Windows 10 vs. Ubuntu Linux With Radeon / GeForce GPUs On The Latest 2018 Drivers

[Vash63]

Wow, those CSGO results are really impressive. I've been playing CSGO on Linux since they updated to 64-bit (fixed my OOM crashing issues), I don't have it installed on Windows anymore so I didn't know it was slower there. As an Nvidia user I haven't seen any of the stuttering or frame pacing issues that others in the thread are talking about, those may be driver related (shader cache?).

[Marc Driftmeyer]

Biggest real takeaway: AMD's Windows drivers *really* suck at OpenGL.

Windows customers don't care about OpenGL. They care about DX11/DX12.

[Goddard]

I did my own crappy tests on an AMD R9 290 and definately noticed most things run better on Windows even though it bugs me. Simple things like just watching a video perform much better under Windows. Things are just smoother and very responsive compared to Linux. I had to switch to Windows on that system because it was severally gimped. Main system still runs Linux though, but it has a 1070 which seems to respond well no matter the OS and your graphs kind of suggest that. It would be interesting to see how things pan out going forward.

[andre30correia]

unity and linux destroying windows

[prazola]

People often says here CS:GO for Linux is shitty port, i am not quite sure... maybe it is - maybe it isn't. Anyhow, so to take user compliants seriosly i guess Valve and/or driver devs should take a look right there to improve that case if possible.

https://github.com/ValveSoftware/csg...ux/issues/1111

Is that really since 64bit client? At least they are watching on this, improving but also breaking it

64bit isn't the problem. The problem is the overhead of the port (this is not a port, it's just a translation layer) that got worse when valve updated CSGO hitboxes with new and more complex versions. This game needs a real opengl/vulkan renderer but developers are scared to touch the code of the game because it's a mess and they could break something.

Wow, those CSGO results are really impressive. I've been playing CSGO on Linux since they updated to 64-bit (fixed my OOM crashing issues), I don't have it installed on Windows anymore so I didn't know it was slower there. As an Nvidia user I haven't seen any of the stuttering or frame pacing issues that others in the thread are talking about, those may be driver related (shader cache?).

Nvidia drivers are better at multithreading so the togl can work well enough. A gtx285 is better than an r9 390x in CSGO_linux.

[monte84]

This is why I hate his testing methodology. From the article "Radeon GPUs Are Increasingly Competing With NVIDIA GPUs On Latest RadeonSI/RADV Drivers" March 19. He tested the RX 580 and GTX 1060 at 4k medium in Bioshcok and the results were pretty close. In this article he does 1920x1080 at high quality. There is now a 80 FPS difference between the GTX 1060 instead of 4. SO the article from March 19th, should really state "Radeon GPUs Are, in certain situations, Increasingly Competing With NVIDIA GPUs On Latest RadeonSI/RADV Drivers" 4K owners are small percentage, and even smaller 4k owners would be using an RX 580 or GTX 1060.

Now look at DOTA2, both articles are in 4k, OpenGL Numbers are a bit closer to the March 19th article, but instead of both being tied at 72FPS the RX 580 is at 70 and the 1060 at 78 now. Vulkan really shows some huge discrepancies. Now,at 4K still the RX 580 is getting 61 in this article, but was getting 72 on the march 9th article while the 1060 was more consistent at 78 march 19 to 80 in this article.

This is why i complained about consistency awhile ago. In the March 19th article he was using an i7 8700K and this article, an i9 7890XE with lower clock speeds. Most of the tech sites I read use a common system when testing graphics cards. How much of the discrepancy is related to the CPU difference?

[dungeon]

What you expect from methodology for Bioshock, when both RX 580 & GTX 1060 gives you about 60 fps on Windows @ 4K... so again about 60 fps for both

While on Linux (at least for Michael here) that is 90 to 100 fps on "same" high settings Is that Bioshock port so much better at 4K for Linux or something else happen there - something like that high settings is sort of fake on Linux, i could only guess what happened there no idea else really... This case probably require sort of visual comparison video to be maded before anything else

This is why i complained about consistency awhile ago. In the March 19th article he was using an i7 8700K and this article, an i9 7890XE with lower clock speeds. Most of the tech sites I read use a common system when testing graphics cards. How much of the discrepancy is related to the CPU difference?

Consistency is actually very easy to fix, by using one and only native resolution of your display for about anything and never ever use anything else - strict & fixed, like gaming consoles

There is nothing complex to explain there, either your are CPU or GPU bound, both happens sooner or later (but this can't be consistently explained exactly when) and that mostly depends on what resolution is used.

[haagch]

This game needs a real opengl/vulkan renderer but developers are scared to touch the code of the game because it's a mess and they could break something.

https://www.gamingonlinux.com/articles/counter-strike-global-offensive-is-moving-to-source-2.9498

[Tsih]

I did my own crappy tests on an AMD R9 290 and definately noticed most things run better on Windows even though it bugs me. Simple things like just watching a video perform much better under Windows. Things are just smoother and very responsive compared to Linux. I had to switch to Windows on that system because it was severally gimped. Main system still runs Linux though, but it has a 1070 which seems to respond well no matter the OS and your graphs kind of suggest that. It would be interesting to see how things pan out going forward.

This still bothers me, my PC performs well with R290 on Linux, better than with W10 when I last tested Windows couple of years ago. Desktop usage on Windows was fine, but in CS:GO there was microstuttering that doesn't appear with Linux. I can run CS:GO on 2K resolution with pretty much full graphics and still get 144fps+ on my Linux. Have been playing CS:GO on Linux without any problems for a year plus, that's after they fixed frame dropping that came with high resolution smokes.

[mitcoes]

Thanks a lot for this benchmarks, but I miss WINE benchmarks: If the gamer community notice WINE gaming (without antivirus) is faster than MS WOS bare metal gaming (with antivirus) and sometimes available GNU/Linux game versions are faster we will see a great increase in Steam Linux use.