X1900 XT perfromance in Quake 4

[titaniumx3]

This is the first OpenGL game I've tried on the X1900 and I was suprised to see that the framerate rarely goes above 60fps and can go as lows as 25fps, even though the benchmarks on various websites (with similiar specs and settings) indicate average framerates of ~100 fps!

My system specs:

A64 3500+ @ 2.2Ghz
1GB RAM
Radeon X1900 XT
Omega Drivers (6.5)

Running the game @ High Quality, 1280x1024, 4xAA, 16xAF, Vsync OFF, everything else on.

Is this normal?

 

[MeatLoaf]

16x AF is the culprit there i think. Try knocking it down to 8x or 4x and see what happens.

 

[titaniumx3]

I tried reducing the AF and even tried application controlled, still no difference...but I did turn off the AA and performance improved quite a bit, though I think the <60 fps is in-game limit. The wierd thing is that the slowdown to <30fps only occurs in certain areas, like in the very first level where there is a big fan casting a shadow on the floor. Maybe Quake doesn't support ATI anti-aliasing very well???

Anyone else with an X1900 card get similiar results?

 

[MeatLoaf]

Its the same with my 7900GT, running at 1440x900, hi quality, 4x AF and it never goes over 60fps, even with vsync off. And it drops to 30ish during intensive moments.

 

[Jimbo Mahoney]

The game is limited to 60fps and also is rather fond of a powerful CPU.

A timedemo won't be 60fps limited.

 

[speeduk]

The game is limited to 60fps and also is rather fond of a powerful CPU.

A timedemo won't be 60fps limited.

Yes this is true. I would upgrade the cpu or try and overclock it if you can.

 

[titaniumx3]

Wow, I didn't expect Quake 4 to be so CPU intensive, but in any case it seems the performance is normal for my system. According to this online review, Quake 4 can be pretty taxing even on top spec comps. I guess some benchmarks can be quite misleading.

I also tried the HardwareOC benchmark utility (for Quake 4 v1.2) using supplied demo @ 1280x1024, High Quality, 4xAA, 16xAF and get around 55fps. It would be cool if these benchmarks were standardised somehow.