AMD Catalyst A.I. Useless Under Linux?

[mitcoes]

more resolutions please

Thanks, i do like a lot this kind of articles but ...

I do like at gaming benchmarks more resolutions as you usually do.

And of course benchmark VS other OSs, recently one in Tom's Hardware showed that ATI performs better in linux than MS WOS, and Nvidia has better drivers for MS WOS than for linux.

Also I read a comment that said that Linux kernel must be optimized first for gaming.

"Make a preemptible Ubuntu kernel with a 300Hz/1000Hz tick ratio (or CK\'s BFS) to see what\'s good for gaming and encoding/decoding tasks"

Perhaps a "normal kernel" vs this "games kernel" would be a good benchmark and of course a way of doing it - I do not know how to compile this kind of kernel. And i am sure that it will also improve compiz Mutter and other composite windows, and of course future Wayland, and Web browsers, and flash games and videos, that will be good for Linux desktop home users and QuakeLive players.

i would like a lot a Quake live test with differnt resoultions and full vs pro - low - configs specially 800x600 120 Hz and 120 fps as limit at config.

[Rallos Zek]

Aren't the 7950 and 7970 the only card built on GCN? The 77** are Southern Island design and it was built on VLIW4.

Nope anything 76xx and under will be VLIW, 77xx and up are all GCN...

[Sonadow]

AFAIK Catalyst A.I. mainly activates some shader replacements for specific games which run faster or fix bugs compared to the original shader programs delivered with the game. Therefore it is nearly useless to test 2 benchmarks and 1 game with moderate graphics quality.
Maybe you see some differences when it is tested against popular windows games which proivde a Linux binary such as Doom 3 or UT 2004 - but maybe they are also just too old. And as there are nearly no new Linux games with high-end graphics (only Oil Rush AFAIK), this feature is actually useless but not because of the fact that Nexius doesn't run much faster.

Seconded...the test should ideally be done by getting baseline results for the same games under Linux and Windows with Catalyst AI disabled first, and then redone with Catalyst AI enabled.

The difference in performance between the Windows and Linux versions will then be an indicator as to how effective the AI works under different OSes.

Also, having come from a Windows evironment some 5 years ago before making the switch, I can safely say that some games exhibit issues under Catalyst AI, but run fine when it is disabled. So AMD's reputation for subpar drivers is rather well known among the users in every OS's camp.

[GreatEmerald]

Try Doom3. The original id Software client.
Run a timedemo with Catalyst AI disabled, enabled, and after renaming the executable (doom.x86).
You get *much* more FPS with AI enabled and the original executable name.
So obviously at least part of AI are hacks that are program-specific and that are enabled by checking the executable name.

Well yes, that's the whole idea of Catalyst AI, it's program-specific workarounds. That is rather plainly stated on the Catalyst AI tab. It's a rather nifty idea - instead of hardcoding hacks, have them be optional.

Advanced also creates major problems in Warcraft 3 in wine.

I didn't ever see noticeable improvement anyway...

That raises a fair point - does it work on Windows games running through Wine?

[ChrisXY]

It would be good if we at least had a list of applications that are supposed to be "improved" by Catalyst AI.

Also, if Catalyst AI is supposed to only "improve" such games as the original doom3, then why would it try to interfere with the benchmarked games in the article?

[Nasenbaer]

Also, having come from a Windows evironment some 5 years ago before making the switch, I can safely say that some games exhibit issues under Catalyst AI, but run fine when it is disabled. So AMD's reputation for subpar drivers is rather well known among the users in every OS's camp.

Sometimes it's also vice versa. Games do not work well without the fixes integrated by Catalyst A.I.

[Ibidem]

It would be good if we at least had a list of applications that are supposed to be "improved" by Catalyst AI.

Also, if Catalyst AI is supposed to only "improve" such games as the original doom3, then why would it try to interfere with the benchmarked games in the article?

Last I knew, you didn't get checks for free.
It has to check the programs, and that imposes overhead automatically.