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

[jpiniero]

They said march for server chips...

They also said October and then 2016 for Summit Ridge too.

 

[lolfail9001]

I picked the E5 with the best MT score when i did my comparision, the one you have picked for whatever reason is an outlier compared to the other 44 core E5 results.
https://browser.primatelabs.com/v4/cpu/45716

but then look at the sub score differences between your one and my one and you will see the likely difference is memory speed. So what speed memory is Zen running?

Well, the GB4 leak had ridiculously bad memory characteristics, considering it was running a 128GB setup (your example with bad memory performance runs a 1TB setup).

Who knows. Anyway, I'm going to run my BDW-E at 1.44GHz (DDR4-2400) and see what single core score I get...

Post a link, please.

 

[itsmydamnation]

you should get quite a bit higher, one RLDDR4 isn't normally that fast and two you dont have the extra 54 cores of cache coherency to worry about.

Well, the GB4 leak had ridiculously bad memory characteristics, considering it was running a 128GB setup (your example with bad memory performance runs a 1TB setup).

Amount of memory shouldn't make a big difference unless memory channels aren't populated, granted 1tb on E5 would need both dimms for each channel populated, while the 128gb likely just had 8x16. Think its far more likely those 16gb dimms are 2400 and the 64 of the result i used are 1600/1866.

 

[Sweepr]

So, Desktop 6-core Coffee lake will come first (H2 2017) but its the Laptop 6-Core SKU coming one year later (Q2 2018) thats the reason we have the desktop part ??? is this some form of surrealism or what ??

No, it's actually very simple. The reason you see 6C/12T Coffe Lake for desktops is - mobile is getting it - Intel might be confident about its performance at 35-45W TDP level. They wouldn't bother otherwise, especially with 6C/12T HEDT coming close to LGA 115x i7 pricing. And let me refresh your memory: Skylake launched first for desktop PCs so don't play surprised here. I expect the same for Coffee Lake based on the DigiTimes rumor.

 

[Concerned Citizen]

What is this? WCCFtech or something?

 

[jpiniero]

No, it's actually very simple. The reason you see 6C/12T Coffe Lake for desktops is - mobile is getting it - Intel might be confident about its performance at 35-45W TDP level. They wouldn't bother otherwise, especially with 6C/12T HEDT coming close to LGA 115x i7 pricing. And let me refresh your memory: Skylake launched first for desktop PCs so don't play surprised here. I expect the same for Coffee Lake based on the DigiTimes rumor.

Wait, you are thinking Intel will release Coffee Lake in 2017 on the desktop? Uh, no. Kabylake-S in January, yes, and then Kaby/Skylake-X in the middle of the year.

 

[Sweepr]

Wait, you are thinking Intel will release Coffee Lake in 2017 on the desktop? Uh, no. Kabylake-S in January, yes, and then Kaby/Skylake-X in the middle of the year.

Already explained this in an earlier post. DigiTimes is saying H2-2017, but I think early 2018 (around CES) is more likely - so 1 year after Kaby Lake-S. Mobile parts following in Q2.

 

[Abwx]

In GB4 Zen @ 1.44GHz gets a single-core score of 1141.

https://browser.primatelabs.com/v4/cpu/488835

Xeon E5 2699v4 @ 3.6GHz single core gets a single core score of 3616. 3616*(1.44/3.6) = 1446.

How do you know that Zen is actually clocked at 1.44GHz in those tests..?..

What if it s 1GHz for instance.?.

Indeed the link i posted with a comparison to a 2.5GHz K10 Magny Cours show this latter often getting over 2x the single core result of Zen in some INT tests, FTR INT IPC is much more difficult to improve than FP, this show in the FP tests where Magny cours does not show as big advantage as in INT, wich suggest that Zen FP improvement is much higher in % than INT, wich is expected when looking at the published diagrams.

https://browser.primatelabs.com/geekbench3/compare/8073317?baseline=8076878

 

[Concerned Citizen]

Nevermind.
How is the IPC,aye?

 

[Arachnotronic]

Well, the GB4 leak had ridiculously bad memory characteristics, considering it was running a 128GB setup (your example with bad memory performance runs a 1TB setup).

Post a link, please.

https://browser.geekbench.com/v4/cpu/598163

CPU frequency @ 1.4GHz, L3$ @ 1.4GHz, memory speed reduced to DDR4-1600.

