AMD accuses BAPCo and Intel of cheating with Sysmark benchmarks

[Prysin]

back in 2011, what exact markets were nVidia and Intel competing in?!

HPC

 

which is based upon complex specialized codes. Sort of like, oh yeah, exactly the kind of stuff synthetic benchmarks tests

[zMeul]

Tabletmark tests tablets, including Android of course.

 

Anyway, downplaying GPU performance on PC negatively affects Nvidia too.

I stand corrected, but Intel has a flimsy presence in the Android market - so, your argument doesn't stand

some manufacturers (ASUS if I recall) do some Android HW with Intel CPUs, but they're not great - I don't recall ever seeing Intel saying benchmarks weren't fair to them 

[Prysin]

Intel would be leaving itself open to slander and liable as anticompetitive practices. AMD is way too small to fall into contention with that branch of antitrust law. And Intel doesn't throw wild accusations. When was the last time Intel sued someone or claimed abuse of software/hardware standards? Intel doesn't have a habit of bringing up accusations for appearance. AMD does. Look no further than Richard Huddy's lies about Gameworks making it impossible for AMD to optimize. Nvidia replied saying they do most of theirs on binary files anyway, and now in a couple cases AMD is beating Nvidia in some Gameworks games on the high end. This accusation smells like a red herring to me. And AMD would stand to win more money from such a suit than it would lose. Intel is big enough that the damages payable for antitrust behaviors are in the ~100 million USD range for simply farting in the wrong direction. This is getting free advertisement and is borderline slander until proof is presented. At least Apple is publicly quiet about its accusations of other companies. They keep it in the courtroom and let judges decide. If AMD had a habit of doing that, I'd be far less skeptical.

BUT, if we look at intels previous strategy. Then they have always come down on AMD the moment AMD reaches roughly 20-25% marketing share, and just by brute force either tanked prices, cheated in benchmarks or simply marketed so hard that AMD drops down to their "usual" 5-10% market share.

 

Intel hasnt always been using foul play to crush AMDs market share, but given the hype about ZEN, and Intel KNOWS how ZEN performs by now. Ofcourse they do. So they KNOW whether it would be a hurdle to them or not. If Intel decided to preemptively push to keep ZENs traction lower then what could possibly be, then that would be serious.

 

also, to those of you who are just shitposting.

AMD compared FX (piledriver/bulldozer) with Intels offerings.

They admitt openly to having a WORSE product, what they contest is the gap between Intels offerings in some tests versus others. And i can agree with that possibly being the case.

Look at gaming benchmarks, Intel is on average 30% faster in gaming then AMD (8350 vs whatever intel has that scores higher). Yet some synthetic tests show intel as much as 70% faster. Odd dont you think?

 

i did the math based on 12 different gaming benchmarks. Difference in FPS output is roughly 27.8% higher score for intel then AMD. (this includes 5960x scores vs 8350 scores. The whole intel lineup, not just i5s).

So why can one test show 70% difference, whilst the average of 12 other tests show an puzzling 27.8% difference? (only ONE test had AMD score higher by a few FPS. They lost ALL other tests)

 

Sure, intel crushed AMD in minimum fps in some games, and in others, they didnt really do that much better if at all. So what gives? Why the fuck aint intel scoring better average FPS in games if they are THAT much faster?

 

or, maybe the synthetic benchmarks IS rigged... After all, it IS cheaper and less risky in terms of being caught, to pay off a single company, then it is to try pay off multiple game developers and publishers.

[ivan134]

Marketing. Quite a few people buy Skylake over Haswell because of DDR4 as way to "futureproof" their PC builds. We all know it's BS but it works.

If you really want to get specific, lower power draw and faster data transfer rates.

Its not just marketing. Digital Foundry has done a few videos on this and has shown that ddr4 does increase performance significantly in some games. It also scales with higher speed ddr4. Ddr4 would be more than just a marketing tool for Zen.

On Zen's peformance, iirc, the performance increase is from architecture changes only and doesn't take node shrink into acct. When you add 14 nm in the equation, I dont see why it wouldnt beat haswell.

While i dont see them beating skylake, i can think of a way for them to compete. People forget that iGPU adds to the cost of intel CPUs. AMD has already said that Zen CPUs will go all the way up to 8 cores. Something they could is offer a 6 core overclockable CPU for the same price as a K series intel cpu. 6 core overclockable cpu with a slightly worse IPC than a k series quad core for the same price would be attractive to a lot people.

[Paragon_X]

First of all I am brand agnostic I always go for whatever is best for me.

