AMD accuses BAPCo and Intel of cheating with Sysmark benchmarks

[patrickjp93]

AMD has only given some approximate numbers about IPC. We don't know what clocks Zen will run at yet. The problem with Intel architectures after Sandy Bridge is that they've struggled to reach the same (over)clocks, which means the improvement in IPC has been partially counteracted by lower clocks.

In addition, people saying they'll move to Zen but not Skylake also has to do with pricing, as well as perceived performance and maybe even brand preference. If a Zen quad core is significantly cheaper than the currently $250+ Core i5-6600K, well then it could be a more attractive upgrade even if performance is slightly lower.

Bottom line, we don't know yet. We don't know the price, and we only have vague ideas about the performance.

We do know, and 14nm FF is not going to maintain the same clocks that 32nm Bulk did. At best they'll be in the same range. I predict low-mid 3GHz for the 8-cores.

We have a very solid idea about performance. It will be 3% over Haswell at the absolute best per-clock.

[Sakkura]

We do know, and 14nm FF is not going to maintain the same clocks that 32nm Bulk did. At best they'll be in the same range. I predict low-mid 3GHz for the 8-cores.

We have a very solid idea about performance. It will be 3% over Haswell at the absolute best per-clock.

 

Please show all the benchmarks then. The ones not from AMD or WCCF.

 

Real talk, we don't know jack. We can speculate.

[Tomsen]

No they haven't. Cinebench was a matter of Maxon using a compiler that didn't optimize for AMD's architecture. That said, Intel's compiler still doesn't optimize for anyone but Intel, and you don't see a fresh lawsuit over that. Maxon screwed that up, not Intel. The judge was incompetent. As for Pentium 4, that was under Otellini more than a decade ago. He and Richard Huddy are both gone. Intel is innocent until proven guilty and AMD is full of crap until it puts its money where it's mouth is or provides proof I can't debunk in my sleep.

Yes they have. Look into the courtcase. It is not regarding Intel not optimizing their compiler for AMDs or VIA architecture, rather how open they were about it. You might be the only one calling that judge imcompentent in that exact case.. 

Intel is always innocent according to you, even so if they were proven to be guilty.

 

We have a very solid idea about performance. It will be 3% over Haswell at the absolute best per-clock.

You sure?  :rolleyes:

[ivan134]

Yes they have. Look into the courtcase. It is not regarding Intel not optimizing their compiler for AMDs or VIA architecture, rather how open they were about it. You might be the only one calling that judge imcompentent in that exact case.. 

Intel is always innocent according to you, even so if they were proven to be guilty.

 

You sure?  :rolleyes:

Lmao, I love your sig.

[patrickjp93]

Please show all the benchmarks then. The ones not from AMD or WCCF.

Real talk, we don't know jack. We can speculate.

I provided the bench results a while back. I only used open source benchmarks including the Linpack Suite.

[patrickjp93]

Yes they have. Look into the courtcase. It is not regarding Intel not optimizing their compiler for AMDs or VIA architecture, rather how open they were about it. You might be the only one calling that judge imcompentent in that exact case..

Intel is always innocent according to you, even so if they were proven to be guilty.

You sure? :rolleyes:

No. Intel was guilty in paying off Dell to not sell AMD's parts, but the judge was incompetent. The warning about ICC not optimizing for other architectures has been in the manual since Intel bought the startup that made the base of it. Maxon chose to use it. That lawsuit and the decision were both B.S..

Yes. I ran the tests myself, and the INteger Linear Program algorithm is the same one you'd find in any algorithms textbook. The benchmarks were all open source and included the Linpack suite. The absolute best AMD will do is 3% over Haswell per clock assuming their cache system is equally as good as what's on Haswell.

[Tomsen]

No. Intel was guilty in paying off Dell to not sell AMD's parts, but the judge was incompetent. The warning about ICC not optimizing for other architectures has been in the manual since Intel bought the startup that made the base of it. Maxon chose to use it. That lawsuit and the decision were both B.S..

Yes. I ran the tests myself, and the INteger Linear Program algorithm is the same one you'd find in any algorithms textbook. The benchmarks were all open source and included the Linpack suite. The absolute best AMD will do is 3% over Haswell per clock assuming their cache system is equally as good as what's on Haswell.

They were in the wrong. It was a grey area that, where Intel took the risk, and got a slap on the wrist.

