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

[F-Rex]

Which were confirmed by the AoTS leak (at least for 95W AM4 part).

The AoTS leak (if true) only shows an AMD ES clocking @ 2;8/3,

Based on a rumor I heard, the 32C/64T part has base clock of around 1.5GHz. Those turbo frequencies (single core) should be accurate, based on the same rumor.



Based on what?
The partners made their AM4 motherboards using Bristol Ridge SKUs, meaning they didn't have the need to have access to extremely early prototypes. Since the leaks obviously originate from China (AotS), it means that AMD has now shipped Zeppelins to the motherboard / system manufacturers. And as I said, they didn't and still have no need to do that unless it is a silicon representing the final product. There will definitely be some improvements, but I would expect them to come purely from the maturing process.

Who said this ES does not represent the final product?
You think this ES Will be the highest frequency part ( at least @ launch). But so far we don't know amd plan regarding zen fréquence.

 

[The Stilt]

The thing is, when you see multiple leaks indicating the same thing there is usually at least some truth in them. So far the info leakaged by "AMD Polaris" and AotS match perfectly.

 

[lefty2]

Except you can't really use the regular framerate as a guide since the 480 is mostly GPU limited, even on High. 3930K for instance gets roughly 5600-6100 in general; While the 5960X gets into the 9000s and the 6700K gets into the high 8000s. Even the regular 6500 tends to get about 5800-6000. Granted you don't know how high those unlocked processors are overclocked....

The 9590 gets roughly 5000-5400 btw.

Are you sure those are benched on the same version of AotS? The Zen sample was benched on version 1.30. You can select the game version from the "more filters" tab.

 

[jpiniero]

Are you sure those are benched on the same version of AotS? The Zen sample was benched on version 1.30. You can select the game version from the "more filters" tab.

The version was all over the place really. But assuming they haven't changed the test at all, I think it would be fine as a comparison since you would think the CPU portion would only get more optimized over time but probably not significantly.

Guess we'll have to wait until someone does more extensive benchmarking before we can draw any real conclusions, then?

That's true but it doesn't look good.

 

[Sweepr]

That's true but it doesn't look good.

Indeed. It might perform better (overall) relative to Intel CPUs in the end, but these results in particular are not impressive. Leaving the 6700K out, a locked 2013 Haswell i7 outperforms it here, and based on PCWorld testing there's a >30% gain going from 4C/8T -> 8C/16T (same chip).

- Summit Ridge 8C/16T ES - 2.8-3.2 GHz (2016/2017 Zen)
Average: 58.9 FPS

- Core i7-4770 4C/8T 3.4-3.9 GHz (2013 Haswell)
Average: 66.0 FPS

 

[Sheep221]

If that ES has some 20-35W TDP it's fine, if it is 95, 125 or higher - not impressed.

 

[Sweepr]

AMD ZEN Engineering Sample AOS - Further Performance Analysis

http://www.guru3d.com/news-story/amd-zen-engineering-sample-aos-further-analysis.html

 

[SAAA]

I have to admit looking at these graphs there seem to be an issue with FPS scaling and AMD cpus because 5960X literally doubles frames going from extreme to low settings, so something might be fixed with drivers or software changes alone.
The CPU frame and whatever it means still don't paint a good picture tough, unless Ashes runs that faster on Intel side for AVX instructions only, AMD cores are left behind even with a decent 50% advantage for Zen over FX.
What's to like? Apparentely Zen is up ad running games on some guy's machine (Doom too?), better than nothing.

 

[Erenhardt]

Why would they put 2.8 Zen against 3.2GHz competition? That is 15% disadvantage for the start.

 

[Sweepr]

Don't forget FX has a multi-thread performance penalty because of CMT and Summit Ridge (Zen) has 8C/16T' ('HT'), so that's a big advantage from the start if you want to do an IPC comparison.

 

[lefty2]

The version was all over the place really. But assuming they haven't changed the test at all, I think it would be fine as a comparison since you would think the CPU portion would only get more optimized over time but probably not significantly.