Getting that out of the way, why people are so adamant to the fact the companies lie, Intel has lied before AMD has lied before (remember on the presentation of the Fury X that they boasted it was an overclockers dream?) Nvidia lied as well, in general ALL companies have lied about one thing or two over the years (things we got to know about anyway)

My point is the most fragile company at the moment is AMD, and they are betting all in, on the Zen architecture as I see it. And if they know that a Benchmark is rigged (which is really possible) they are trying to eliminate that risk. They don't mention that their CPUs are superior, they only mention that the difference in performance is not true, which I tend to believe is true, in day to day workloads. 

People are calling  AMD a crybaby in many posts and are mentioning that AMD is doing this in order to give an excuse to the luck of performance of the upcoming Zen. Which might again be true, but how the hell do you know how the price to performance ratio and raw performance is for an unreleased product? Also in contrast with many people's speculations AMD provided 2 other benchmarks to support their claims (which of course would be nice if they could be verified by a third party).

Another interesting fact is that their clash with Sysmark is going even back to the Athlon years where AMD had the upper hand in the CPU market. So i do not believe its fair to call them liars at this point for a problem they have been mentioning for a decade. 

Only time will tell, with real world testing (games, editing software etc.) how the new products are going to perform.

[Sakkura]

I stand corrected, but Intel has a flimsy presence in the Android market - so, your argument doesn't stand

some manufacturers (ASUS if I recall) do some Android HW with Intel CPUs, but they're not great - I don't recall ever seeing Intel saying benchmarks weren't fair to them 

 

My point stands, because it applies to the PC space as well.

[Sakkura]

also, to those of you who are just shitposting.

AMD compared FX (piledriver/bulldozer) with Intels offerings.

They admitt openly to having a WORSE product, what they contest is the gap between Intels offerings in some tests versus others. And i can agree with that possibly being the case.

Look at gaming benchmarks, Intel is on average 30% faster in gaming then AMD (8350 vs whatever intel has that scores higher). Yet some synthetic tests show intel as much as 70% faster. Odd dont you think?

 

i did the math based on 12 different gaming benchmarks. Difference in FPS output is roughly 27.8% higher score for intel then AMD. (this includes 5960x scores vs 8350 scores. The whole intel lineup, not just i5s).

So why can one test show 70% difference, whilst the average of 12 other tests show an puzzling 27.8% difference? (only ONE test had AMD score higher by a few FPS. They lost ALL other tests)

 

Have you heard about GPU bottlenecking?

[zMeul]

My point stands, because it applies to the PC space as well.

what the hell are you talking about, nVidia's Tegra is not a x86 chip

on PC market Intel and nVidia had nothing overlapping, absolutely nothing

by 2011, nVidia stopped providing nForce chipsets for Intel CPUs

their latest chipsets were the GeForce 9xxx for LGA775 and notebook variants - 2008

[Sakkura]

what the hell are you talking about, nVidia's Tegra is not a x86 chip

on PC market Intel and nVidia had nothing overlapping, absolutely nothing

by 2011, nVidia stopped providing nForce chipsets for Intel CPUs

their latest chipsets were the GeForce 9xxx for LGA775 and notebook variants - 2008

 

You may not have heard about it, but Nvidia makes chips for PCs too.

[zMeul]

You may not have heard about it, but Nvidia makes chips for PCs too.

please enlighten me

[Sakkura]

please enlighten me

 

They have things like Geforce, Quadro, and Tesla.

[Prysin]

Have you heard about GPU bottlenecking?

sure, there can be other bottlenecks such as GPU, memory (slower memory = lower minimum FPS), PCIe bus speeds (gen2 vs gen3... sure its only 2-3 FPS difference at max, but still), then there is background tasks, test methodology, temperatures and cooling solutions (throttling?).

but hey, here is the post. Go redo my math if you'd like to.

 

http://linustechtips.com/main/topic/520052-negatives-about-amd-cpu/?p=6920328

 

Either way, the scores doesnt lie. Aslong as ALL tests are using same GPU and same config, CPU bottlenecking is the one factor that is going to manifest itself.

 

yes the benchmarks i pulled these results from are OLD. but back then, we didnt have the 900 series, and AMDs 200 series didnt have nearly as good drivers as now. So the test is valid, for when it was done.

 

770SLI = Custom 970/reference 980

[Sakkura]

http://linustechtips.com/main/topic/520052-negatives-about-amd-cpu/?p=6920328

 

Either way, the scores doesnt lie. Aslong as ALL tests are using same GPU and same config, CPU bottlenecking is the one factor that is going to manifest itself.

 

Actually, using the same GPU is not enough, since there can still be partial GPU bottlenecking placing the CPUs closer together.

[zMeul]

They have things like Geforce, Quadro, and Tesla.

