AMD Ryzen (Summit Ridge) Benchmarks Thread (use new thread)

[Doom2pro]

We don't know if this is real Zen ES benchmarks, if they are we don't know which SKU (Could be 4c/8t) and we pretty much know Turbo is disabled on these as they aren't finished validating that part yet.

Tip for Future leakers: SHOW US CPU-Z AT LEAST!

 

[krumme]

It could be a 6 core or a toaster. People fake this stuff all the time.

I didn't see one reliable leak from these Chinese forums when the RX480 was about to launch.

Yep. Wasnt it the same guy with the 1.7GHz rx480.
Perhaps this is just a bunch of people working together earning money on making fake news.
Its sad all the clicks it generates vs how much work is put into many AT articles.

 

[Doom2pro]

Yep. Wasnt it the same guy with the 1.7GHz rx480.
Perhaps this is just a bunch of people working together earning money on making fake news.
Its sad all the clicks it generates vs how much work is put into many AT articles.

Sure looks dubious to me, for someone claiming to make an effort to leak benchmarks they sure went out of their way to show the least amount of details. If I had my hands on some Zen Hardware I would have run a full suite of benchmarks, CPU-Z screenshots, hardware info, driver versions, system pictures (if possible), the whole enchilada.

 

[Asterox]

Isn't Cinebench biased against AMD? I remember there being a big stink about it, and Agner being rather miffed. http://www.agner.org/optimize/blog/read.php?i=49#49. Sure, Intel can't require benchmarks to have a biased cpuid check, but that doesn't stop 'em pushing for it.

We need open source benchmarks.

Yes it is in the past and even today, Powered by lobbying-bribery Intel compiler.

 

[TheELF]

We don't know if this is real Zen ES benchmarks, if they are we don't know which SKU (Could be 4c/8t) and we pretty much know Turbo is disabled on these as they aren't finished validating that part yet.

Tip for Future leakers: SHOW US CPU-Z AT LEAST!

We only know that they didn't use it in the presentation and gave this explanation.We do not know if it is generally disabled or if they just chose not to use it.

 

[TheELF]

Yes it is in the past and even today, Powered by lobbying-bribery Intel compiler.

Well pretty much everything is powered by lobbying-bribery Intel compiler even amd uses it for it's drivers.

 

[Doom2pro]

We only know that they didn't use it in the presentation and gave this explanation.We do not know if it is generally disabled or if they just chose not to use it.

Possible, but considering they did disable it, it may not even be performing correctly even if it was enabled, so it's suspect right off the bat. Then again the entire "leak" is suspect so I guess it doesn't matter either way.

 

[.vodka]

Tip for Future leakers: SHOW US CPU-Z AT LEAST!

So close, yet so far. Can't even leak properly.

-----------------------------


On another note,

http://fotoforensics.com/analysis.php?id=ffb9ca9a264eeef6b28d05e0ec4df84c8a9c6688.45097
http://www.izitru.com/Z978S

http://fotoforensics.com/analysis.php?id=ad445d8b6bb73a7f53ecb816db8a0f82c9228dc0.44744
http://www.izitru.com/fcOUj

There's so much JPEG compression/destruction here that you couldn't even tell if this is 'shopped. Everything in the ELA processed picture looks edited. I don't know about you guys, the 6s in the fritz chess picture seem too fuzzy for my liking. Even with a potato quality camera that has focusing problems at a range this close, you're taking a picture of a LCD screen, with the pixels and subpixels matrix up close. It's gonna be sharp... unless it's using windows scaling and the app doesn't play nice with it. I don't know. It's a lousy job.

Just so you know I managed to find the original source, it's a Baidu user and he says he will post more benchmarks in a few days. I will ask him to include a CPU-Z screenshot.

Well, let's see if the leaker can deliver properly.

 

[krumme]

Sure looks dubious to me, for someone claiming to make an effort to leak benchmarks they sure went out of their way to show the least amount of details. If I had my hands on some Zen Hardware I would have run a full suit of benchmarks, CPU-Z screenshots, hardware info, driver versions, system pictures (if possible), the whole enchilada.

Its understandable its limited what we get if this was actually a real leak.
But as i can tell we got excactly nothing. Is that correct?
At the same time i can observe agressive linking from here and H and also eg. observe google ranking keywords up on page 1 to this crapsite and the article in question.

 

[The Stilt]

There is no need to Photoshop the Cinebench screenshots, since the results can be forged without any third party tools (fire and forget).

 

[Doom2pro]

There is no need to Photoshop the Cinebench screenshots, since the results can be forged without any third party tools (fire and forget).

Not only that but I'm sure video game cheat utilities (Memory editors, similar to Game Genie) can easily do the same thing.

 

[The Stilt]

Not only that but I'm sure video game cheat utilities (Memory editors, similar to Game Genie) can easily do the same thing.

Sure they can, but for Cinebenches you don't need one.

 

[vissarix]

LOL

https://twitter.com/BitsAndChipsEng/status/810862715233255424

LMAO!!!

yeah definitely a 4core 8 threads zen smashing an i7 7700k oc at 5ghz??

can we get back to reality please?

