ATI driver quality

[divVerent]

I benchmarked some test systems on Nexuiz 2.3's latest test version with different configurations... but see for yourself:




Source: http://mensa.ath.cx/~polzer/temp/benchresults.txt

As you see, for nvidia cards, there are no big driver performance differences between the operating systems - on Linux you basically get the same performance as on Windows. OTOH, look at the ATI drivers... 2.5x as slow on Linux than on Windows.

ATI, make working drivers!

[LordHavoc]

Not that I advocate the destruction of any ATI cards, but they have consistently proven unworthy on Linux, and as a dedicated Linux-based developer I have no interest in cards that are poorly supported on my preferred development platform.

I really hope AMD and ATI can craft some good GPL drivers for ATI cards on Linux, I'd switch to a good card with GPL drivers in an instant.

[Ed]

@div: Can we have some numbers please? Did the open source drivers not run the higher details tests at all? Any artifacts?

I know most people will automatically jump to looking at the low detail numbers (just because they're higher) but those will be the least reliable as they are too dependent upon other factors. The higher settings will give the more honest reflection of rendering performance. The only use of the lower resolutions is that as they are more CPU limited, they give, by comparison, indication of the CPU overhead of the driver.

[Psychcf]

I know this is off topic but what benchmarking program did you use?

[KadaverJack]

@div: Can we have some numbers please?

Look beneath the pictures, there's a link

I know this is off topic but what benchmarking program did you use?

Obviously some game called Nexuiz

[torus]

Obviously some game called Nexuiz

Where can I get that?

[Psychcf]

Obviously some game called Nexuiz

Where can I get that?

that and how much does it cost?

[divVerent]

I ran a timedemo on silvercity on my keyhunt server. However, I have good reasons to not hand out that demo file (mainly, a player called "P**** ** ** ***** *** **y").

[leileilol]

then hexedit it

[divVerent]

Would need to change all player names, including possible typos of them in the chat messages. That's much work... and any other timedemo on silvercity with about 10 to 12 players will yield the same relative differences.