Nvidia/AMD/VIA leave Sysmark company Bapco

[jvroig]

IDC is arguing with himself now. This should be good.

I don't see why/how you would/could misconstrue it as such.

From reading this thread, all I got was that he merely presented both sides of the situation, then got called out for it, for not sticking to one company over the other (i.e., criticized for not acting like a fanboy, which seems to be a dearly-held tradition of some sort since all camps appear to be well-represented with various personalities here)

It is in no way different from how I would weigh both sides of the story myself:
1.) AMD/Nvidia are leaving supposedly because Intel is paying off BAPCo / benchmark is "unfair". There seems to be evidence that this is actually true, and have been happening for 10 years.
2.) On the other hand, the existence of a metric means there will be losers, and these losers will need to do "damage-control", whether it is justified (valid) or not.

In Intel's case, it is a bit more complex since it appears to be a mix of both justified and unjustified cases. They are fielding superior CPU products. The benchmark reflects that. This does not mean, however, that there is no problem with the benchmark. It could be that they are fielding superior CPU products, while at the same time "fixing" the benchmark to further exaggerate the difference. If this benchmark will suddenly be less-favorable for them due to other factors (e.g., GPU issue), then they most certainly will fight against it, at least until they are also fielding products that are similarly superior.

Whatever the scenario, the losers in this case (AMD) need to raise a stink about it, whether it is justified or not, which is why it is hard to read anything into this without being inundated with PR.

My opinion in the matter is rather simple: that fact that it is used as a standard government metric means it is probably being "fixed" by Intel as much as they can. If there is an indicator as important as that (basis for government contracts), I do not know any company who would not try to fix or game it, short of breaking the law, of course. It is simply an obvious business decision. This is why I am inclined to be more receptive to the "BAPCo is paid by Intel / is Intel / is controlled by Intel" issues. It just makes a whole lot of business sense, like "giving incentives" to Dell (and others), or making biased compilers.

It's just business as usual, for both parties here (Intel/BAPCo vs AMD/NV).

Now, if someone(AMD/NV/VIA/whoever) were to say "Sysmark is full of crap, so now we made a better benchmark" and that benchmark turns out to be really better [insert criteria for "better"-ment here], then that would be impressive and really newsworthy. Aside from that, this is still just PR at work, whether it is 100% justified (the "good" sort of PR) or not (the "bad" PR people are used to ignoring).

 

[Genx87]

Do you read? Tessellation was nVidia's strong point so they pushed it as hard as they can in synthetic benchmarks such as Stone Giant and Heaven. AMD made a big stink because of nVidia's marketing trying to sell their cards based on synthetic benchmarks rather than real world performance.

AMDs "tesselation done right" BS started months after the release of 400 series parts. I am reading just fine what you are saying. Which is AMD attacked tesselation performance(a cornerstone of DX11) and benchmarks that showed Nvidia doing better because Nvidia was 6 months late and 10% faster. And AMD did this months "after" the release of Fermi, right before the release of the 6800 and 6900 series. A pattern is emerging.

Like I said, you have an interesting theory. I other words, I am mocking the theory that AMD did pretty much the same thing they did then with tesselation, as they are doing now, because of your stated reasons. Nvidia late, hot, and barely faster. Clearly they did it because their new arch didnt fix an obvious deficiency in tesselation performance. Instead of fixing their problem, they attacked the messenger.

 

[podspi]

Companies are always going to protect their interests and play to their strengths. If this was just AMD leaving, I would say they were probably unhappy with the results BD showed on SysMark and were throwing FUD.

The fact that NV, VIA, AND AMD left suggests to me that something is going on in that organization that benefits Intel, and only Intel. When you look at the partner list they boast about, its mostly software and OEMs. No more chipmakers...

 

[Arkadrel]

AMDs "tesselation done right" BS started months after the release of 400 series parts. I am reading just fine what you are saying. Which is AMD attacked tesselation performance(a cornerstone of DX11) and benchmarks that showed Nvidia doing better because Nvidia was 6 months late and 10% faster. And AMD did this months "after" the release of Fermi, right before the release of the 6800 and 6900 series. A pattern is emerging.

Like I said, you have an interesting theory. In otherwords, pulling things right out of your ass.

That because nvidias tessellation performance is only really clear under extreme situations that DONT reflect real world use (in games ect).

They become appeart (as a nvidia strength) only when your "doing it wrong", which is why AMD did the "tessellation done right" thingy.

Basically Nvidia where useing sub pixel triangles that had no visable effects, to showcase they had stronger tessellation. Its a fact they do, its also a fact that there is no image quality gained from doing so (which means worse performance for no gain in image quality = stupid).

So nvidia are testing their tessellation unit and saying its faster, under situations that have nothing to do with the real world performance.

