AMD Vega (FE and RX) Benchmarks [Updated Aug 10 - RX Vega 64 Unboxing]

[Oddzz]

My try to put Vega FE against Titan X into prospective:

I used the numbers from GamersNexus and PCPerspective benchmarks (@ stock settings).

However, this chart is showing all benchmarks of GN and PCP :

 

[Gideon]

Not sure if it deserves its own topic:
AMD RX Vega Leaked Benchmark Shows It Ahead Of GTX 1080 – Specs Confirmed, 1630MHz Clock Speed, 8GB HBM2 & 484GB/s Of Bandwidth

RX Vega, not Vega FE.

wccftech is being wccftech. IMO Videocardz does a better job explaining:

Well, I didn’t really want to post this because I think it’s still too early, but since the highest score started to float around the web I think it’s worth to clear some misunderstanding. The highest score the 687F:C1 has achieved is an overclocked chip. 3DMark11 does not recognize unreleased overclocked graphics cards very well. The good news is that this puts RX Vega above overclocked GTX 1070, bad news, it might still be slower than overclocked GTX 1080. I guess time will verify those results.

Speaking of 3DMark11 scores, compared to some older benchmark results (~3 months old) the performance has increased by around 15% (quite impressive).

The Memory clock is a let-down, being same as FE edition. I was hoping the memory fabs can reach their claimed clocks (2 GHz), as this alone would have improved performance ~10%

 

[tg2708]

Let's go AMD prove the naysayers wrong and release a smash hit. I'm team AMD until rx vega releases then I'm back to being a neutral party. I think being competitive with the 1080 is rather good if priced right as the ti might be a bit of stretch until drivers mature. Power consumption might be a tad high but if the card runs without sounding to horrible it might be somewhat manageable.

 

[parvadomus]

Which smelled like?

A combination of underperforming / power hog

 

[Geegeeoh]

wccftech is being wccftech. IMO Videocardz does a better job explaining:

The Memory clock is a let-down, being same as FE edition. I was hoping the memory fabs can reach their claimed clocks (2 GHz), as this alone would have improved performance ~10%

Why put the 30400 score above 306xx+/31873 ?

 

[IEC]

Why put the 30400 score above 306xx+/31873 ?

GTX 1080 vs Vega (some flavor)

 

[Geegeeoh]

GTX 1080 vs Vega (some flavor)

For that you have the colors, imho the scores should be ordered... or, at first glance, you get tricked.

 

[leoneazzurro]

Primitive Shaders are mentioned. Just under the Pixel engine.

It is all part of Geometry Pipeline.

I know they are part of the geometry pipeline, but they are not mentioned explicitely in that page. They were quite explicely presented in the architecture presentation, anyway.

 

[Bacon1]

No one will buy either for pro tasks, they will buy a quadro because certified drivers are much more important then a little extra performance. The whole Vega FE for pro is a complete red herring fired out their by AMD marketing to make Vega look better at something. Both the Titan and the Vega FE are gaming cards. The titan sells because it is the absolute fastest, Vega FE won't because it isn't.

Sorry but that is just wrong. People have already bought it for Pro tasks and GN's Titan XP article was only made because someone loaned them their GPU they had purchased for machine learning tasks. You are being ridiculous by saying that no one would buy these for Professional tasks as not everyone needs specific application signed drivers. That's why the Titan line was first created after all!

Vega FE kills the Titan XP for productivity tasks and 1080 Ti kills Titan XP for gaming. Nvidia's top end no longer has a point.

 

[Glo.]

Told you that compute performance per clock is higher in Vega than it is in Fiji.

In games it shows reduction per clock versus Fiji.

http://www.gamersnexus.net/guides/2977-vega-fe-vs-fury-x-at-same-clocks-ipc
Food for thought.

 

[Bacon1]

My try to put Vega FE against Titan X into prospective:

I used the numbers from GamersNexus and PCPerspective benchmarks (@ stock settings).

However, this chart is showing all benchmarks of GN and PCP :

Why aren't you using the actual Power numbers?

...the Vega FE stock configuration draws an average of 384W load...

...Adding the Titan Xp stock card to the test, we see an average power consumption of about 357W – so that’s 27W lower than the Vega: FE system...

That's a 7% difference not 20% like your chart shows. Remove the rest of the system (est 84w) you get: 300 vs 273 or still under a 10% difference.

Also why only include 3 productivity related tasks? They both tested more than that.

 

[Phynaz]

Vega FE kills the Titan XP for productivity tasks and 1080 Ti kills Titan XP for gaming. Nvidia's top end no longer has a point.

Cuda.

End.

 

[Bacon1]

Cuda.

End.

Glad you can see that people did use Titans for productivity, now hopefully you can also realize that not every tasks requires CUDA either

 

[Phynaz]

Glad you can see that people did use Titans for productivity, now hopefully you can also realize that not every tasks requires CUDA either

I'm not sure how this post is a counter argument to Cuda keeping Titan relevant, but whatever.

