4K AMD/NVIDIA High-End GPU Comparison On SteamOS Linux

[dcc24]

No it doesn't use 16 threads, but the frequency of that CPU is high enough to not be the bottleneck. Also CPU usage is one of qualities of a driver, so even if there is a bottleneck AMD driver's handles it better in this specific test.

As a long time community server operator I can tell you that TF2, on even a cheap $80-$100 GPU, is completely bottlenecked by the CPU. You can test this yourself with a cheap GPU; just downclock/overclock your CPU and your FPS will change accordingly. At some point, you'll reach the Source1 FPS cap of 300 and you'll barely have ~10% GPU load.

[wizard69]

Michael this is an interesting article but it brought something to mind, 4K video cards aren't of much use unless you have a 4K display. Since 4K TV's are widely available these days I was thinking that testing a few of those monitors/TV's would be in order. The reason to test TV's over discreet 4K monitors is due to the idea that we are testing hardware in the context of gaming.

Now I know full well that testing monitors can end up being very hardware intensive if you do a complete job. However maybe a bit of subjective testing and minor inspections would sort good from bad. Considering the ability of TV's to act as quality monitors varies widely you may be able to weed out the obviously bad examples.

Just an idea here.

[theriddick]

Nice benchmarks, I look forward to the next round with Kernel 4.3. I actually run a 390x and game at 4k, although lately less so under Linux due to obvious reasons. Looks like we should all be playing Team Fortress LOL. I hope Valve manages to make progress with AMD videocard performance and SteamOS. It is actually looking up for Linux gaming, and 4k Linux Gaming. (AMD drivers still terrible for allot of titles however)

There is a Dying Light demo for Linux available atm, can that be benchmarked also? I don't know if it has any automated system or not (contact dev?) but it sure gives videocards a run for their money!

wizard69 you can run in super resolution mode on 1080p screens. So 4k is accessible to all, however I dunno if it works under Linux just yet.

[Azpegath]

Did you read the article? End of article or end of first page?

One of the next Steam Linux tests planned for this week would be comparing these SteamOS benchmarks with the proprietary graphics drivers to the same system running Ubuntu 15.10 while upgrading to the Linux 4.3 kernel, Mesa 11.1-devel Git, and LLVM 3.8 SVN.

Haha, sorry Michael, I just quickly browsed through the article and read the text inbetween the tests. I'm at work and just had time to glance at it for half a minute or so. My bad!

[andre30correia]

tf2 have cpu problem nothing more than that, and like is normal the opengl catalyst performance is horrible so much users saying miracles about amd ate the end is always the same thing bugs, bad performance and bad support nothing change and when vulkan arrives it will be the same story

[profoundWHALE]

tf2 have cpu problem nothing more than that, and like is normal the opengl catalyst performance is horrible so much users saying miracles about amd ate the end is always the same thing bugs, bad performance and bad support nothing change and when vulkan arrives it will be the same story

Having AMD's GPUs performing well is good, considering how bad they've been in the past. However, seeing them beat Nvidia at any game, let alone a popular Source engine game is praiseworthy. Nvidia is currently the reigning champion. Seeing them beaten even a little bit is refreshing after all this time.

[profoundWHALE]

Considering the ability of TV's to act as quality monitors varies widely you may be able to weed out the obviously bad examples.

http://www.displaylag.com/display-database/

http://www.rtings.com/tv/reviews/by-test-results/input-lag

[Daktyl198]

tf2 have cpu problem nothing more than that, and like is normal the opengl catalyst performance is horrible so much users saying miracles about amd ate the end is always the same thing bugs, bad performance and bad support nothing change and when vulkan arrives it will be the same story

I'm pretty sure DX12 and Vulkan are going to be different stories, actually. From the current AMD driver devs, we've heard nothing other than how they're afraid to touch anything of the current disgusting mess that is the Catalyst driver... well DX12/Vulkan (or any new API really) gives them the chance to throw all of that off and start (somewhat) fresh. Especially on Linux, with the new AMDGPU kernel-side component and we've got ourselves a ton of brand new code being written.

[edoantonioco]

on a laptop using r600, around september 2013 (I know it since I took a screenshot of the game claiming good performance) I used mesa-git and I was able to run tf2 at full graphics (with r600 backend and radeon dpm enabled if I recall correctly), today I cant even play it on linux at minimum settings, regressions sucks. I'm sure than the amd free driver can still offer more performance.

[Vassili]

This review and the other one about SteamOS are really high quality reviews! Not only that, but I think there is great interest in these kind of reviews... more than the average review on this site, I've seen them relinked on other big site's as well.

If I might give a recommendation, make these graphs:

http://openbenchmarking.org/embed.php?i=1510257-HA-4KSTEAMOS33&sha=2dc7523&p=2

But then instead of power consumption, with minimal FPS, 50% average FPS, average FPS and maximal FPS.