Your spinning things, or mis-interpriateing things because of fanboyisme.
Your Also derailing a thread, thats about Intel screwing Via/AMD/Nvidia over with a benchmark, that they own and are cheating in and have been for the last 10years or so (to make their cpus look even better (even if they already are)).

The ironical thing here is Intel are cheaters, without a need to be, because they already have the fastest cpus on the market.

 

[Nemesis 1]

Well there is Nights Corner release in 2012. I would imagine that it being x86 has something to do with all of this . From reading the PDF I posted in the Larrabbee not dead yet topic . Nights corner should be 3x faster than what shown in that PDF as the nights ferry that is being used for those benchies in that PDF is in fact larrabee renamed.

 

[Nemesis 1]

All zoners know that all benchmarks favor intel . Intel has bought them all.Zoners all know that in real world AMD is faster smoother and can leap tall buildings. Intel fanbois need a reality check .

 

[Topweasel]

Didn't I address both of these possible progenitors in my post?

You probably did. But combination of poor TLDR skimming ability and what kind of felt like brushing off of the AMD/Nvidia/VIA portion at the end didn't seem more like a counterpoint but more of a "just they could be right, but until they prove it to me they are just a bunch of whiners". I was merely trying to enforce the two obvious portions, real world tests matter more, and its not just AMD its three companies that compete with each other in different ways. That lends credence to me at least that it isn't just whining, but a legitimate complaint. You don't lose a third of your partners and all of the impactful component manufacturers outside Intel otherwise.

 

[Nemesis 1]

No its not a legit complaint . Befor this happened and befor the FTC reached agreement with Intel all 3 companies were complaining about intels compilers . Now that that issue was settled by the FTC . they are not happy . They didn't get what they wanted Intel won on the compiler issue bigtime .

 

[Nemesis 1]

That because nvidias tessellation performance is only really clear under extreme situations that DONT reflect real world use (in games ect).

They become appeart (as a nvidia strength) only when your "doing it wrong", which is why AMD did the "tessellation done right" thingy.

Basically Nvidia where useing sub pixel triangles that had no visable effects, to showcase they had stronger tessellation. Its a fact they do, its also a fact that there is no image quality gained from doing so (which means worse performance for no gain in image quality = stupid).

So nvidia are testing their tessellation unit and saying its faster, under situations that have nothing to do with the real world performance.

Your spinning things, or mis-interpriateing things because of fanboyisme.
Your Also derailing a thread, thats about Intel screwing Via/AMD/Nvidia over with a benchmark, that they own and are cheating in and have been for the last 10years or so (to make their cpus look even better (even if they already are)).

The ironical thing here is Intel are cheaters, without a need to be, because they already have the fastest cpus on the market.

If this infact an intel owned benchmark . Intel would be crazy not to optimize for it . But its still not going to stop people from using it . Same as GPU benchmarks that both NV and AMD use . They pay developers to optimize for their GPus . Both run the games fine but not equally .

 

[waffleironhead]

All zoners know that all benchmarks favor intel . Intel has bought them all.Zoners all know that in real world AMD is faster smoother and can leap tall buildings. Intel fanbois need a reality check .

is that nvzone or slizone?

 

[Genx87]

That because nvidias tessellation performance is only really clear under extreme situations that DONT reflect real world use (in games ect).

They become appeart (as a nvidia strength) only when your "doing it wrong", which is why AMD did the "tessellation done right" thingy.

Basically Nvidia where useing sub pixel triangles that had no visable effects, to showcase they had stronger tessellation. Its a fact they do, its also a fact that there is no image quality gained from doing so (which means worse performance for no gain in image quality = stupid).

So nvidia are testing their tessellation unit and saying its faster, under situations that have nothing to do with the real world performance.

Your spinning things, or mis-interpriateing things because of fanboyisme.
Your Also derailing a thread, thats about Intel screwing Via/AMD/Nvidia over with a benchmark, that they own and are cheating in and have been for the last 10years or so (to make their cpus look even better (even if they already are)).

The ironical thing here is Intel are cheaters, without a need to be, because they already have the fastest cpus on the market.

Real world performance based on what? Todays games? That is irrelevant comparing the tesselation performance between the two camps. Toss geometry at the two pieces of hardware. AMD loses to Nvidia and doesnt scale even between their own hardware.

And just like this situation now where they cry about benchmarks showing their inferior hardware in a bad light. They did the same last Fall. Buck up AMD, playing the victim card is getting old.

 

[Lonyo]

Real world performance based on what? Todays games? That is irrelevant comparing the tesselation performance between the two camps. Toss geometry at the two pieces of hardware. AMD loses to Nvidia and doesnt scale even between their own hardware.

And just like this situation now where they cry about benchmarks showing their inferior hardware in a bad light. They did the same last Fall. Buck up AMD, playing the victim card is getting old.

NV also quit. Are they playing the victim too?