 

[CatMerc]

LMAO!!!

yeah definitely a 4core 8 threads zen smashing an i7 7700k oc at 5ghz??

can we get back to reality please?

He said that as a joke. It's his way of saying that the rumor is bull.

 

[vissarix]

He said that as a joke. It's his way of saying that the rumor is bull.

look at how many people upvoted him, they took it seriously...

 

[Doom2pro]

look at how many people upvoted him, they took it seriously...

Or liked the joke?

 

[krumme]

LMAO!!!

yeah definitely a 4core 8 threads zen smashing an i7 7700k oc at 5ghz??

can we get back to reality please?

Ehh... Perhaps you might want to read it again. The twitter post was exactly a call for that - read the comments below.

 

[Abwx]

EDIT: Am I slow or what? [/QUOTE]

Sure looks dubious to me, for someone claiming to make an effort to leak benchmarks they sure went out of their way to show the least amount of details. If I had my hands on some Zen Hardware I would have run a full suite of benchmarks, CPU-Z screenshots, hardware info, driver versions, system pictures (if possible), the whole enchilada.

Dunno but in Fritzchess the FX ST score is 1982 at 3.4GHz..

Given the soft scaling this would imply that IPC + SMT gain are barely 24% in respect of a Piledriver core, and even less than this compared to a Steamroller or XV core..

http://www.hardware.fr/getgraphimg.php?id=385&n=1

Besides HW has 40.3% advantage in ST compared to Piledriver..

 

[prtskg]

I still find it quite an improvement over bulldozer. it was never going to reach BW level IPC in single generation. Hopefully by 3rd iteration Zen derivatives will be much nearer to the Intel cores. The thing is, some people will always promote worst possible case for AMD, some will show best possible and truth will lie somewhere in between.

 

[krumme]

Dunno but in Fritzchess the FX ST score is 1982 at 3.4GHz..

Given the soft scaling this would imply that IPC + SMT gain are barely 24% in respect of a Piledriver core, and even less than this compared to a Steamroller or XV core..

http://www.hardware.fr/getgraphimg.php?id=385&n=1

Besides HW has 40.3% adavantage in ST compared to Piledriver..

Thought the same. It gives zero sense vs piledriver score. Like the branchpredictor is even worse in zen vs pd. Yet excactly the branch predictor is hyped by amd as the most important improvement. As i wrote yesterday its on top of their list if we look at their own slides.

 

[DrMrLordX]

2 Zeppelin in MCM in server sku == snowy owl ( upto 16 cores, 4 memory channels)

I thought Snowy Owl was the server APU part? Though technically what you are posting does not contradict that . . .

BTW it is my understanding that CB15 and 11.5 both have problems working with CMT chips. I don't recall Stars chips having that much trouble with it, at least not compared to contemporaries.

 

[Abwx]

Thought the same. It gives zero sense vs piledriver score. Like the branchpredictor is even worse in zen vs pd. Yet excactly the branch predictor is hyped by amd as the most important improvement. As i wrote yesterday its on top of their list if we look at their own slides.

Even with the same branch predictor the numbers are doubtfull to say the least.

At 3.4GHz and a software 0.9x scaling 8 non CMTed Piledriver cores would score 1982 x 8 x 0.9 = 14270 knodes/s.

That would imply that in respect of Piledriver a Zen core throughput would be only 24% higher in an INT workload, despite 2x the ALU count and 2 threads loading the core, not only that would put Zen IPC at single digit distance of XV IPC, but moreover there would be no way that Zen could match BDW in Handbrake, let alone outmatch it like in AMD s demo...

 

[dark zero]

I still find it quite an improvement over bulldozer. it was never going to reach BW level IPC in single generation. Hopefully by 3rd iteration Zen derivatives will be much nearer to the Intel cores. The thing is, some people will always promote worst possible case for AMD, some will show best possible and truth will lie somewhere in between.

Actually the improvements went FAR better than expecting. At least they are back on the game. They won't won billions, but they could win enough to pay their debts.

 

[krumme]

I still find it quite an improvement over bulldozer. it was never going to reach BW level IPC in single generation. Hopefully by 3rd iteration Zen derivatives will be much nearer to the Intel cores. The thing is, some people will always promote worst possible case for AMD, some will show best possible and truth will lie somewhere in between.

You still treat this as if it was legit. And we got excactly no information just indicating that. On the contrary the source as i can tell leaked the 1.7GHz rx 480. So be carefull.

What we got is some blender and handbrake score at bwe 8c level and a base freq at 3.4 or up. While that says only a bit about performance i can point at a lot of this board who wouldnt expect that to happen if you asked us beforehand. My self included. And i still find base 3.4 or up a bit crazy for 8c on gf 14lpp. But it stands as a fact.

It might be that zen branch predictor tanks and that explains the results but today i wouldnt bet on it. Thats for sure. I dont know the technical reasons but surely Lisa su looks like she have found a goldmine.

 

[TheELF]

Yet excactly the branch predictor is hyped by amd as the most important improvement. As i wrote yesterday its on top of their list if we look at their own slides.

They also said that it will work together with their "neuro network" and that that will have to learn to give the best improvements...maybe they will have a driver for the CPU in the future with better performance,they already said that oc will only be possible through windows so maybe it will be a package.