A general thread for Q4 stuff would be cool.
-------------------------------------
Just running Quake4 on my new video card, have realised a few things:
In your Doom3 v Q4 article
, you say:
I'm not terribly authoritive, but have run many benches on various hardware this week.. Your Quake4 ceiling at around 49-53fps is CPU bound, not GPU. Without going into the wherefore, here's my results (where game is also cpu bound)
Video: 6800agp/128M/12pixel/5vertex/325MHz.
CPU: socket 754 sempron 3400+/512Meg DDR400 (at 2000MHz , and overclocked to 2150MHz)
Settings: High Quality 1280x1024x32, no AA
Demo:Quake4-1.3, playNetTimeDemo id_demo001
-------------------------------
Sempron 3400+ 2000 MHz, 51 fps
Sempron 3400+ (overclocked) 2150 MHz, 56 fps
Unlocking video to 16pipes/6vertex/350MHz
Produces no gain in FPS, except in 4xAF,4xAA, when my results go up from 38 fps (similar to the results of your 6800gt results) back to the CPU limit of 51 fps.
Windows Quake4 benchmark for this CPU is around 80-90 fps. Much higher than linux, probably due to linux port (and especially) SDL overhead.
Hmmm..
-------------------------------------
Just running Quake4 on my new video card, have realised a few things:
In your Doom3 v Q4 article
, you say:
Although no results can be decisively concluded due to the slight difference in testing demos, in general we've found Quake 4 to be more GPU demanding than CPU bound.
Video: 6800agp/128M/12pixel/5vertex/325MHz.
CPU: socket 754 sempron 3400+/512Meg DDR400 (at 2000MHz , and overclocked to 2150MHz)
Settings: High Quality 1280x1024x32, no AA
Demo:Quake4-1.3, playNetTimeDemo id_demo001
-------------------------------
Sempron 3400+ 2000 MHz, 51 fps
Sempron 3400+ (overclocked) 2150 MHz, 56 fps
Unlocking video to 16pipes/6vertex/350MHz
Produces no gain in FPS, except in 4xAF,4xAA, when my results go up from 38 fps (similar to the results of your 6800gt results) back to the CPU limit of 51 fps.
Windows Quake4 benchmark for this CPU is around 80-90 fps. Much higher than linux, probably due to linux port (and especially) SDL overhead.
Hmmm..
Comment