after being caught at it, their C compiler STILL turns off CPU optimizations for AMD chips? All they did was post a disclaimer at the bottom of the download page. And tellingly, the disclaimer isn't text, but a graphic image of text, so you can't copy and paste it if you're outraged.

Did you know the Intel C compiler gives AMD the most performance boost via optimizations then any other compiler?

*crickets*

Intel doesn't as aggressively optimize for AMD because ICC is very tightly integrated with Intels CPUs. Because the underlying HW is different, low level optimizations that benefit Intel chips could theoretically hurt AMD ones.

And I again note (because this isn't the first time this came up): Intel is under no obligation to even support AMD with their own proprietary compiler.

 

[jdwii]

Its just an excuse for people to use gamer. We both have been on this forum for years the bulldozer forum and this one they use that excuse over and over again. I've been to Amdzone(never again tomshardware FTW) and i'll never go back they use excuses after excuses to support their claims.

Luckily Amd knows better and they are getting rid of their current design in favor of Zen and K12 we will see how that goes a lot of rumors are coming up with skylake and i personally think its a bit much. Gamer we both know this time around Amd isn't hyping Zen or K12 its hush hush and that has me worried and at the same time it makes me happy. With bulldozer we had a lot of bad marketing and with the 7970 we had a rush release. It makes me wonder how Zen will perform.

 

[palladin9479]

The ICC issue was a problem years ago, and no it has nothing to do with "tight integration", it was a business strategy and like all business strategies is motivated purely by monetary gain. This isn't a bad thing, business's exist primary to make money and should be motivated by profit, otherwise they would be charities. It's our job as consumers to be ever vigilant for shenanigans and vote with our wallet accordingly. They were caught, got bad press, were forced to publicly inform consumers and that's the important part.

Nowadays for the MS Windows platform, compiler typically won't make much of a difference one way or another. The largest gains are from x87/MMX to SSE2 and now even the ICC compiles for SSE2 on AMD CPUs. Only in some exotic math applications that utilize an older Intel math library would you see issues, and those are far enough removed as not to be a common case. Agner did a very good breakdown of this awhile back.

 

[gamerk316]

Well:

http://www.pcper.com/news/Graphics-Cards/GPU-Market-sees-20-point-swing-2014-NVIDIA-gains-AMD-falls

Ouch. Wonder how much of that is the ending of the crypto-currency bubble?

 

[juanrga]

Did you know that even after being caught at it, their C compiler STILL turns off CPU optimizations for AMD chips? All they did was post a disclaimer at the bottom of the download page. And tellingly, the disclaimer isn't text, but a graphic image of text, so you can't copy and paste it if you're outraged.

Did you know the Intel C compiler gives AMD the most performance boost via optimizations then any other compiler?

*crickets*

Intel doesn't as aggressively optimize for AMD because ICC is very tightly integrated with Intels CPUs. Because the underlying HW is different, low level optimizations that benefit Intel chips could theoretically hurt AMD ones.

And I again note (because this isn't the first time this came up): Intel is under no obligation to even support AMD with their own proprietary compiler.

Yearsa go Intel ICC was ahead of any other compiler, specially on numeric code. This is no longer true. There are even workloads where ICC gives less performance than alternative compilers.

Intel ICC issue was not technical but business (unfair competition practice):

Intel has designed its compiler purposely to degrade performance when a program is run on an AMD platform. To achieve this, Intel designed the compiler to compile code along several alternate code paths. Some paths are executed when the program runs on an Intel platform and others are executed when the program is operated on a computer with an AMD microprocessor. . . . By design, the code paths were not created equally. If the program detects a "Genuine Intel" microprocessor, it executes a fully optimized code path and operates with maximum efficiency. However, if the program detects an "Authentic AMD" microprocessor, it executes a different code path that will degrade the program's performance or cause it to crash.

It was proved that AMD and VIA processor run faster when the default behavior of the Intel compiler was disabled. The unfair practice was discovered by Agner and others experts and Intel tried to negate/hide it for years. Finally Intel was obligated by court to add the famous disclaimer.

