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

[The Stilt]

Can Zeppelin get as low as 45W TDP within that clockspeed range using an 8c/16t part?

Not even close, allegedly.

 

[inf64]

You don't have to be rocket scientist to know that ES clock speeds do not equal shipping clocks. Plus now with this (likely) CES launch AMD will have more time to get the clocks where they want them to be (which was hinted at the event to be 3+Ghz).

 

[cbn]

In those Zen related slides there s this one :

So Finfets used for Zen not only consume less but also have higher working frequency in respect of a 28nm that can only be the one used for Kaveri/Carrizo/Bristol Ridge.

Comparing to graphs below, yes it does look like the light blue dashed line does corresponds to the curve Excavator uses.

The question is whether AMD is using the 25W or 20W Excavator for comparison to Zen? I think it might be the 20W Excavator in the second graph. (notice the 20W Excavator crosses the same number of horizontal lines as the light blue dashed in the Zen vs. Excavator graph)

So maybe the point where the tachometer dial is on the Zen graph corresponds to 12.5W or so?

 

[sirmo]

You don't have to be rocket scientist to know that ES clock speeds do not equal shipping clocks. Plus now with this (likely) CES launch AMD will have more time to get the clocks where they want them to be (which was hinted at the event to be 3+Ghz).

Bulldozer engineering samples were 2.8Ghz.

 

[HurleyBird]

Aside from the grain of salt necessary for any vendor supplied benchmark, the fact that AMD is comparing Zen to a $1000 octo core Broadwell-E processor (favourably at that), and not a quad core Skylake, already places Zen far above my original expectations.

Assuming that Zen ends up at least somewhat competitive against Intel's best in reality as well, the big question is whether Intel and AMD will cooperate in keeping each other's margins high, or if we'll see a return to the competition of old and these $1000+ processors dropping to a more appropriate price. It would sure be nice if the answers were yes, and the later.

 

[swilli89]

That seems right :



http://techreport.com/review/30540/amd-gives-us-our-first-real-moment-of-zen

Zen is about 3% faster...

Edit : According to Techreport article :

WOW. Incredible, this is Intel's current high end and AMD is winning in a common benchmark. Very good sign and probably why AMD stock rallied so high today.

 

[The Stilt]

You don't have to be rocket scientist to know that ES clock speeds do not equal shipping clocks. Plus now with this (likely) CES launch AMD will have more time to get the clocks where they want them to be (which was hinted at the event to be 3+Ghz).

You don't have to be a rocket scientist to realize that the CPUs don't turn up over night exactly. I wouldn't hold my breath for any clock improvements over the current levels, since most likely the main limitation is the process itself. The CPUs which would be shipped in January will be manufactured in October - December time period. Not a terribly long time for the process to mature, when considering that the peak maturation of 32nm SHP SOI was achieved three years after the process was entered mass production. Meanwhile the GF28A took ~ 18 months to do a similar improvement.

I have several ES CPUs of final stepping, which have been manufactured 4-8 months prior their launch date. If AMD demonstrates performance benchmarks to the press, they definitely have the final silicon available. And all signs point to the same direction, to A0c revision seen in Zauba.

Bulldozer engineering samples were 2.8Ghz.

The facts still have not changed...

Except that's wasn't the case. The second major prototype stepping (OR-A1, ZD262046W8K43 SKU) already reached 3.6GHz, which is 400MHz lower than the fastest retail part (FX-8120) did at the same TDP.
A1 stepping of Bulldozer was taped out a year (September-October 2010) before the final product was launched. That's why the situation is different to Zeppelin, since obviously we are not talking about a product still year away (hopefully at last).

 

[Abwx]

Not even close, allegedly.

That s right, It should be more like 32-35W/2GHz rather than said 45W...

The question is whether AMD is using the 25W or 20W Excavator for comparison to Zen?

Dosent really matter, if we take a worst case figure with the Athlon X4 845 it use 46W/3.5GHz on Cinebench R15, given that according to AMD XV and Zen consume the same power a 3.5GHz Zen should be barely 90W and a 2GHz one at 32-35W or a little less given the steep slope of the finfets at lower frequencies.

 

[The Stilt]

Took me a while to realize something from what AMD stated:

How much does dual channel vs. quad channel hurt Broadwell-E? (i.e "comparable configuration", Zen has just dual channel memory).

 

[SAAA]

So this is competition, finally. I'm also glad that the 40% IPC uplift was over the latest arch. and not Piledriver, it will help mitigate lower clocks, if that happens at least.
I mean I'm not expecting 5GHz products like FX line at all, but a ~150W TDP 4GHz stock part at some point might not hurt.

Now just tell me Zen+ is another 10-15% and with process getting better it will be a good Skylake-X competitor too.

