A Look At CS:GO & TF2 On AMD GPUs With The Open-Source Driver

[phoronix]

Phoronix: A Look At CS:GO & TF2 On AMD GPUs With The Open-Source Driver

I'm in the process of running open-source RadeonSI Gallium3D tests on the new MSI Radeon R7 370 4G graphics card. That initial Linux review of the AMD Radeon R7 370 with the open-source driver will be published later this week (still waiting on an updated Catalyst driver for those proprietary driver tests). However, as an excerpt of the Gallium3D testing, here are results from some AMD graphics cards with Counter-Strike: Global Offensive and Team Fortress 2...

http://www.phoronix.com/scan.php?page=news_item&px=CSGO-TF2-370-Preview

[Kano]

Is tesselation active in the oibaf ppa for radeonsi? I would not consider this a valid test to compare against binary drivers if not. But there is no r600g tesselation for now at all, not even in the tesselation-review branch. I guess the the desktop Broadwell with 4rd level cache can beat the R7 370 with Source (2) engine.

[Rakl?dder]

The Gun Mettle Update is Team Fortress 2: Global Offensive on steroids!

[smitty3268]

Is tesselation active in the oibaf ppa for radeonsi? I would not consider this a valid test to compare against binary drivers if not. But there is no r600g tesselation for now at all, not even in the tesselation-review branch. I guess the the desktop Broadwell with 4rd level cache can beat the R7 370 with Source (2) engine.

Oibaf doesn't even have LLVM 3.7 support yet, so tesselation isn't anywhere close to being in.

[dungeon]

Oibaf doesn't even have LLVM 3.7 support yet.

He should really consider using it by now , in a week head will be llvm 3.8 branch

[mmstick]

Is tesselation active in the oibaf ppa for radeonsi? I would not consider this a valid test to compare against binary drivers if not. But there is no r600g tesselation for now at all, not even in the tesselation-review branch. I guess the the desktop Broadwell with 4rd level cache can beat the R7 370 with Source (2) engine.

The Source engine does not use tessellation and is therefore only D3D9/OG3, which the open source drivers fully support.

[haagch]

Do you people not see this?
http://openbenchmarking.org/embed.ph...ha=43656d0&p=2
The HD 7950 (!!!!) gets 52.5 fps in CSGO. Do you know how strong this GPU is?

I haven't seen much comparison, but here is september last year, which includes the HD 7950 with the proprietary driver:
http://www.phoronix.com/scan.php?pag...ux_blobs&num=3
142.5 fps
That's about 270% of the open source driver. That's a way bigger difference than what the difference typically is.

The R7 370 seems to suffer much too. At least in relation to TF2. Look at the R9 290. Almost no difference in CSGO and TF2. Then look at 7950 and 370: Almost double the performance in TF2.

Can it be due to SI/GCN 1.0 not working right at the moment? I have looked at wikipedia and for the R7 370 it says

June 18, 2015

Trinidad PRO(Cura?ao PRO)(Pitcairn PRO)

GCN 1.0

My HD 7970M is underperforming a lot in CSGO too and it is a PITCAIRN/SI/GCN 1.0 GPU. Also the rendering is very glitchy on my GPU. Can you look at the rendering of SI GPUs whether there is any flickering/other corruption?

Michael can you benchmark with windows too? People are saying they get 200+ FPS with the HD 7950 in CSGO on windows, but of course not with the same benchmark...

With the higher-end graphics cards, the current open-source AMD Linux stack does deliver a playable CS:GO experience on Ubuntu at 1080p.

Oh really? Have you actually tried playing CSGO with < 60 fps online? That's a very real impairment.

[dungeon]

Do you people not see this?
http://openbenchmarking.org/embed.ph...ha=43656d0&p=2
The HD 7950 (!!!!) gets 52.5 fps in CSGO. Do you know how strong this GPU is?

Yeah something very bad happen there on HD 7950, see those nouveau results where GTX 650 behave like that - similar fps rate on both, i guess you should be something like 150% faster then that nvidia card.

http://openbenchmarking.org/result/1...BE-NOUVEAU6801

[Kano]

@mmstick

I know that. But why would you want to compare Unigine Valley without tesselation against the same test with tesselation? That sounds somehow useless.

[tomtomme]

Yeah something very bad happen there on HD 7950, see those nouveau results where GTX 650 behave like that - similar fps rate on both, i guess you should be something like 150% faster then that nvidia card.

http://openbenchmarking.org/result/1...BE-NOUVEAU6801

at least better than GCN in 2013 where it did not work at all with source engine games:
http://www.phoronix.com/scan.php?pag...ing_gpus&num=3

I am still searching for an early gcn source engine benchmark, where it worked - can someone fine one?
#edit:
there we have it http://www.phoronix.com/scan.php?pag...on_12nov&num=3
same frame rate half a year ago in TF2 but with 1600p instead of 1080p for the HD7950 - thats a hell of a regression!