ATI Technologies explained today the optimisations they implemented in its CATALYST drivers in order to improve performance in 3DMark03 Game 4. An official from the company told us that ATI’s optimisations are really optimisations, but not cheats as we thought before.
In contrast to the rival, ATI Technologies’ drivers do not avoid any workload for the RADEON-based graphics cards. The final image quality is exactly as FutureMark intends, so, this more resembles optimisations, but not simple cheating, when the final image quality is lower than it should be and the graphics processor avoids certain workload. In fact, ATI rendered the 3DMark03 in such a way that allowed the R300 and R350 VPUs to perform 3 pixel shading instructions in parallel. According to ATI, this also works in real games as well when the Developer Relations department works with the developers of such games. In fact, it may be hard or impossible to perform such optimisations via compiler, ATI said, so, they need to work with the developers to achieve such efficiency.
Note that ATI decided to avoid such optimisations for 3DMark03 in future and will remove this one from the next version of CATALYST drivers.
Comments currently:
4
Discussion started: 05/28/03 11:32:42 AM
Latest comment: 05/29/03 05:48:30 AM
Expand all threads |
Collapse all threads
[1-2]
1.
there we go , i repeat ATI DID NOT CHEAT!
[
Posted by: Seabook
|
Date: 05/28/03 11:32:42 AM]
+ expand thread (2 answers)
- collapse thread
optmizing and cheating are 2 different things. what ati did was NOT cheating.
[
Posted by: Orion12
|
Date: 05/28/03 11:39:13 AM]
Given the fact that ATI is member of the special 3dMark Development team, they could have taken the time to verify the reaction first with FutureMark. I am not saying they cheated but I do find the way that the company handled this situation is questionable as had a little more work been done, many hassles may have been avoided.
[
Posted by: AJW
|
Date: 05/28/03 07:18:54 PM]
2.
ATI searches for the specific shader used in 3DMark03 and replaces it with one they've hand written. That's what Carmack labelled "grungy". You can't say it's definitely not a cheat, because they're no longer executing the same workload as other cards in the test, therefore how can you make a comparison? If they had a generic assembler that did this for all input shaders it would be a valid optimisation. But as it stands it only works for this one test and exists solely to increase 3DMark score which may trick people into buying their card thinking that better performance will translate over into games.
To say it's the same thing that happens with any game "if ATI works with the developers" is a lame cop-out - ATI did work with Futuremark, they're on the beta program. If they'd found an optimisation they should have told FM about it so that all cards could benefit, bot kept it to themselves.
[
Posted by: Myrmecophagavir
|
Date: 05/29/03 05:48:30 AM]
[1-2]
Make Home Page
Add to Favorites
XbitLabs is using Twitter
Windows Live Alerts