Of course, Intel is not obligated to support AMD. That is not the point gamerk. The point is that Intel lied and disseminated false documentation, and was caught doing it. This is part of the US Federal Trade Commission (FTC) antitrust complaint against Intel:

Intel sought to undercut the performance advantage of non-Intel x86 CPUs relative to Intel x86 CPUs when it redesigned and distributed software products, such as compilers and libraries.
[...]
To the public, OEMs, ISVs, and benchmarking organizations, the slower performance of non-Intel CPUs on Intel-compiled software applications appeared to be caused by the non-Intel CPUs rather than the Intel software. Intel failed to disclose the effects of the changes it made to its software in or about 2003 and later to its customers or the public. Intel also disseminated false or misleading documentation about its compiler and libraries. Intel represented to ISVs, OEMs, benchmarking organizations, and the public that programs inherently performed better on Intel CPUs than on competing CPUs. In truth and in fact, many differences were due largely or entirely to the Intel software. Intel’s misleading or false statements and omissions about the performance of its software were material to ISVs, OEMs, benchmarking organizations, and the public in their purchase or use of CPUs. Therefore, Intel’s representations that programs inherently performed better on Intel CPUs than on competing CPUs were, and are, false or misleading. Intel’s failure to disclose that the differences were due largely to the Intel software, in light of the representations made, was, and is, a deceptive practice. Moreover, those misrepresentations and omissions were likely to harm the reputation of other x86 CPUs companies, and harmed competition.
[...]
Some ISVs requested information from Intel concerning the apparent variation in performance of identical software run on Intel and non-Intel CPUs. In response to such requests, on numerous occasions, Intel misrepresented, expressly or by implication, the source of the problem and whether it could be solved.
[...]
Intel’s software design changes slowed the performance of non-Intel x86 CPUs and had no sufficiently justifiable technological benefit. Intel’s deceptive conduct deprived consumers of an informed choice between Intel chips and rival chips, and between Intel software and rival software, and raised rivals’ costs of competing in the relevant CPU markets. The loss of performance caused by the Intel compiler and libraries also directly harmed consumers that used non-Intel x86 CPUs.

 

[juanrga]

@ blackkstar, this is an AMD thread. I will answer your off-topic questions about the evolution of Intel chips with PM.

If I am allowed to change my thoughts about Kaveri refresh, I have new data. The 7850K will not be the faster model, but AMD will release a A10-8000K series with improved clocks. The CPU clock could be increased to 3.9GHz (4.1GHz turbo) and the GPU could increase to 867MHz.

This model will be more marketing than real performer part. Apparently AMD want to present its first teraflop APU for consumers. The A10-8000K series would hit near 1013 GFLOP/s.

 

[etayorius]

@ blackkstar, this is an AMD thread. I will answer your off-topic questions about the evolution of Intel chips with PM.

If I am allowed to change my thoughts about Kaveri refresh, I have new data. The 7850K will not be the faster model, but AMD will release a A10-8000K series with improved clocks. The CPU clock could be increased to 3.9GHz (4.1GHz turbo) and the GPU could increase to 867MHz.

This model will be more marketing than real performer part. Apparently AMD want to present its first teraflop APU for consumers. The A10-8000K series would hit near 1013 GFLOP/s.

If they want to present an Upgrade to the 7850k they could at least make it 4GHZ base and 4.3/4.4 Turbo, otherwise this upgrade is BS. Also, why not call it 7870k? they want to make it sound as if it was a much better performer.

Why the hell have they not make the PS4 APU version into desktop with faster clocks? they also promised this waaay back.

 

[logainofhades]

Also, why not call it 7870k? they want to make it sound as if it was a much better performer.

Why the hell have they not make the PS4 APU version into desktop with faster clocks? they also promised this waaay back.

If it is Excavator, it would get a new number, just like Kaveri did, since it was steamroller, and not Piledriver. The one maybe to complain about is Trinity and Richland. Same CPU cores, but improved GPU, in Richland.

 

[blackkstar]

@ blackkstar, this is an AMD thread. I will answer your off-topic questions about the evolution of Intel chips with PM.

If I am allowed to change my thoughts about Kaveri refresh, I have new data. The 7850K will not be the faster model, but AMD will release a A10-8000K series with improved clocks. The CPU clock could be increased to 3.9GHz (4.1GHz turbo) and the GPU could increase to 867MHz.