 

[Topweasel]

Well there is Nights Corner release in 2012. I would imagine that it being x86 has something to do with all of this . From reading the PDF I posted in the Larrabbee not dead yet topic . Nights corner should be 3x faster than what shown in that PDF as the nights ferry that is being used for those benchies in that PDF is in fact larrabee renamed.

Now its about Nights Corner and not Haswell?

Its doubtful that AMD or Nvidia is worried in a benchmark about a product they have no real performance information on that isn't coming out for nearly 18 months.

What they are worried about is Sysmark is releasing a benchmark that won't get a real update for 4-5 years, that has no eye toward the shift in computing to how much work is offloaded on to the video cards and a scoring system that weights application and tests that have no relevance in today's uses. Its pretty simple, Anand himself he commented that it seems wierd to use and old Firefox and an old IE in testing for example when if they used a slightly newer version of both, work is GPU accelerated. So for the next 5 years they will already be using in two of the most used applications worldwide an already out of date software. Software that the new versions might not be the best examples of further GPU usage, but set the stage of what they feel the market is shifting to.

 

[Nemesis 1]

@post 67.

This is a differant subject and for your information. beside other things SB /IB that are common to both processors in the Word Bridge. A bridge to What? Haswell .

Did you read the pdf on nights ferry /nights corner. It wasn't tested with SB or IB it was tested with old tech. Given intels compilers and the fact this is about benchmarks and what the 3 muskateers are cring about . I would imagine that X86 has a lot to do with the cry in the wild . Why would NV be cring about this benchmark ?

If Haswell has ondie vector unit combine that with Nights corner. It should be something to look at as to whats going on here on this benchmark . SB introduces AVX IB continues . Haswell brings AVX II . I am simply connecting the dots . The one thing that ties all on the intel side of things is compilers .

My speculation is built on firm ground where as your swimming in the ocean being pulled out to sea by undercurrent were you will likely drown. You screamed and screamed for aid but no lifeguard came to your aid.

 

[lifeblood]

I don't know if AMD is just whining or rather they have a real gripe, I just know Sysmarks overall score doesn't represent my average use either at work for the government or at home. I don't use excel for giant spreadsheets, I use it for purchase orders. I don't compress or decompress files. At home I use Firefox 4 with GPU acceleration for watching YouTube and other videos. Sounds to me like AMD has a point, at least as it relates to me.

 

[formulav8]

The stupid tool says 2012. There is NO excuse not to include IE9 and firefox 4. Among other things. Its pathetic that the head sysmark guy is also on Intels payroll.

Via and Nvidia along with AMD left probably for good reason. Label them babys if you want but you can be sure Intel is the bully in his case

 

[Genx87]

NV also quit. Are they playing the victim too?

I am not sure why they were in the group. Their x86 CPU lineup, is, well, lacking

 

[Phynaz]

The stupid tool says 2012. There is NO excuse not to include IE9 and firefox 4.

It's been in development for two years. Adding relatively newly released software would delay its release that much longer.

 

[Lonyo]

I am not sure why they were in the group. Their x86 CPU lineup, is, well, lacking

So AMD quit because they are crybabies. NV quit because you're not sure why they were there in the first place.

 

[Topweasel]

I am not sure why they were in the group. Their x86 CPU lineup, is, well, lacking

Complaint seems to be lack of testing for GPU offset computing. Nvidia wants to see systems with larger video cards test better. But when you have very few apps that interact with a card and the ones they do poorly weighted so that a computer with a GMA 2000 ties a GTX580, but you add more cores and and the score doubles then there is an issue. Seems like it has nothing to do with BD on the AMD side but Llano and their Radeon lines.

Its legitimate to a degree when several highly used apps use graphics now (Flash, FF, IE, office) that there is a shift and should have some relevance in the suite. The one avenue intel isn't competitive in is the one that doesn't get measured is fishy.

 

[formulav8]

It's been in development for two years. Adding relatively newly released software would delay its release that much longer.

Then name it 2010 or 2009. 2012 implys all of the latest and greatest we currently have. Even have tech in the future not invented yet...

Blah to them I say

 

[Phynaz]

Then name it 2010 or 2009. 2012 implys all of the latest and greatest we currently have. Even have tech in the future not invented yet...

Blah to them I say

So now your issue is with its name? :\

 

[formulav8]

So now your issue is with its name? :\

Its name is Sysmark isn't it? 2012 is the 'verson'.

 

[GaiaHunter]

It's been in development for two years. Adding relatively newly released software would delay its release that much longer.

That is why it is falls into the "synthetic benchmark" category.

 

[Madcatatlas]

So AMD quit because they are crybabies. NV quit because you're not sure why they were there in the first place.

Yes Lonyo, apparantly this is some fast food type of thing where you can just comment left and right without an ounce of reflection ever entering the equation.