I wouldn't make that assumption. There were a good few changes in 1.3

Just out of curiousity I filtered results just for RX 480 GPU and i7-6700K and I get 5 differeent results ranging from 6000 to 5,400. So, the results actually do vary widely for different versions of the game.

 

[deasd]

Testing CPU by using game bench is silly, no matter what game you bench, not to mention the whole platform including CPU are still not fully functioned.

 

[know of fence]

Why would they put 2.8 Zen against 3.2GHz competition? That is 15% disadvantage for the start.

Agreed. How unfortunate, Guru3D's Hilbert Hagedoorn made all those graphs for nothing. Maybe that's a natural assumption that people always set their CPU to just clock to max turbo, or more likely - it isn't.
Perhaps 2.9 or 2.8 GHz would be fair, given that the game occupies up to 16 threads, even though performance doesn't improve significantly past 4 cores.

Source: http://semiaccurate.com/2016/03/01/investigating-directx-12-cpu-scaling/

 

[Sweepr]

Very different results compared to PCWorld:

 

[bystander36]

Why would they put 2.8 Zen against 3.2GHz competition? That is 15% disadvantage for the start.

Because it's the best we have to compare to. Because that's the clock they decided it would run at. It's not the final product, but if you are interested in what Zen can do, this is the best we have to go off of now. Show us a better comparison if you don't like it.

 

[know of fence]

Very different results compared to PCWorld.

Yes, well you can immediately see from the numbers. Scaling up to around 60-ish frames that would be GPU framerate, around 120-ish that would be their heavily fluctuating "CPU-Framerate".
There is a bit of a problem, it really should be the other way around, Zen ES compares relatively well in FPS tests and not so well with CPU-framerate (just looking at the one core- i5 it manages to best, ignoring the deltas), which may be down to differences in bandwidth of cache or memory. FX CPUs may have 8 integer cores but they still are only dual channel systems.
A hint of this we get from the huge score gap between Skylake and Haswell, which basically shouldn't exist based on clocks or IPC, except Skylake can have up to twice as much bandwidth due to DDR4.

 

[DrMrLordX]

That's true but it doesn't look good.

Eh? The results might be inconclusive, but it looks to me like it beats the ever loving hell out of Vishera. The first thing AMD had to do was beat its own Con cores (including hypothetical XV products) and it laid down the smack right there. Again, assuming AotS reaches peak performance on 8 physical cores or less (which Sweepr's graph seems to indicate that it does).

Of course, there is also the issue of the 8350 only having the equivalent of 4 core's worth of FP resources . . .

Yes, well you can immediately see from the numbers. Scaling up to around 60-ish frames that would be GPU framerate, around 120-ish that would be their heavily fluctuating "CPU-Framerate".
There is a bit of a problem, it really should be the other way around, Zen ES compares relatively well in FPS tests and not so well with CPU-framerate (just looking at the one core- i5 it manages to best, ignoring the deltas), which may be down to differences in bandwidth of cache or memory. FX CPUs may have 8 integer cores but they still are only dual channel systems.
A hint of this we get from the huge score gap between Skylake and Haswell, which basically shouldn't exist based on clocks or IPC, except Skylake can have up to twice as much bandwidth due to DDR4.

Hmm, interesting observation.

 

[Beer4Me]

Zen just proves AMD should really just fold their CPU division as it's a losing enterprise altogether. They've not been *any* competition to Intel since the glory of the "Athlon/Duron" days. Their GPU division OTOH (formerly ATi) is in much better shape, and I think the money saved from scrapping the CPU division is better spent on marketing/RD in the GPU space fighting nVidia.

EDIT: If it is any consolation, the AMD APU arm could be spared and rolled into the GPU division IMHO as the lines get blurred as GPUs take on more compute tasks.

 

[frozentundra123456]

Why would they put 2.8 Zen against 3.2GHz competition? That is 15% disadvantage for the start.

