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

[Abwx]

but you would have to have confidence that AMD didn't rig the test beyond downclocking the Broadwell-E machine and disabling AVX2 like they did with that Polaris power consumption demo.

How did AMD rig the Polaris power consumption video?


Trolling is not allowed
Markfw900
Anandtech Moderator

New Blender results with an ES (Snowy Owl? / Naples?)

AMD Engineer Sample render time: 69 seconds
E5-2699 v3 (Haswell-EP, 2014) render time: 35 seconds

http://blenchmark.com/cpu-benchmarks

Where does it say that this is a Zen ES..?..
There are tons of ES at Geekbench, check the models..

As for Haswell EP rendering time being 35s so much, the i7 6900K is at 99s and the i7 6900 at 84s, as "accurate " as usual apparently...

 

[Burpo]

Why don't you think that is accurate? 36 thread Haswell vs 8 or 16 thread Skylake/Broadwell?

 

[Glo.]

New Blender results with an ES (Snowy Owl? / Naples?)

AMD Engineer Sample render time: 69 seconds
E5-2699 v3 (Haswell-EP, 2014) render time: 35 seconds

http://blenchmark.com/cpu-benchmarks

http://ark.intel.com/products/91317/Intel-Xeon-Processor-E5-2699-v4-55M-Cache-2_20-GHz

This is the CPU in question. Is there any information on AMD Engineering Sample? Any core count, clocks, anything?

Sorry bro, but until then your post is considered as misinformation with comparing top tier Intel CPU with anonymous engineering Sample from AMD.

 

[Burpo]

Sweepr pointed to a v3 CPU from 2014, not a 2016 v4 as your link shows.

 

[Abwx]

Why don't you think that is accurate? 36 thread Haswell vs 8 or 16 thread Skylake?

The 35s is with a dual socket plateform, if we extrapolate from the result of the i7 6900 at 3.2 then it would require 31 cores/62T at 2.3GHz to get to 35s assuming perfect scaling of the software.

This is likely 2 x 18 cores/72T/2.3GHz, so that s not the score of a single CPU as impliyed by Sweepr post.

This is the CPU in question. Is there any information on AMD Engineering Sample? Any core count, clocks, anything?

If that s a Zen ES then this is likely 16C/32T at 2.3GHz, given that they seems to have 8C/16T samples that are within 95W/3.15GHz then 16C/32T/2.3GHz should be within the same TDP.

 

[Glo.]

Your post is misinformation. Sweepr pointed to a v3 CPU from 2014, not a 2016 v4 as your link shows.

Thank you for correction, proper link: http://ark.intel.com/products/81061/Intel-Xeon-Processor-E5-2699-v3-45M-Cache-2_30-GHz

18 core CPU, in the end.

Also, in that list there is 6950X with 84 s score.

Would be quite funny if the AMD Engineering Sample was actually the 8 core version of the CPU...

i7-5960X - 91s.

 

[Sweepr]

Would be quite funny if the AMD Engineering Sample was actually the 8 core version of the CPU...

i7-5960X - 91s.

You can safely rule out this possibility - even in the AMD controlled demo with a i7-6900K gimped to 3 GHz (and god knows what other changes) - 8C/16T Summit Ridge barely matches it (tiny bit faster). This is most likely a server CPU, maybe two of them.

 

[Glo.]

If that s a Zen ES then this is likely 16C/32T at 2.3GHz, given that they seems to have 8C/16T samples that are within 95W/3.15GHz then 16C/32T/2.3GHz should be within the same TDP.

You can be right. There is in the list score for this CPU: http://ark.intel.com/products/81057/Intel-Xeon-Processor-E5-2695-v3-35M-Cache-2_30-GHz
The score is 58 s.
vs 69. for that engineering Sample.

You can safely rule out this possibility - even in the AMD controlled demo with a i7-6900K gimped to 3 GHz (and god knows what other changes) - 8C/16T Summit Ridge barely matches it (tiny bit faster). This is most likely a server CPU, maybe two of them.

If the scores were on the same level for 8 core versions, we need to look at other clues that could point to 16 core version. And the 14 core version of Intel Broadwell CPU is very close to that score of AMD Eng. Sample.

 

[Abwx]

You can be right. There is in the list score for this CPU: http://ark.intel.com/products/81057/Intel-Xeon-Processor-E5-2695-v3-35M-Cache-2_30-GHz
The score is 58 s.
.

14 cores at 2.3 cant get this score if the i7 6900 is at 99s at 3.2..

The combined cores x frequency is 14 x 2.3 = 32.2 while it s 8 x 3.2 = 25.6 for the i7, so the Xeon should be at 78.7s if the software has perfect scaling, wich is not the case and time should be closer to 80s.

Also compare with the i7 6950 wich has Core x frequency ratio of 30 and is at 84s...

Obviously the referenced frequencies are not accurate, this Xeon run close to its max turbo frequency, the score suggest 3.17GHz with perfect software scaling, so likely that it s at 3.3....

 

[lolfail9001]

Folks going all defensive over this are funny.