In the same sense, it would be if AMD release a compiler, did a vendor ID check, disabled AMD64 on other x86 vendor chips, then start pushing it to benchmark tools. They would initially deny all the allegation, in how they would simply disable the entire instruction set extension.

Yeah, like Dell accepted Intels deal, doesn't give Intel a easy out. Lets not be foolish, and maxon should also be in the blame.

I'm much more concerned with how you think you can calculate a best-per-clock with no reliable data. AMDs claim 40%. Is this the maximum, average?? How did they get the numbers they did? What test/stimulations did they run? We have no data, beside amount of units. We known the amount of ALU/AGU, size and amount of SIMD/FPU units. That however, dont tell much regarding performance. Sure, it would be easy to calculate zen theoretical SIMD throughput.

[patrickjp93]

They were in the wrong. It was a grey area that, where Intel took the risk, and got a slap on the wrist.

In the same sense, it would be if AMD release a compiler, did a vendor ID check, disabled AMD64 on other x86 vendor chips, then start pushing it to benchmark tools. They would initially deny all the allegation, in how they would simply disable the entire instruction set extension.

Yeah, like Dell accepted Intels deal, doesn't give Intel a easy out. Lets not be foolish, and maxon should also be in the blame.

I'm much more concerned with how you think you can calculate a best-per-clock with no reliable data. AMDs claim 40%. Is this the maximum, average?? How did they get the numbers they did? What test/stimulations did they run? We have no data, beside amount of units. We known the amount of ALU/AGU, size and amount of SIMD/FPU units. That however, dont tell much regarding performance. Sure, it would be easy to calculate zen theoretical SIMD throughput.

Intel didn't push Maxon to use it. Go ahead and look back over the case. Maxon used it of their own recognizance. Intel made it perfectly clear in the documentation from the word go. That decision was BS, but it was bad press for Intel. Otellini and that entire C-Suite are gone. Intel is not the same company it was in 2004. Jesus people you ask to give Lisa Su a chance even though she herself cancelled the Beema tablets but you think Kirzanich is pulling the same illegal crap just because the company label is Intel? Get bent.

I have reliable data. I have all the exact clock latencies for Excavator courtesy of Agner Fog's documentation and AMD's x86 instruction manual. Assuming the 40% IPC gain is right (and they will fall short as they always do; let's be honest here), then the absolute best they could do is 3% better than Haswell. An integer linear program is a proven optimizing program. I can be 100% confident in my claims because the data and methods are solid. This is the difference between a scientist and a charlatan. I can actually do the work.

Nope. We know the counts, the width, the IPC gain claim, and much more. We also know exactly which instructions are supported. After that it's merely a matter of data input to QEMU to get a virtual CPU and running benchmarks on a virtual Windows 7 instance for the bulk of benches and Ubuntu 14.04 instance for Linpack.

Honestly it's not difficult, just tedious to automate.

[Sakkura]

I provided the bench results a while back. I only used open source benchmarks including the Linpack Suite.

 

So what you're essentially saying is you're a straight-up liar.

[patrickjp93]

So what you're essentially saying is you're a straight-up liar.

No, but that work is now six months old and I'm on winter tour in Florida. It's posted here on LTT, but I'm not digging through my archives on a cell phone.

[Dabombinable]

So what you're essentially saying is you're a straight-up liar.

Wouldn't be the first time he's essentially said that.

[Jahramika]

So what you're essentially saying is you're a straight-up liar.

That dude is a jackass lives on here defending Intel 24/7 proboly works for Intel.

[Dabombinable]

That dude is a jackass lives on here defending Intel 24/7 proboly works for Intel.

Look, AMD deserves all the shit they've been getting because of the FX modules are 2x cores BS. Just like Intel still deserves shit over fucking Netburst (tried my old Pentium 4 631 yesterday, and I'm sure that my 3.2GHz Northwood P4 HT from 2003 isn't that slow).

[patrickjp93]

That dude is a jackass lives on here defending Intel 24/7 proboly works for Intel.

Sure is easy slandering someone on the other side of a keyboard. I keep the facts straight, nothing more nor less.

[ivan134]

Sure is easy slandering someone on the other side of a keyboard. I keep the facts straight, nothing more nor less.

 

How much longer until Jim Keller announcement?

[Belgarathian]

There may be some merit in their accusations. This wouldn't be the first time Intel has done something like this.