This model will be more marketing than real performer part. Apparently AMD want to present its first teraflop APU for consumers. The A10-8000K series would hit near 1013 GFLOP/s.

If they want to present an Upgrade to the 7850k they could at least make it 4GHZ base and 4.3/4.4 Turbo, otherwise this upgrade is BS. Also, why not call it 7870k? they want to make it sound as if it was a much better performer.

Why the hell have they not make the PS4 APU version into desktop with faster clocks? they also promised this waaay back.

I'd wager it's just a marketing thing. AMD looks to be ending their bulldozer family chips with 9000 series parts. We already saw this with 9590 and 9370 chips. AMD will sort of do a rebranding type of thing if they have to.

My guess is

8000 series APUs are Kaveri rebrands
9000 series APUs are Excavator
AMD will come up with a new naming scheme when they release Zen after Excavator.

As for Intel 14nm, I'll believe it when I see it. Someone mentioned Broadwell is not so bad because Haswell has VRMs on CPU die, but Broadwell removed them.

 

[juanrga]

@ blackkstar, this is an AMD thread. I will answer your off-topic questions about the evolution of Intel chips with PM.

If I am allowed to change my thoughts about Kaveri refresh, I have new data. The 7850K will not be the faster model, but AMD will release a A10-8000K series with improved clocks. The CPU clock could be increased to 3.9GHz (4.1GHz turbo) and the GPU could increase to 867MHz.

This model will be more marketing than real performer part. Apparently AMD want to present its first teraflop APU for consumers. The A10-8000K series would hit near 1013 GFLOP/s.

If they want to present an Upgrade to the 7850k they could at least make it 4GHZ base and 4.3/4.4 Turbo, otherwise this upgrade is BS. Also, why not call it 7870k? they want to make it sound as if it was a much better performer.

Why the hell have they not make the PS4 APU version into desktop with faster clocks? they also promised this waaay back.

Marketing, the APU 8000 series will be APU 7000 with minor clock increases just to sat that they have something new. Just as the FX-9000 series was FX-8000 series with higher clocks.

The desktop version of the PS4 APU is in the folder of canceled products.

If it is Excavator, it would get a new number, just like Kaveri did, since it was steamroller, and not Piledriver. The one maybe to complain about is Trinity and Richland. Same CPU cores, but improved GPU, in Richland.

Kaveri refresh is Steamroller all the day.

 

[logainofhades]

So it is the same as Trinity to Richland then? I really haven't kept up on the FM2+ stuff, as I haven't had much interest in them. I am generally not a fan of IGP.

 

[etayorius]

So it is the same as Trinity to Richland then? I really haven't kept up on the FM2+ stuff, as I haven't had much interest in them. I am generally not a fan of IGP.

Less than that since at least Richland got better iGPU and some really small tweaks to the Arch, in the A10 8000 it is just a small bump of Speed on both CPU and iGPU, very disappointing... like 95% of the stuff that AMD throws now a days.

 

[jdwii]

Also, why not call it 7870k? they want to make it sound as if it was a much better performer.

Why the hell have they not make the PS4 APU version into desktop with faster clocks? they also promised this waaay back.

If it is Excavator, it would get a new number, just like Kaveri did, since it was steamroller, and not Piledriver. The one maybe to complain about is Trinity and Richland. Same CPU cores, but improved GPU, in Richland.

Lets hope for some better prices, and more clock speed. Be nice to make a 350-400$ build using this, wondering if GTA5 will perform well on it.

 

[con635]

I've read the 7850k's cpu down clocked when the gpu was loaded regardless of bios settings and it shows no performance gain in benchmarks when clocked above 4.5ghz, is this true? Is this where the refresh gets more performance?

 

[juanrga]

I've read the 7850k's cpu down clocked when the gpu was loaded regardless of bios settings and it shows no performance gain in benchmarks when clocked above 4.5ghz, is this true? Is this where the refresh gets more performance?

Due to a bug on the power management, the CPU and GPU cannot sustain the frequencies at once. The bug has not been completely fixed in Kaveri refresh but there are some firmware quirks that would help the new models (it is exactly the same die after all) to get more performance.