 

[SAAA]

How much does dual channel vs. quad channel hurt Broadwell-E? (i.e "comparable configuration", Zen has just dual channel memory).

I don't know but if it's even 15% with Zen being this close in a else identical comparison it doesn't look bad.
Who cares at that point if they give us 8-core Haswell-Ivy performance for less than Intel ever did? I don't see much depreciation on those parts... similar specs for even 70% of the price could do wonders to AMD high end sales and margins, unlike current huge 32nm cpus that are sold for less than competition quad cores.

 

[The Stilt]

I'd say it is better that we wait until we declare that AMD has finally catched Intel in IPC. Blender is open source and there is plenty you can do to prevent two different builds performing identically. Not saying that AMD did such thing, but so far nothing AMD has said (40% over Excavator) implies that they would get anywhere close (within ~20%) to Broadwell / Skylake IPC. If they really have, then that is one of the biggest achievements in the history of x86 CPUs. If not, it will be one of the biggest dissapointments in the history of x86 CPUs. Hero or zero, just watch your steps AMD dammit....

 

[plopke]

If it brings more affordable 4/8 cpu or making it the new midrange , no matter which side you buy , zen is a win for me .

 

[sirmo]

The facts still have not changed...

Except that's wasn't the case. The second major prototype stepping (OR-A1, ZD262046W8K43 SKU) already reached 3.6GHz, which is 400MHz lower than the fastest retail part (FX-8120) did at the same TDP.
A1 stepping of Bulldozer was taped out a year (September-October 2010) before the final product was launched. That's why the situation is different to Zeppelin, since obviously we are not talking about a product still year away (hopefully at last).

I am going by what was widely reported about BD before the launch, which was 2.8Ghz. 32nm was brand new at the time, while AMD already has products under its belt on 14nm FF. But I could be wrong. I personally don't think we can even begin to guess what Summit Ridge's final clocks will be.

 

[cbn]

They should do better than that for enthusiasts.

Looking at the Zen graph in this post it looks like the Zen Core at the point of the tachometer corresponds to around 12.5W.....but notice it is not leveling off at this point.

So maybe there is some decent room left in the process and design.

8 x 12.5W = 100W. Add in another 15W for the integrated chipset and maybe TDP would be 115W. Assuming extra room is indeed left then I think 125W+ could be possible for 8C/16T. With that mentioned, I am mostly concerned with the 4C/8T. Considering AMD already has nice copper cored 95W coolers available it would be good to at least see a cTDPup option (maybe 95W, if not then 80W) on the 65W SKUs.

 

[sirmo]

I'd say it is better that we wait until we declare that AMD has finally catched Intel in IPC. Blender is open source and there is plenty you can do to prevent two different builds performing identically. Not saying that AMD did such thing, but so far nothing AMD has said (40% over Excavator) implies that they would get anywhere close (within ~20%) to Broadwell / Skylake IPC. If they really have, then that is one of the biggest achievements in the history of x86 CPUs. If not, it will be one of the biggest dissapointments in the history of x86 CPUs. Hero or zero, just watch your steps AMD dammit....

I agree.. except don't think they can ever top Bulldozer when it comes to disappointments. Zen is looking solid so far, but we'll see.

 

[The Stilt]

8 x 12.5W = 100W. Add in another 15W for the integrated chipset and maybe TDP would be 115W. Assuming extra room is indeed left then I think 125W+ could be possible for 8C/16T. With that mentioned, I am mostly concerned with the 4C/8T. Considering AMD already has nice copper cored 95W coolers available it would be good to at least see a cTDPup option (maybe 95W, if not then 80W) on the 65W SKUs.

It doesn't work like that, even remotely.

Two Excavator CUs require > 45W (other SoC HW excluded) to sustain 3.4GHz during Cinebench R15, and that's not even the most stressful real world workload you can throw at it.
Not to mention that the Fmax vs. Vdd curve on 14nm LPP is vastly steeper than on 28nm HPP, close and beyond the optimal Fmax range. 14nm LPP is superior to 28nm HPP when it operates at it's optimal range, but anything outside this range is far inferior.

 

[powerrush]

All seems in the direction that AMD wants to get the most optimized platform for Zen , that is the reason of the engineering samples sent to the manufacturers. Zen will launch at 3.3ghz 3.6 turbo. Remember me.

 

[VirtualLarry]

There is no free lunch. Its 95w tdp and 8c. Dense and lean. It will probably be cheap then. I gladly take 8 cores with hsw like ipc at base 3ghz if it stays at 95w oc 125w tdp and i can get it instead of the usual 4 core i7 for the same cost of cpu plus mb. Not that i can use all the power but at least i dont have to shell out 1000euro for the cpu only to get there.
Cheer up !

Sounds good to me. Time to start saving up some cash... for Zen!