I'm sorry .. what?!

since when does Intel provide alternatives to those and on the PC market

if you are referring to Xeon Phi, that's quite a recent product, it's aimed at very specific applications and I'm fairly sure BAPCo doesn't even cover those

[Sakkura]

I'm sorry .. what?!

since when does Intel provide alternatives to those and on the PC market

if you are referring to Xeon Phi, that's quite a recent product, it's aimed at very specific applications and I'm quite sure BAPCo doesn't even cover those

 

Intel has integrated graphics that competes against dedicated graphics. In addition, their x86 cores compete against GPUs, in that GPGPU is encroaching on more compute tasks over time. Intel has an incentive to make graphics performance look less important in benchmarks.

[patrickjp93]

BUT, if we look at intels previous strategy. Then they have always come down on AMD the moment AMD reaches roughly 20-25% marketing share, and just by brute force either tanked prices, cheated in benchmarks or simply marketed so hard that AMD drops down to their "usual" 5-10% market share.

Intel hasnt always been using foul play to crush AMDs market share, but given the hype about ZEN, and Intel KNOWS how ZEN performs by now. Ofcourse they do. So they KNOW whether it would be a hurdle to them or not. If Intel decided to preemptively push to keep ZENs traction lower then what could possibly be, then that would be serious.

also, to those of you who are just shitposting.

AMD compared FX (piledriver/bulldozer) with Intels offerings.

They admitt openly to having a WORSE product, what they contest is the gap between Intels offerings in some tests versus others. And i can agree with that possibly being the case.

Look at gaming benchmarks, Intel is on average 30% faster in gaming then AMD (8350 vs whatever intel has that scores higher). Yet some synthetic tests show intel as much as 70% faster. Odd dont you think?

i did the math based on 12 different gaming benchmarks. Difference in FPS output is roughly 27.8% higher score for intel then AMD. (this includes 5960x scores vs 8350 scores. The whole intel lineup, not just i5s).

So why can one test show 70% difference, whilst the average of 12 other tests show an puzzling 27.8% difference? (only ONE test had AMD score higher by a few FPS. They lost ALL other tests)

Sure, intel crushed AMD in minimum fps in some games, and in others, they didnt really do that much better if at all. So what gives? Why the fuck aint intel scoring better average FPS in games if they are THAT much faster?

or, maybe the synthetic benchmarks IS rigged... After all, it IS cheaper and less risky in terms of being caught, to pay off a single company, then it is to try pay off multiple game developers and publishers.

Because the FX 8350 has only 2 ALUs for every 4 per Haswell core, and because it lacks newer instructions. Software should be keeping up with new instructions. AMD can get bent or prove it. Also, games are GPU-bound. That's not a valid way to test CPU power.

[zMeul]

Intel has integrated graphics that competes against dedicated graphics.

no, they do not

Intel's iGP had barely any notable performance compared to a dGPU; only recently with Iris PRO iGPs Intel started to get serious - not 2011

as for compute tasks, Intel's offering is based on x86 architecture, nVidia's aren't; they don't even use the same software platform

and, as I said, BAPCo doesn't even cover this market

---

as for who's screwing who ..

do you recall Intel's Larrabee? nVidia fckd over Intel by retracting their licences, killin Larrabee in 2010

what did Intel do? complain? made YT videos about it? no, they started working on co-processors; they recently bought Altera and with that purchase they will put the screws to nVidia - not by "cheating" in benchmarks

[Sakkura]

no, they do not

Intel's iGP had barely any notable performance compared to a dGPU; only recently with Iris PRO iGPs Intel started to get serious - not 2011

as for compute tasks, Intel's offering is based on x86 architecture, nVidia's aren't; they don't even use the same software platform

and, as I said, BAPCo doesn't even cover this market

 

Yes they do. And Intel was indeed far behind in 2011, giving them that much more incentive to minimize the impact of GPU performance in benchmarks. BAPCo does cover this market.

[zMeul]

Yes they do. And Intel was indeed far behind in 2011, giving them that much more incentive to minimize the impact of GPU performance in benchmarks. BAPCo does cover this market.

Intel doesn't offer the same CPU with and without the iGP, so that comparison doesn't exist

it makes absolutely no sense to put a dGPU on a system that has no need for it

and Intel doesn't need to lie, they already have the GPU market by the balls with over 70% share

[Notional]

no, they do not

Intel's iGP had barely any notable performance compared to a dGPU; only recently with Iris PRO iGPs Intel started to get serious - not 2011

as for compute tasks, Intel's offering is based on x86 architecture, nVidia's aren't; they don't even use the same software platform

and, as I said, BAPCo doesn't even cover this market

---

as for who's screwing who ..

do you recall Intel's Larrabee? nVidia fckd over Intel by retracting their licences, killin Larrabee in 2010

what did Intel do? complain? made YT videos about it? no, they started working on co-processors; they recently bought Altera and with that purchase they will put the screws to nVidia - not by "cheating" in benchmarks

Intel integrating graphics onto their cpu singlehandedly killed Nvidia's low end graphics on all laptops and almost all desktops. People seem to forget that as apu's are such a commodity today.