Where's your next goalpost move going to be?

 

[tamz_msc]

I'm not sure how this post is a counter argument to Cuda keeping Titan relevant, but whatever.

Where's your next goalpost move going to be?

The Titan cards have never been relevant, except the Kepler one which had strong double precision performance, and to a lesser extent the Maxwell one with good FP16. CUDA works with every NVIDIA GPU, not just the Titans.

 

[tential]

If the 1080ti makes the titan xp irrelevant, then swap the 1080ti out to use instead and the 1080ti is even better vs the fe.....

 

[Bacon1]

I'm not sure how this post is a counter argument to Cuda keeping Titan relevant, but whatever.

Where's your next goalpost move going to be?

Uhh you are the one moving the goalposts. We've seen Vega FE faster than the Titan XP in 5/8 of the Pro tasks tested, and better price/perf in 6/8. Then you stated "Bbuuut Cuda... END" while its clear that CUDA isn't making Titan XP faster in these workloads.

 

[Bacon1]

If the 1080ti makes the titan xp irrelevant, then swap the 1080ti out to use instead and the 1080ti is even better vs the fe.....

Yep for gaming the 1080 Ti is clearly better than the Vega FE. For Productivity Vega FE is clearly better. Vega FE was marketed toward productivity and 1080 Ti was marketed toward gamers. Don't see what's wrong with that.

 

[tential]

Yep for gaming the 1080 Ti is clearly better than the Vega FE. For Productivity Vega FE is clearly better. Vega FE was marketed toward productivity and 1080 Ti was marketed toward gamers. Don't see what's wrong with that.

Yes, but the 1080ti does everything the titan xp does for $634 was the latest sale. According to yourself.

So fe really only has 5/8 of productivity benchmarks in one review.

That's not really that great.

Change the benchmarks and who knows.

It can't compete in power or gaming and only does well in a select number of benches.

It needs a lot more optimization or a massive price cut since it's just not competitive at $1000

 

[Bacon1]

Yes, but the 1080ti does everything the titan xp does for $634 was the latest sale. According to yourself.

So fe really only has 5/8 of productivity benchmarks in one review.

That's not really that great.

Change the benchmarks and who knows.

It can't compete in power or gaming and only does well in a select number of benches.

It needs a lot more optimization or a massive price cut since it's just not competitive at $1000

Wait for the RX version then. It will be basically as fast as frontier for productivity (heck maybe even faster if clocks are higher), will have better gaming performance and be cheaper. Its almost as if the FE was for productivity and RX was for gaming...

 

[tential]

Wait for the RX version then. It will be basically as fast as frontier for productivity (heck maybe even faster if clocks are higher), will have better gaming performance and be cheaper. Its almost as if the FE was for productivity and RX was for gaming...

If rx Vega is even better for productivity you've further proven my point... The fe has very little reason to exist

 

[raghu78]

Told you that compute performance per clock is higher in Vega than it is in Fiji.

In games it shows reduction per clock versus Fiji.

http://www.gamersnexus.net/guides/2977-vega-fe-vs-fury-x-at-same-clocks-ipc
Food for thought.

Thats a very useful comparison. The lowest perf increase at same clock is 15% in showcase-01 and the highest increase is 576% in snx-02 . Even after excluding snx-02 as an outlier we get an avg perf increase of slightly above 66% at same clocks and same sp count for Vega vs Fury X. this proves that the architecture is vastly better. We just need to wait for RX Vega launch to see how those improvements translate to gaming.

 

[Bacon1]

If rx Vega is even better for productivity you've further proven my point... The fe has very little reason to exist

It will have two sets of drivers, Pro and Gaming. If you want to say that the gaming only 1080 Ti drivers are fine, then the gaming only RX drivers should be as well right? But for those tasks that need the more accurate "Pro" drivers, FE will have them while RX will not.

 

[Malogeek]

Thats a very useful comparison. The lowest perf increase at same clock is 15% in showcase-01 and the highest increase is 576% in snx-02 . Even after excluding snx-02 as an outlier we get an avg perf increase of slightly above 66% at same clocks and same sp count for Vega vs Fury X. this proves that the architecture is vastly better. We just need to wait for RX Vega launch to see how those improvements translate to gaming.

That graph is only relevant to those applications/tests which is nothing to do with games. The other graphs indicate a zero architecture improvement for games at the same clock.

 

[Glo.]

That graph is only relevant to those applications/tests which is nothing to do with games. The other graphs indicate a zero architecture improvement for games at the same clock.

It shows that throughput of the cores is actually higher than in Fiji, which also shows that registry file sizes, and cache size has been increased, which should affect gaming performance, also.

What we see in games is that performance per clock vs Fiji - decreases.

It should not be this way.