Discrepancies between review sites

[mjtdevries]

While most reviews seem to agree that the XT is slightly faster than the 6800 Ultra, I did notice some interesting differences between some sites.

For Example Halo benchmarks:
At Tom's the 6800U wins. Same result at Xbitlabs. But at Anandtech the XT wins.

Call of Duty: The 6800U wins at Tom's, but even the X800 Pro beats the 6800U at Xbitlabs. That is a huge difference. Does anybody have any idea what could be causing that. (other then the review site simply lying about the numbers, which I don't believe)

Tom uses a P4 and Xbit and Anand both use Athlon 64. So the difference between the Halo results at Xbit and Anand don't seem to be caused by the CPU.

 

[Veridian3]

Different levels used....possibly different detail settings ingame too.

 

[Diplo]

Most games don't have a standard benchmark map to run, meaning you can get wildly different results depending on what part of the game is benchmarked. For instance, indoor levels in 'Far Cry' are more dependent on per-pixel lighting than outdoor ones, but outdoor levels often have a much higher poly-count. Also whether the games has been patched can make a difference, and some patches introduce different paths for different video cards etc.

Other factors might be whether a 'fly-by' is used for a benchmark (where there's no sound, AI or other factors effected by CPU) or whether it's an actual level with full AI, sound etc. Then resolution, anti-aliasing, filtering, detail settings all make a difference.

 

[Dave Baumann]

Check waht drivers are being used as well.

 

[Atomahawk]

Check waht drivers are being used as well.

Ya to bad we have to worry, I mean check that!

 

[jimmyjames123]

Check waht drivers are being used as well.

Forceware 61.11 improves performance noticeably for FarCry, but as far as I know, doesn't really improve performance noticeably for most other games. In fact, in some games, the performance goes down vs version 60.72. However, apparently version 61.11 does not allow for disabling of trilinear optimizations.

The discrepancies could be due to many different reasons: difference in system hardware, resolutions tested, AA/AF levels tested, different custom demos tested, trilinear optimizations turned on or off in the control panel using version 60.72, whether or not version 61.11 was used, whether or not anisotropic filtering was set via control panel or via application for the ATI cards, etc.

 

[jb]

Forceware 61.11 improves performance noticeably for FarCry, but as far as I know, doesn't really improve performance noticeably for most other games.

However 61.11 introduce a number of bugs as well to Farcry. Granted some may not effect frame rates. Thus its not a good idea to compare the two scores unitl the bugs get fixed any case there is any change (fog could be one).

BTW I find it a bit ironic that for a "TWIMTBP" game, why they have some many issues with NV cards as I thought part of the deal was to make games run better on certian IHV hardware