[Sakkura]

Intel doesn't offer the same CPU with and without the iGP, so that comparison doesn't exist

it makes absolutely no sense to put a dGPU on a system that has no need for it

and Intel doesn't need to lie, they already have the GPU market by the balls with over 70% share

 

The comparison is between integrated GPUs from Intel, and dedicated GPUs from Nvidia and AMD (as well as integrated GPUs from the latter).

 

Intel did have an incentive to lie about GPU performance in 2011, even more so than they do today. They were FAR behind in 2011, and are still not in a good position today. The marketshare is artificial, it's just based on their tight grip on the CPU market. Very few people actually pick an Intel CPU because of the integrated GPU.

[Prysin]

Because the FX 8350 has only 2 ALUs for every 4 per Haswell core, and because it lacks newer instructions. Software should be keeping up with new instructions. AMD can get bent or prove it. Also, games are GPU-bound. That's not a valid way to test CPU power.

i7 4790k supports: https://en.wikipedia.org/wiki/Haswell_%28microarchitecture%29

 

Instructions MMX, AES-NI, CLMUL, FMA3 Extensions

• x86-64, Intel 64
• SSE, SSE2, SSE3, SSSE3, SSE4, SSE4.1, SSE4.2
• AVX, AVX2, TXT, TSX(Xeon E5 and Xeon E7 only)
• VT-x, VT-d

 

FX 8350 supports: https://en.wikipedia.org/wiki/AMD_FX

 

Instruction set AMD64/x86-64, MMX(+), SSE1, 2, 3, 3s, 4.1, 4.2, 4a, AES, CLMUL, AVX, XOP, FMA3, FMA4, CVT16/F16C, BMI1, ABM, TBM

 

 

AFAIK, in day to day tasks. normal, mundane day to day tasks, intel doesnt hold a huge advantage over FX in regards to instruction sets and their respective extensions. I could be wrong, but most commonly used instructions/extensions are x86/x86-64 (AMD64), AVX, SSE,MMX, AES and FMA3. The others i havent heard mentioned in any programming tutorials ive read or seen on youtube. Sure i dont have your education, so you can prolly prove me wrong here. But since its not mentioned, i take it the others arent THAT popular or known atleast in basic programming anymore...

 

i have no issues seeing that intel would have notable advantages in specialized tasks. However, i also recognize and agree that benchmarks SHOULD focus more on realistic, day to day "average joe" types of tasks.

 

I mean, go ask some random middle aged man or woman what they use to "validate" how fast a computer is. The majority would likely say "start up time" or "time to open tasks"....

The average person doesnt care if their CPU can calculate the weigth an elephant exerts on the ground with every footstep while it stampedes. The average person just want to know "would it hurt? yes or no".......

 

it is fine that such things ARE tested and validated, however putting emphasis on useless functions for marketing purposes are simply wrong.

 

let me ask you patrick. If you buy a harddrive or SSD, would you rather want to know how much data it can hold or how reflective the metal shroud is compared to the competitor?. This is what we are possibly seeing here. But until we have proof, it is all but "ifs and buts".

But unless someone dares to call Intel/BAPCo out on it, nobody ever will bother to dispute the results.

[Jahramika]

Ya so in the end all company's lie to make there product look better. Thief's that loose in court lie about others products/fake tests to make there's look than they are.

[Fetzie]

Because the FX 8350 has only 2 ALUs for every 4 per Haswell core, and because it lacks newer instructions. Software should be keeping up with new instructions. AMD can get bent or prove it. Also, games are GPU-bound. That's not a valid way to test CPU power.

AFAIK, in day to day tasks. normal, mundane day to day tasks, intel doesnt hold a huge advantage over FX in regards to instruction sets and their respective extensions.

My parents both have laptops with pretty much identical specs, but my Mum has an AMD A10 processor and my Dad has an Haswell-gen i5 (can't remember the model). I've used both, and in every-day applications (browsing, word processing, watching videos etc) they are indistinguishable. You simply cannot tell the difference.

[Delicieuxz]

People questioning AMD's damning, demonstrated and data-backed arguments... what are you doing? Is your objectivity switched not flipped in the 'On' position?