Considering it goes back to August i can certainly bet this is the same sample that popped in geekbenches and it performs as expected

 

[lifeblood]

Latest rumors say the 8c/16t chip will have clocks at 3.15 base and 3.3 (all core) boost. Am I allowed to be disappointed with that? The worse I was expecting was 3.2 base with 3.5 boost, and I was hoping for 3.5 base with a 4.0 boost. What good is a 40% IPC boost if you lose 20% clock speed?

If this is true it looks like AMD will be back to claiming “more cores” and competing on price. Performance wise it will only be able to compete with Intel’s lower speed i5’s and i7’s, which is better than the current situation with Vishera, but not nearly what I had hoped for.

 

[Redentor]

So even Broadwell-E is competing on price against Skylake-S? :|

Broadwell-E has lower IPC, lower frequencies, but costs a lot more than Skylake-S. Why?

Zen will be a fuc*ing 8C/16T with a Broadwell IPC and under 95W of TDP!

I think that you are crazy, guys ... your minds are lost.

 

[bjt2]

Latest rumors say the 8c/16t chip will have clocks at 3.15 base and 3.3 (all core) boost. Am I allowed to be disappointed with that? The worse I was expecting was 3.2 base with 3.5 boost, and I was hoping for 3.5 base with a 4.0 boost. What good is a 40% IPC boost if you lose 20% clock speed?

If this is true it looks like AMD will be back to claiming “more cores” and competing on price. Performance wise it will only be able to compete with Intel’s lower speed i5’s and i7’s, which is better than the current situation with Vishera, but not nearly what I had hoped for.

This is only an early ES of a completely new chip on a completely new process.
For comparison, early ES of Bulldozer ran at 2.6-2.8GHz, and now we know the actual frequency...
Let's wait final silicon...

 

[krumme]

Latest rumors say the 8c/16t chip will have clocks at 3.15 base and 3.3 (all core) boost. Am I allowed to be disappointed with that? The worse I was expecting was 3.2 base with 3.5 boost, and I was hoping for 3.5 base with a 4.0 boost. What good is a 40% IPC boost if you lose 20% clock speed?

If this is true it looks like AMD will be back to claiming “more cores” and competing on price. Performance wise it will only be able to compete with Intel’s lower speed i5’s and i7’s, which is better than the current situation with Vishera, but not nearly what I had hoped for.

No. You are not allowed. Because it nonsense.

The same latest rumors you are refeering says hsw bw like ipc at a base of 3.15 for a 8c part at 95w tdp. You "missed" the ipc part. Thats bwe 8c core part perf that is otherwise like 1000 usd. And ipc and max freq is as we know tightly connected. If a 180mm2 like cpu delivers that its way beyond darn good. And i would expect something a good deal less performant even from a sound architecture at the same footprint. Bwe is not exactly cheap tech on a cheap node in a cheap size. And it shows.

 

[Asterox]

Latest rumors say the 8c/16t chip will have clocks at 3.15 base and 3.3 (all core) boost. Am I allowed to be disappointed with that? The worse I was expecting was 3.2 base with 3.5 boost, and I was hoping for 3.5 base with a 4.0 boost. What good is a 40% IPC boost if you lose 20% clock speed?

If this is true it looks like AMD will be back to claiming “more cores” and competing on price. Performance wise it will only be able to compete with Intel’s lower speed i5’s and i7’s, which is better than the current situation with Vishera, but not nearly what I had hoped for.

What you do not understand, or it is preferable that you are quite confused, or what is possible or realistic impossible?

- AMD Zen "FX-9100" 8 Cores/16 Threads, 3.1ghz, TDP 95W, Samsung 14nm

- Intel i7 6900K 8 Cores/16 Threads, 3.2ghz, TDP 140W, Intel 14nm

Why this Intel CPU does not work at 3.5GHz, and what would actually be the TDP on that CPU operating frequency?

In the end what to write, yes AMD is again marching with "More Cores" no doubt.

 

[lolfail9001]

Why this Intel CPU does not work at 3.5GHz, and what would actually be the TDP on that CPU operating frequency?

A. 6900k actually works at 3.5Ghz and actually consumes way less than 140W in real life except Prime95. But want to make a signature bet that Zen 8 core will consume way more in Prime95 than 95W?

 

[Asterox]

A. 6900k actually works at 3.5Ghz and actually consumes way less than 140W in real life except Prime95. But want to make a signature bet that Zen 8 core will consume way more in Prime95 than 95W?

Not in highly multithreaded apps, if you do a video rendering(H265 very CPU hungry guy), or 3D modeling this CPU 8 Cores is pushed to maximum.When this CPU is working on maximum, it will work at base CPU frequency 3.2ghz/CPU is not overclocked.i7 6900K is not a gaming CPU, this processor is say first step is some serious video editing or 3D modeling, etc.

TDP is not a simple or direct CPU power consumption number.

If CPU has very good performanse, high or higher power consumption is not very important.

- FX-8150 as we know, low performanse and high power consumption=swearing and breaking wine glasses

- FX-8150+25% higher IPC vs Phenom II, high power consumption=not important let's have a Czech beer this is good CPU

 

[lolfail9001]

Not in highly multithreaded apps, if you do a video rendering(H265 very CPU hungry guy), or 3D modeling this CPU 8 Cores is pushed to maximum.When this CPU is working on maximum, it will work at base CPU frequency 3.2ghz/CPU is not overclocked.