 

Didn't Intel pull some shit by doing some shady dealings with sole supply agreements, anti-competitive behavior by paying off developers to optimize only for Intel, and to prevent access for AMD to certain vendors? 

[Briggsy]

Well, I am not mistaken as I was just responding to the scenario where Zen was only just as good as Haswell in performance. I do doubt that would ever be the case, otherwise I don't know why AMD would bother. If they release IPC levels of Haswell with 6-8 cores and they are priced competitively then it will obviously be a great addition to the market and will make a lot of people consider AMD over Intel. 

 

AMD's focus in the past year has been to change public opinion about being the cheaper brand, or "pc parts for poor people." If they release Zen and it matches Intel performance, I would expect AMD to match Intel pricing for enthusiast/enterprise tiers, and slightly undercut Intel at the lower end - Although AMD still has a more robust iGPU, so there is room for charging more and getting away with it if CPU performance is on par with Intel. I'm fairly sure the days of AMD being the go-to cheaper brand are nearing their end, unless Zen is a flop.

[patrickjp93]

How much longer until Jim Keller announcement?

It's been on his LinkedIn for weeks. I posted this on Christmas.

Edit: apparently sarcasm tags are required for everything.

[patrickjp93]

Didn't Intel pull some shit by doing some shady dealings with sole supply agreements, anti-competitive behavior by paying off developers to optimize only for Intel, and to prevent access for AMD to certain vendors?

Yes, and all the leadership responsible for it were kicked out of the company. Paul Otellini had to leave over that. He now works for Google.

[pixeldensity]

It's been on his LinkedIn for weeks. I posted this on Christmas.

I can't seem to find his linkedin account nor anything that says he is currently working at intel. If he was at intel you would think that a news article would say so. Can you link his page, so we can see for ourselves?

[Demonking]

You are mistaken, Zen was stated to have IPC of around Haswell level or a little above that with 6 and 8 cored mainstream CPUs, not to be around Haswell level of performance. If they're priced aggressively, they would not be a disappointment but a valid alternative to what Intel has/will have to offer, I'm not sure if those statements are true, but that's what has been revealed so far.

How are they going to get haswell level 6 core & actual 6 core not that BS they did with FX & 8 core cpu priced below skylake without taking a lost, because the CEO said they're are tired of being know as the cheaper solution & AMD is not in a place to be taking losses. Now intel can which would put AMD in a bad spot. 

[ivan134]

How are they going to get haswell level 6 core & actual 6 core not that BS they did with FX & 8 core cpu priced below skylake without taking a lost, because the CEO said they're are tired of being know as the cheaper solution & AMD is not in a place to be taking losses. Now intel can which would put AMD in a bad spot. 

Because people seem to forget that iGPUs increase the cost of Intel's CPUs. Zen is coming in 2 flavors: APUs and just CPUs. A 6 core priced similarly or a little bit higher than Intel's quad core is not far-fetched.

[Demonking]

Because people seem to forget that iGPUs increase the cost of Intel's CPUs. Zen is coming in 2 flavors: APUs and just CPUs. A 6 core priced similarly or a little bit higher than Intel's quad core is not far-fetched.

& a 5820K is $389 USD without an igpu, so tell me how are they going to do that for less than the 5820K without taking a lost, because i don't see it. The last thing AMD needs to to be selling cpu's at a lost. Corners will have to be cut in order to get a 6 core haswell level IPC CPU  for less than a 4690K.

[ivan134]

& a 5820K is $389 USD without an igpu, so tell me how are they going to do that for less than the 5820K without taking a lost, because i don't see it. The last thing AMD needs to to be selling cpu's at a lost. Corners will have to be cut in order to get a 6 core haswell level IPC CPU  for less than a 4690K.

It's 6 cores and 12 threads. An i5 is 4 cores, 4 threads and an iGPU. I'm not saying I know the configurations for Zen CPUs, but a 6 core/6 thread(or whatever AMD decides to call it) and no iGPU for the same price is possible.

[Demonking]

It's 6 cores and 12 threads. An i5 is 4 cores, 4 threads and an iGPU. I'm not saying I know the configurations for Zen CPUs, but a 6 core/6 thread(or whatever AMD decides to call it) and no iGPU for the same price is possible.

it's has to be priced above the 4690K around the 4790K which is 339.99 USD maybe even above that reaching close to the 5820K that the only way i see it. but if they do it which i highly doubt good,but they are taking a lost on it which isn't good for them.