Extra Steam Linux Gaming Benchmarks Of The AMD Radeon RX 480

[phoronix]

Phoronix: Extra Steam Linux Gaming Benchmarks Of The AMD Radeon RX 480

Our just-published 12-page AMD Radeon RX 480 Linux review features a wealth of interesting OpenGL/OpenCL/Vulkan Linux benchmarks along with performance-per-Watt and performance-per-dollar metrics and more. Check it out if you haven't already. This article are just some extra Steam Linux gaming benchmarks for the RX 480 vs. various NVIDIA graphics cards in the games that can't be fully-automated like Tomb Raider, Shadow of Mordor, and Company of Heroes 2 where the benchmarks had to be manually executed...

http://www.phoronix.com/scan.php?page=news_item&px=RX480-Extra-Linux-Games

[Ehvis]

From the Windows benchmarks it appears that the RX480 should be in the same range as the GTX970. AMDGPU PRO isn't that far off, so I it's looking pretty good.

[grigi]

Hmm, does anyone know why the Shadow of Mordor performance appears out-of-line for the Mesa drivers compared to most other games?
I get better perf/smoothness with the Mesa driver on Tomb Raider/Shadow Warrior/Kotor2, but Shadow of Mordor... ugh.
FYI: I'm running mesa/llvm git (it is actually solidly stable)

[Nobu]

grigi Do you have AA enabled? Shadows and AA hit pretty hard in SoM.

[grigi]

Nobu Nope, never doing AA, my card can only do that on ancient games. it is a Mobility 7770 (512Shader-units 675Mhz 1GB) typically I try run at 1600x900, because that still has good clarity on a 15" screen. 1280x720 seems very blurry (so if the game isn't acceptable at 1900x600 in perf, I don't play it)

[Veerappan]

I feel like I saw some recent patches on the mesa list related to improving compiled code of a compute shader for SoM in either llvm or the GLSL compiler. The shader in question improved by something like 8x in performance. I wish I could find the patch/bug, but my searches all seem to be failing.

I've been seeing a lot of performance optimizations from Marek/Nicolai in the AMD camp recently, and I wouldn't be surprised to see this cleared up if it's a radeon-specific issue. If it's in Mesa Core, then it's possible that the Intel/Collabora/Igalia guys might also take care of it.

[bridgman]

Was that the change related to Bioshock Infinite perhaps ? That's the only one I remember seeing...

... but as usual would be happier if I were wrong

[grigi]

I know there was some async copy, and later the off-chip tesselation improvements that significantly sped up Shadows of Mordor on radeonsi, I haven't seen anything else.

What I mean is that I am starting to get used to radeonsi performing more-or-less at the closed-source driver in terms of overall performance, that seeing SoM/F1 stood out like a sore thumb a bit. The original prediction of ~70% of closed-source performance was way off, as the current reality is considerably better, most of the time.

Is it that those games are using new features that now _works_ but hasn't been optimised yet, or are they just very Nvidia-centric ports and won't ever perform well on anything that isn't the Nvidia closed-source driver?

[Veerappan]

Was that the change related to Bioshock Infinite perhaps ? That's the only one I remember seeing...

... but as usual would be happier if I were wrong

Yeah, I'm probably thinking of that one. Too many performance-related improvements lately to keep track of them all

[Aiua]

Hi Michael, it would be nice to have the benchmarks for TR and SoM in 1080p/high too. Bioshock Infinite benchmark would be nice too

Anyway, thanks for all the articles