I am warned to don't expect miracles. Everyone is saying me that the improvement Kaveri --> KaveriRefresh will be smaller than Trinity --> Richland.

 

[juanrga]

http://www.benchmark.pl/aktualnosci/amd-fx-8800p-wyniki-wydajnosc-3dmark-11-procesor-apu-carrizo.html

Prototype laptop with FX-8880P Carrizo (successor to FX-7600P Kaveri).

GPU score is around 28% better, which agrees with the expected 30% gain from texture compression.

 

[Reepca]

~30% performance increase over fx-7600p, if it applies in the same way to real-world gaming benchmarks, would put battlefield 4 @ 1280x720 low settings at 33 fps (average). Some would say that's playable, but it *is* a first-person shooter, and typically requires smoother play to be fully enjoyable.

It's a shame, maybe mainstream low-end mobile gaming is gonna have to wait till post-carrizo. That said, battlefield 4 is a fairly demanding title, so plenty of other older games should run smoothly - I'm sure there's a place for it, but it's in an awkward place between low-end mainstream gaming and entry-level gaming. It can't run AAA titles very smoothly yet, but is *really* close. If I can find it in a laptop with a 1080p screen for < $700, I'd consider it worth getting. Assuming it shows up at all, unlike the fx-7600p, which *STILL* is not for sale in America.

 

[griptwister]

@faye__kane, From what I understand, they're developing a new architecture as it is. Why not throw something on that's better than their competition? Maybe offer a bit more future proofing. I could see them doing 256-bit instead how ever, as 512 is a large number. Haha! And yeah, But I think we should remember, if AMD was in the same position Intel was in, they'd be doing the same thing. I have nothing against Intel, my next upgrade might be one of their CPUs. (Skyrim framerate dips get annoying haha). however, my current rig will be put to life in a steam box mod I am scheming up.

 

[de5_Roy]

AMD outs a few Carrizzo circuitry details
ISSCC 2015: How and why not speeds and feeds
http://semiaccurate.com/2015/02/23/amd-outs-carrizzo-circuitry-details/
http://www.anandtech.com/show/8995/amd-at-isscc-2015-carrizo-and-excavator-details
http://www.tomshardware.com/news/amd-carrizo-apu-excavator-processors-cpu,28608.html
http://www.pcworld.com/article/2887932/amd-claims-carrizo-chip-offers-substantial-performance-battery-life-boosts.html
same promo slides, better text.

 

[-Fran-]

Interesting news around Carrizzo.

Sounds like it wasn't all gloom and doom after all. Now we just need some benchies. It's a pity though there are no news about desktop.

Cheers!

 

[de5_Roy]

Interesting news around Carrizzo.

Sounds like it wasn't all gloom and doom after all. Now we just need some benchies. It's a pity though there are no news about desktop.

Cheers!

most of the doom and gloom was BS anyway.

amd had already said that it'd use HDL for excavator, now we know in more detail why HDL was chosen. amd said back then that it'd get 30% more area gain on the same process, they got 29% (i think there's still some PR trickeries in how amd measured the gains ). from 2012 AT article:

The power savings comes from not having to route clocks and signals as far, while the area savings are a result of the computer automated transistor placement/routing and higher density gate/logic libraries.

The tradeoff is peak frequency. These heavily automated designs won’t be able to clock as high as the older hand drawn designs. AMD believes the sacrifice is worth it however because in power constrained environments (e.g. a notebook) you won’t hit max frequency regardless, and you’ll instead see a 15 - 30% energy reduction per operation. AMD equates this with the power savings you’d get from a full process node improvement.

We won’t see these new libraries and automated designs in Steamroller, but rather its successor in 2014: Excavator.

the lower peak clockrate isn't surprising.
carrizo is even more mobile oriented than kaveri. all those miliwatts of gains turn into better battery life and the upgraded power and clockrate management help too. but the 5% IPC gain won't mean much for desktop power users. however, i think we'll see carrizo in "desktop pc"s.

edit2:
in the AT article amd didn't seem too eager to reduce bd's L3 latency issue even though it may have figured those out already....