TechReport chimes in on the HL2 benchmark

banksie said:
bdmosky said:
Additionally, I've yet to see how the 51.75 series drivers affects IQ in HalfLife 2, if at all. My question is, would a loss in precision in any way affect the variation of colors displayed?

Yes.

We saw this with 3Dmark03 where shader replacement of the sky texture for Game Test 4 reduced the sky from a 'sunny day with clouds' to a much duller 'overcast day' look. The colour palette was much more muted and it was one of the giveaways that something dubious was going on.

Indeed. If we look at the PNG screenshots available here

http://www.tech-report.com/reviews/2003q2/geforcefx-5800ultra/index.x?pg=12

and check the number of unique color values in each screenshots they are:

Radeon 159539
Nvidia 43.00 (no app-specific cheats) 174286
Nvidia 43.45 (the "dark sky" cheats) 126740

A clear difference in color range due to reduced precision.

(For some reason the 43.00 shot is 20 frames later than the others, but it really does not matter here.)
 
That's what I thought... although in your previous post, you got these results:

Bolloxoid said:
With an image viewer program you can check the number of unique color values in each screenshot. The results from 3DGPU's PNG files:

ATI 83029
NV 45.33 75430
NV 51.71 75754

So it would clearly seem to be about precision/variation, not just about gamma. This gives some quantitative support to the claim that Ati's output is more vibrant.

This lends evidence towards the fact that the 51.75 drivers are not degrading quality any more than what the 45.33 displayed (and actually show more color variation than the older drivers). What's interesting to me is how much variation the 43.00 drivers displayed in 3DMark compared to the Radeons. Perhaps whatever degredation is evident in the 45.33 is still in the 51.75 drivers and performance is increased because of different reasons than precision adjustments.

*edit* I just wanted to note that my wording above probably sucked a little. Quality is obviously degraded from 45.33 to 51.75, I'm just saying that pehaps this degradation isn't caused by precision adjustments but something else.
 
Back
Top