OpenGL: Windows vs Linux

Hi all,
Not quite sure where this belongs so soz if I’ve put it in the wrong forum!

I was just wondering if there are any good resources out there that compare the performance of OpenGL systems on Windows systems vs Linux ones? Or failing that any opinions you might have on OpenGL’s performance?! Assuming that there is a difference :slight_smile:


My untested opinion is that you’ll get results that are similar on Linux to those on Windows where you have a driver that’s supported by the manufacturer.

Here’s an interesting article on driver architecture from NVIDIA for example and the shared codebase they use across platforms:

Here’s an interesting review of some cards on Linux with benchmark results but it’s a year old now and was using 2.2 GHz CPUs:

Compare with Windows results here:

Make sure you account for the different CPU speeds and compare the Viewperf 7 results not viewperf 6.1.2.

If you google for: viewperf results Linux

You’ll find a few results you can compare and they’re usually pretty close once you account for other differences like CPU.

on linux, graphic cards that are faster than others on windows, tend to get slower than those :wink: . At least, so does my GF 5900 XT. It is slower than the 5700 non-ultra I had before, you can find a thread about that in the linux forum.

What this means is that linux often is a step behind windows in terms of drivers. Performance and functionality itself is the same, which seems logical because in OpenGL, the graphics chip does most of the work, and it will not even know which OS the computer it is working in is running on.

One thing that might in fact affect performance in “real” programms (not demos but programs that also require some CPU work) ist that as far as I know, the linux c compiler (gcc) does not produce as fast code as the microsoft c++ compiler does, so this might make a difference.

it might have been true at some point in the past that gcc was slower than cl, but, my recent experience has been that it handily beats at least the msvc6 version. I took a set of custom cpu-limited benchmarks and ran gcc vs icc (intel on linux), and cl vs icl (intel on windows). The results were that icc produced instructions that performed on par with gcc, but icl handily beat cl by about 10-20% on average.

I haven’t run numbers on the msvc7 version, so they might have improved a bit, but it’s probably still in the same ballpark.

VC.NET optimization is significantly better than VC6, however I don’t consider that a valid issue when comparing GFX performance.

And of course gcc continues to improve.