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

[The Stilt]

SuperPI is nothing but cache & memory bandwidth / latency test, as is any application which uses AGM to calculate Pi. That's the very reason why 15h sucks in it.

I would prefer any of the Cinebenches over anything else, despite them being legacy workloads.

 

[Concerned Citizen]

So do you have an early sample or what and what is your opinion?
That means something.
To be honest ,probably you,me,and Flank3r will be earliest adopters.
I want to see how it do,1st.

 

[The Stilt]

So do you have an early sample or what and what is your opinion?
That means something.

Why on earth would I have a sample?

 

[Concerned Citizen]

It's just about October

 

[The Stilt]

Carrizo was the last AMD product I'm laying my hands on.

 

[Concerned Citizen]

Oh so not looking forward to new stuff? Me and flanker are.

 

[The Stilt]

I'm interested to see what AMD can do with Zen, but not up to a point where I would put any specific efforts into it. However business is business, so that might or might not cause the interest to wake at some point. Regardless, AMD's past incompetence and ignorance (Steamroller, Excavator, Polaris and the various cancelled projects) and the choices regarding the design (narrow FPU) and execution (14nm LPP) unfortunately made me lost all of the personal interest in Zen.

 

[Concerned Citizen]

It was pretty bad,and I thank you for your efforts to improve.

 

[LTC8K6]

Funny they will release a 6 core mainstream SKU just after ZEN launch when they had the opportunity to do it for more than 3-4 years (2017-2018)

If it was such a great idea, why has AMD waited so long?

 

[majord]

it's nice that you did the comparison Arachntronic, but I don't think even the biggest critic would think BW-E has a ~50% avg higher IPC (which is what your compatison graph essentially says). That's basically Excavator IPC levels.

So unless anyone believes the've created a lower clocked Excavator, the only logical conclusion is these leaked GB results are basically useless.

 

[Fanatical Meat]

So when will we know if Zen is great or sucks?

 

[Concerned Citizen]

February I'm thinking.
That sucks.

 

[crashtech]

oops, missed some replies...

 

[Azuma Hazuki]

it's nice that you did the comparison Arachntronic, but I don't think even the biggest critic would think BW-E has a ~50% avg higher IPC (which is what your compatison graph essentially says). That's basically Excavator IPC levels.

So unless anyone believes the've created a lower clocked Excavator, the only logical conclusion is these leaked GB results are basically useless.

Oh, I dunno, it could be useful if you have a raging, 15 year long hateboner for AMD. Should really see a doctor about that...


Not appropriate language, and insults are not allowed
Markfw900

 

[Nothingness]

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.

Are you sure GB4 MT tests share data? If not then coherency shouldn't impact performance.

 

[Concerned Citizen]

I wanna see SuperPi and Pifast or no go
Also Wprime.
Post times,pls.Thx

 

[cdimauro]

Albeit GB4 is much improved (compared to version 3) and looks similar to SPEC in some ways, synthetic benchmarks should be avoided, and real-world application used for testing CPUs.
So, let's wait for some serious test.

 

[Abwx]

Comparison with Bristol Ridge in GB3, this latter should be at 4GHz in ST and scaling its ST scores put it at the same perf/frequency than Zen if this latter was actually clocked at 1.44GHz, wich is obviously not the case....

http://browser.primatelabs.com/geekbench3/compare/6032158?baseline=8076878

 

[KTE]

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

CPUs, for many years, turbo boost with low threaded loads by default. It could well be running at 2-3GHz for ST and even higher than stock for MT.

Scaling of clocks vs performance is rarely linear to deduce, too.

Let me run GB4 on my IVB and Zacate...

Sent from HTC 10
(Opinions are own)

 

[KTE]

Comparison with Bristol Ridge in GB3, this latter should be at 4GHz in ST and scaling its ST scores put it at the same perf/frequency than Zen if this latter was actually clocked at 1.44GHz, wich is obviously not the case....

http://browser.primatelabs.com/geekbench3/compare/6032158?baseline=8076878

My IVB laptop locked at 1.4GHz/DDR3 800:
http://browser.primatelabs.com/v4/cpu/606061
http://browser.primatelabs.com/geekbench3/compare/6032158?baseline=8080029

GB3 1275 ST
GB4 1365 ST

Sent from HTC 10
(Opinions are own)

 

[The Stilt]

Someone (who is interested enough) recalculate those Zeppelin GB ST figures with assumption that the CPU worked at 1.0GHz, instead of 1.45GHz? 1.0GHz is a valid frequency state for the SKU used in that leak and the lowest (plausible, non PG) of the available states the CPU could have operated at.

If it still doesn't put it within ~20% of the IPC of Intel wells and lakes, then I'd say it is doing a DAR while running the benchmark. If not, AMD's lies have reached a completely new level.

 

[jpiniero]

Someone (who is interested enough) recalculate those Zeppelin GB ST figures with assumption that the CPU worked at 1.0GHz, instead of 1.45GHz? 1.0GHz is a valid frequency state for the SKU used in that leak and the lowest (plausible, non PG) of the available states the CPU could have operated at.

If it still doesn't put it within ~20% of the IPC of Intel wells and lakes, then I'd say it is doing a DAR while running the benchmark. If not, AMD's lies have reached a completely new level.

Well, there's one result of a 2961Y (Haswell, 1.1 Ghz) which gets 1258 ST in GB4. Couple of results for 847 (Sandy Bridge, 1.1 Ghz) which the median is roughly 1122 once you throw out the really lowball scores. Ignoring AES, it's pretty competitive to that Zen Server result. So it's plausible that the ZS is running at only 1 Ghz.

 

[Arachnotronic]

If I were AMD, I would conveniently "leak" a "real" Zen GB4 result ASAP if Zen really performs much better than this. If it doesn't, then I would do nothing because it's better to let people believe that this isn't real/representative to build up hype than to let the truth be known.

Anyway, I am really excited for the Zen launch, can't wait to see how it does when it's in the hands of consumers

 

[DrMrLordX]

I wouldn't read too much into GB3, that was a pretty flawed benchmark. .

I agree with that, and I still think there's something seriously wrong going on with the Zen ES system being used to run GB4 in these leaks.

it s not in AMD s interest to disclose anything about the final perfs,

The hell it isn't. Intel leaked a bunch of consumer-level Conroe ES chips into the hands of community overclockers and got millions in free advertising as they wowed the crowds with how bad-ass were those chips. On launch they were all that and a bag of chips.

If 8c/16t Summit Ridge is matching similarly-clocked Broadwell in FP-heavy tasks like Blender, at the very least AMD could wipe out all the "oh no a Zen ES sucked at Geekbench it's terrible AMD resign now" type posts we're seeing expressly because AMD won't let anyone in the community play with one. All we get are scraps taken out of context from someone who may be blatantly crippling the performance of their test platform with who-knows-what kind of background tasks.

Maybe the boards just aren't "there" yet for proper Summit Ridge operation, I don't know. But the secrecy around Zen is really hurting AMD at this point.

That's the very reason why 15h sucks in it..

XV is a lot stronger in SuperPi than SR or PD though. A lot stronger.

 

[The Stilt]

XV is a lot stronger in SuperPi than SR or PD though. A lot stronger.

Is, thanks to the larger L1D and lower latency L2. It would be even faster if AMD wouldn't have botched the DRAM latency on Streamroller and newer designs. Steamroller and Excavator have around 35% higher DRAM latency than Piledriver at the same exact settings.

If Excavator had faster and lower latency caches and memory interface than Skylake, it would probably outperform it in any AGM based benchmark despite it's completely inferior FP performance.