 

[Arachnotronic]

I put the data in a nice chart to compare with the leaked Zen GB4 results:

 

[The Stilt]

I'd say the Naples rig that did a re-appearance in the Geekbench database is still either doing some low-level debugging in the background (SBDAR), or alternatively operating at significantly lower speed than we are expecting (based on the SKU) it to. In case the previously leaked results (Geekbench + AoTS) would represent the actual relative performance on Zeppelin, AMD's board better start packing. Including the CEO.

 

[Arachnotronic]

I'd say the Naples rig that did a re-appearance in the Geekbench database is still either doing some low-level debugging in the background (SBDAR), or alternatively operating at significantly lower speed than we are expecting (based on the SKU) it to. In case the result would represent the actual relative performance on Zeppelin, AMD's board better start packing. Including the CEO.

Geekbench 4 seemed to detect that my 6950x was running at 1.4GHz, so I think it was running at 1.44GHz.

Anyway, it could very well have been doing some stuff in the background or something else to nuke the performance. If these results really are representative of the kind of perf/MHz that Zen will deliver, then...yeah, that'd be really bad.

 

[Concerned Citizen]

What does The Stilt say? Does he have an ES?
Enquiring minds like to know.

 

[Concerned Citizen]

It it runs @ 800 Mhz and beats 4.4 Ghz Kaby Lake;I want to know.

 

[The Stilt]

Geekbench 4 seemed to detect that my 6950x was running at 1.4GHz, so I think it was running at 1.44GHz.

Anyway, it could very well have been doing some stuff in the background or something else to nuke the performance. If these results really are representative of the kind of perf/MHz that Zen will deliver, then...yeah, that'd be really bad.

Did you adjust the frequency in flight (i.e after Windows has loaded) or did you boot at that speed?
AFAIK Geekbench shows the frequency which the system has been booted at.

But yeah, their hardware detection / monitoring is pretty pathetic / obsolete. They aqquire the hardware information with CPU-Z SDK, which is dated to 2013 IIRC (i.e extremely outdated), even in Geekbench 4.

 

[Arachnotronic]

Did you adjust the frequency in flight (i.e after Windows has loaded) or did you boot at that speed?
AFAIK Geekbench shows the frequency which the system has been booted at.

But yeah, their hardware detection / monitoring is pretty pathetic / obsolete. They aqquire the hardware information with CPU-Z SDK, which is dated to 2013 IIRC (i.e extremely outdated), even in Geekbench 4.

I adjusted all parameters in the BIOS before booting into Windows.

 

[jpiniero]

I put the data in a nice chart to compare with the leaked Zen GB4 results:

I looked at a random Celeron craptop (1007U, Ivy Bridge 1.5 Ghz) and that's actually not too far off from your result in a good portion of the results, even accounting for the slightly higher clock. Besides AES.

https://browser.geekbench.com/v4/cpu/578348

Anyway, it could very well have been doing some stuff in the background or something else to nuke the performance. If these results really are representative of the kind of perf/MHz that Zen will deliver, then...yeah, that'd be really bad.

I suppose one possibility is that the test is running in a virtual machine. But the performance on this is so bad I don't think the VM overhead alone would cause this.

 

[Arachnotronic]

They also said October and then 2016 for Summit Ridge too.

They never said October, but a leaked internal slide indicated so.

 

[Arachnotronic]

This is why I want a Bristol Ridge desktop to play with. I'd like to do some extensive testing of the Excavator core to see what sort of results we SHOULD be seeing assuming AMD'd 40% claim is true across the board.

On paper Zen looks a heck of a lot better than Bulldozer and its derivatives. I'm not expecting an Intel killer, but being this far behind Broadwell CPU seems...well, it doesn't line up with AMD's public statements at all.

 

[Abwx]

Geekbench 4 seemed to detect that my 6950x was running at 1.4GHz, so I think it was running at 1.44GHz.

.

If it was running at 1.44GHz it would match BDW in the FP tests, or at least in the ray tracing test, yet according to your graph even if Zen had 100% scaling in SMT it would still be roughly 30% below BDW in the Blender test displayed by AMD.

 

[Concerned Citizen]

Is the megahertz really that low?Is it competitive?

 

[jpiniero]

If it was running at 1.44GHz it would match BDW in the FP tests, or at least in the ray tracing test, yet according to your graph even if Zen had 100% scaling in SMT it would still be roughly 30% below BDW in the Blender test displayed by AMD.

AMD's Blender test was deliberately misleading.

 

[deasd]

If you let me choose, I would trust Superpi rather than Geekbench, IMHO.

OTOH, Blender just show Zen's MT performance matches Broadwell, but it tells nothing more, because Excavator with L3 might have similar MT performance.

 

[Concerned Citizen]

Superpi or get out!