 

[Tuna-Fish]

You don't have to be a rocket scientist to realize that the CPUs don't turn up over night exactly. I wouldn't hold my breath for any clock improvements over the current levels, since most likely the main limitation is the process itself. The CPUs which would be shipped in January will be manufactured in October - December time period. Not a terribly long time for the process to mature, when considering that the peak maturation of 32nm SHP SOI was achieved three years after the process was entered mass production. Meanwhile the GF28A took ~ 18 months to do a similar improvement.

I have several ES CPUs of final stepping, which have been manufactured 4-8 months prior their launch date. If AMD demonstrates performance benchmarks to the press, they definitely have the final silicon available. And all signs point to the same direction, to A0c revision seen in Zauba.

Do you have any reason to believe that 2.8/3.2GHz are the top clocks the CPU can reach, and not just whatever AMD decided to set the ES chips at? Early K8 ES were all 800MHz chips. They could have clocked a lot higher, AMD just decided to not ship any high-clocked ES chips to vendors. The argument that final chips are better than ES is not that the CPUs improve between now and release, it's just that there is no actual need or reason to set ES chips to run as high as they can. If all ES chips in the wild have same clock rates, it strongly suggests that those clock rates are arbitrary.

On the other hand, if AMD could make a CPU faster than Intel, I doubt they'd have so many qualifiers in the benchmark. So I don't expect a huge difference, I just don't think that you can estimate how well an ES chip can clock from what clocks it ships at.

 

[rancherlee]

All seems in the direction that AMD wants to get the most optimized platform for Zen , that is the reason of the engineering samples sent to the manufacturers. Zen will launch at 3.3ghz 3.6 turbo. Remember me.

Seems plausible to me, if it launches at those clocks the 4c/8t should sandwich somewhere between the i5 and i7 based on lower clock speeds but +HT vs the i5 on well threaded games/applications and the 8c/16t should do well against the 6c/12t Intel chips. If they are priced at 200-250$ and 325-375$ they should do OK.

 

[itsmydamnation]

Took me a while to realize something from what AMD stated:



How much does dual channel vs. quad channel hurt Broadwell-E? (i.e "comparable configuration", Zen has just dual channel memory).

likely sweet FA, why? look at there server SKU's would Intel gimp there $2K plus cpu's with no enough memory bandwidth..... its not like those dies are pad limited. 2nd point thats the whole point of a cache system prefetchers/ predictors , bandwidth amplification and latency reduction. People Here on Anandtech constantly over exaggerate the need for memory bandwidth, no your workload didn't go faster because of more bandwidth it went faster because of lower latency. We have particularly seen this behavior at every DDR transition.

I'd say it is better that we wait until we declare that AMD has finally catched Intel in IPC. Blender is open source and there is plenty you can do to prevent two different builds performing identically. Not saying that AMD did such thing, but so far nothing AMD has said (40% over Excavator) implies that they would get anywhere close (within ~20%) to Broadwell / Skylake IPC. If they really have, then that is one of the biggest achievements in the history of x86 CPUs. If not, it will be one of the biggest dissapointments in the history of x86 CPUs. Hero or zero, just watch your steps AMD dammit....

Thats because you keep taking a number in a vacuum with no attempt to understand the uarch and then apply linear scaling. Remember its an Average of instruction throughput there will be workloads that get less perf uplift and workloads that get more. But what we can tell form all the information we already know is that in the area of integer, non SIMD FP , MMX/SSE/AVX1/2-128 Zen will be strong, We also know that in 256 AVX/AVX2 and 256FMA Zen wont be near Intel.

These slides help fill in a few holes but dont add anything massively new to what we already know. But its good to see a big L1-i with solid associativity, i dont think anyone guessed the asymmetric L1d/L1i sizings.

What i really want to know is does AMD have the rumored Crypto/compress SOC level engines, thats a massive advantage over intel for the big guys (facebook,google,etc) and hopefully support makes its way to enterprise web stacks in a reasonable time frame.
Will they expose the 10gb interfaces on the consumer product? that could be a massive value add.

 

[VirtualLarry]

Will they expose the 10gb interfaces on the consumer product? that could be a massive value add.

There's 10GE on Zen? *whistle*. That could make these the go-to board / CPU for home server duties, especially if they allow ECC on the home-user boards as well.

 

[itsmydamnation]

There's 10GE on Zen? *whistle*. That could make these the go-to board / CPU for home server duties, especially if they allow ECC on the home-user boards as well.

The rumor is 2 per Zeppelin.

 

[Tuna-Fish]

There's 10GE on Zen? *whistle*. That could make these the go-to board / CPU for home server duties, especially if they allow ECC on the home-user boards as well.

There is 2x 10GBe on die, but honestly that is the first part I'd fuse off in a home cpu to be able to ask for more for the server ones.