You are confusing "AVX2 apps" and highly multithreaded apps. If i launch "make -j 16" on linux kernel tree with 6900k i will have 8 cores working at 3.5Ghz. So, you are wrong.

 

[The Stilt]

Checked out the "Blenchmark". E5-2699 V3 will sustain 2.6GHz when all cores / threads are loaded, regardless if it is a AVX2 workload or not.

2.77a public build:

14C/28T @ 2.3GHz = 90.40s
18C/36T @ 2.3GHz = 77.53s

2.78 MSVC 2015 build (the AVX2 one I posted yesterday):

14C/28T @ 2.3GHz = 62.16s
18C/36T @ 2.3GHz = 53.04s

I wouldn't be so sure that the "AMD ES" result is made on Zeppelin at all.

 

[DrMrLordX]

New Blender results with an ES (Snowy Owl? / Naples?)

AMD Engineer Sample render time: 69 seconds
E5-2699 v3 (Haswell-EP, 2014) render time: 35 seconds

http://blenchmark.com/cpu-benchmarks

I doubt it's Snowy Owl. But it's impossible to tell from the benchmark alone. We don't know core count or clockspeed.

 

[LTC8K6]

What you do not understand, or it is preferable that you are quite confused, or what is possible or realistic impossible?

- AMD Zen "FX-9100" 8 Cores/16 Threads, 3.1ghz, TDP 95W, Samsung 14nm

- Intel i7 6900K 8 Cores/16 Threads, 3.2ghz, TDP 140W, Intel 14nm

Why this Intel CPU does not work at 3.5GHz, and what would actually be the TDP on that CPU operating frequency?

In the end what to write, yes AMD is again marching with "More Cores" no doubt.

Again, let’s take a look at the individual core frequencies for the three configurations we tested. Despite a 3.2GHz base clock rate, Turbo Boost pushes the CPU to 3.7GHz in lightly threaded workloads. Our sample managed to maintain this frequency across all cores during the stress test

So it held 3.7 on all cores during some stress.

And they have the power levels during gaming and torture, etc.

It took torture to get to 140W, and an overclock to exceed that.

http://www.tomshardware.com/reviews/intel-core-i7-broadwell-e-6950x-6900k-6850k-6800k,4587-9.html

 

[Abwx]

So it held 3.7 on all cores during some stress.

And they have the power levels during gaming and torture, etc.

It took torture to get to 140W, and an overclock to exceed that.

http://www.tomshardware.com/reviews/intel-core-i7-broadwell-e-6950x-6900k-6850k-6800k,4587-9.html

Those numbers not only are meaningless but they are artificialy reduced thanks to sub par voltage margin, think about it, according to Computerbase.de undervolting by 7-8% only allow to boot windows and make run such apps, Hardware.fr didnt even bother to make their usual undervolting test because they know that the margin is below specs..

Of course MB manufacturers will increase the voltage to get the plateform within decent stability and not take the blame for Intel tricky marketing, once supplied within the normal voltage margin it will gladly get close to 140W with stress tests.

On the other hand it s not in AMD habits to undersize the voltage margins, quite the contrary, so just this fact is a hint that Zen will have no trouble being competitive perfs and perf/Watt wise.

 

[lolfail9001]

Those numbers not only are meaningless but they are artificialy reduced thanks to sub par voltage margin, think about it, according to Computerbase.de undervolting by 7-8% only allow to boot windows and make run such apps, Hardware.fr didnt even bother to make their usual undervolting test because they know that the margin is below specs..

So, what undervolting has to do with any of it? Also, there are enough 1V for 4Ghz Broadwell-Es as is, so this is a pretty expensive overvolt/bad chip on top of it.

so just this fact is a hint that Zen will have no trouble being competitive perfs and perf/Watt wise.

We'll see when it actually comes out.

 

[deasd]

Folks going all defensive over this are funny.

Considering it goes back to August i can certainly bet this is the same sample that popped in geekbenches and it performs as expected

I guess it is 16 cores Snowy Owl operating at 1.4-2.0Ghz just like that Geekbench leak.


EDIT: here's the info about that platform:
http://www.blenchmark.com/content/amd-engineering-sample

Submitted by Blender on Thu, 08/25/2016 - 13:45
AMD Engineering Sample
Operating System:
Windows 2008ServerR2 64bit
Blender Version:
2.77a
Render Time:
0:01:09
Type:
CPU
Addon Version:
1.0.5

It was an old result, so still it's very likely the same SKU as that on Geekbench.

 

[Dresdenboy]

I guess it is 16 cores Snowy Owl operating at 1.4-2.0Ghz just like that Geekbench leak.

1.44GHz is not the correct frequency. As with earlier AMD CPUs, GB seems to read some power management table but forgets a shift left by 1. So 1.44 seems to be just 2.9/2, as 2.0 for a 4GHz turbo Bristol Ridge is also just half the truth.