The graphs in sweeper's post, #207, are all at 3.2. So actually, that makes Zen compared to vishera look very bad. They actually had to downclock Vishera to show the difference. I assume they did that to show the ipc gain, which is in the range of 40% they promised, but more than half that gain would be lost to decreased clockspeed, comparing stock 4ghz vishera and max demonstrated clock of 3.2 for Zen. Actually, they were being quite generous to zen, running it at it max demonstrated clockspeed vs base clock for intel (which has proven to oc to at least 4.0), and a downclocked Vishera.

 

[Phynaz]

Indeed we do. Or that's the rational thing anyway. For certain values of "we" (notably Phynaz and Arachnatronic...) bashing AMD seems to take priority. I can't imagine why; what do they think their precious Intel chips would cost without AMD around?

I'm betting I've bought more AMD equipment than you have.

My opinions bothering you to such a degree says more about you than they say about me.

 

[DrMrLordX]

The graphs from post 207 are bizarre. I don't think we can draw any meaningful conclusion from those graphs, other than that AotS has serious CPU scaling issues on Vishera and Summit Ridge. The CPU utilization charts are quite telling.

 

[blublub]

The graphs from post 207 are bizarre. I don't think we can draw any meaningful conclusion from those graphs, other than that AotS has serious CPU scaling issues on Vishera and Summit Ridge. The CPU utilization charts are quite telling.

That could likely be solved via a simple micro code update

 

[Doom2pro]

It feels like someone is trolling us hard core... People are so desperate to get some performance measure out of Zen they are diving at this potentially fake benchmark like it's 100 dollar bills falling from the sky.

Why, WHY release a CPU benchmark using a video game, with one of the Intel CPUs overclocked like Charlie Sheen on 10 cent cocaine day? It makes no sense, unless someone is trolling us hard core...

Just because this person knows the SKU of the new CPU doesn't mean he actually has Engineering Samples or has actually benched them... And yet look at all these people trying to make sense of something that might not even be real in the first place.

Sigh, Zen can't come soon enough.

 

[Erenhardt]

The graphs in sweeper's post, #207, are all at 3.2. So actually, that makes Zen compared to vishera look very bad. They actually had to downclock Vishera to show the difference. I assume they did that to show the ipc gain, which is in the range of 40% they promised, but more than half that gain would be lost to decreased clockspeed, comparing stock 4ghz vishera and max demonstrated clock of 3.2 for Zen. Actually, they were being quite generous to zen, running it at it max demonstrated clockspeed vs base clock for intel (which has proven to oc to at least 4.0), and a downclocked Vishera.

No, Zen is 2.8 and the rest is 3.2. The turbo frequency is not kicking in when you have all cores loaded. They did not run it at max clocks. Where you take these things from?
2.8 base and 3.2 turbo is slower than 3.2 base an no trubo 24/7 365d/y

So again. This fx and intel CPUs should be downclocked another 15% to match zen frequency.

 

[The Stilt]

It feels like someone is trolling us hard core... People are so desperate to get some performance measure out of Zen they are diving at this potentially fake benchmark like it's 100 dollar bills falling from the sky.

Why, WHY release a CPU benchmark using a video game, with one of the Intel CPUs overclocked like Charlie Sheen on 10 cent cocaine day? It makes no sense, unless someone is trolling us hard core...

Just because this person knows the SKU of the new CPU doesn't mean he actually has Engineering Samples or has actually benched them... And yet look at all these people trying to make sense of something that might not even be real in the first place.

Sigh, Zen can't come soon enough.

There are several things in the information received from these leaks and all of them are extremely hard, if not impossible to forge. As I've said before Zen has slightly different method the generate the CPU clocks, which leads to somewhat unusual operating frequencies. The clock frequencies posted by "AMD Polaris" match this pattern. Also the CPU name string displayed by AotS is a real deal. And since the stepping identifiers and all the other information in the SKU name string is design specific, it cannot be forged either.

I'd say both of the leaks we've had so far are real, the only question is if the actual scores in AotS are. Most likely they're too, since the leak came from China / Taiwan which at this point makes are perfect sense.