Dota 2 Vulkan vs. OpenGL Performance Redux

[phoronix]

Phoronix: Dota 2 Vulkan vs. OpenGL Performance Redux

Earlier this week I published some Dota 2 Vulkan vs. OpenGL benchmarks with AMD Radeon and NVIDIA GeForce graphics cards under Linux. Since then I received some feedback from Valve with regards to Dota 2 on the Source 2 Engine testing along with a better demo to use for benchmarking and also using the latest Dota 2 Vulkan DLC updates. So here is a fresh look at the OpenGL vs. Vulkan performance for this popular Valve game on an assortment of NVIDIA and AMD graphics cards.

http://www.phoronix.com/vr.php?view=23223

[humbug]

Interesting. So it's when there is a lot going on with big battles, lots of units on screen, lots of work for the game engine etc is when Vulkan gains over openGL. In CPU limited areas.

Whereas the high res GPU limited areas still require more optimization, compared to mature openGL / directX.

[rycodge]

This is great to see. As an avid Dota 2 player and 7870 owner I'm just amazed at how well Dota2 runs with open drivers. I get a rock solid 60 fps (v-synced). I used to cry when I'd install fglrx on my old r600 card and attempt to game on Linux. It's truly amazing to see how far AMD has come when you can just install ubuntu 16.04, install Steam and play Dota2 at 60 fps.

Looking forward to seeing AMD's benchmarks with their new driver model and Vulkan as time goes on.

[Evil Penguin]

Michael, was the 120 FPS cap not removed during the benchmarking? "+fps_max 0"

[humbug]

This is great to see. As an avid Dota 2 player and 7870 owner I'm just amazed at how well Dota2 runs with open drivers. I get a rock solid 60 fps (v-synced). I used to cry when I'd install fglrx on my old r600 card and attempt to game on Linux. It's truly amazing to see how far AMD has come when you can just install ubuntu 16.04, install Steam and play Dota2 at 60 fps.

Looking forward to seeing AMD's benchmarks with their new driver model and Vulkan as time goes on.

Yep although Vulkan looks superior here at least AMD's openGL performance is respectable and smooth on both drivers.

Whereas there are many games which support only Nvidia, and even say so officially. Whether the fault lies with the development approach or with AMD's openGL I don't know.

[Michael]

Michael, was the 120 FPS cap not removed during the benchmarking? "+fps_max 0"

fps_max was 0 during the testing, it's enforced by the test profile.

[tuke81]

Great review. Have you planned to do any CPU scaling tests? Something like one discrete graphics cards from both vendors and opengl vs vulkan tests with 1,2,4,6 and 8 threads enabled on one cpu. Or couple of cpus starting from celeron/am1 apu to more stronger ones. Vulkan should benefit users with weak cpus(well maybe not so much against ogl than it would be against dx11).

[bug77]

God forbid you used the same scale on both CPU usage graphs, cause then it would be easier to, you know, compare...

[Michael]

God forbid you used the same scale on both CPU usage graphs, cause then it would be easier to, you know, compare...

Patches welcome... PTS is open source. PTS came up with it based upon the available data independently, but could come up with it to try to make it match what other graphs in same result file are doing.

[Michael]

Great review. Have you planned to do any CPU scaling tests? Something like one discrete graphics cards from both vendors and opengl vs vulkan tests with 1,2,4,6 and 8 threads enabled on one cpu. Or couple of cpus starting from celeron/am1 apu to more stronger ones. Vulkan should benefit users with weak cpus(well maybe not so much against ogl than it would be against dx11).

If enough premium members request such tests.