Dx 8 cards in 3dmark2003 - 330 and Villagemark

A

Anon Lamer

Newcomer
Radeon 9100 at 250/250 = 1100 3dmarks with 148 cpu score.
Gf 4200 at 250/250 = 1110 3dmarks with 210 cpu score.

Radeon 9100 at 250/250 in Villagemark at 1024x768x32 = 83 fps
GF 4200 at 250/250 in Villagemark at 1024x768x32 = 99 fps WTF?

Villagemark does not appear to be very cpu intensive..... My guess is that
Nivida has corrupted *all* major and minor benchmarks. If futuremark wants to make guerilla warfare against Nvidia, then they should release a new patch for every detonator upgrade. Would it not be sweet if Nivida promised "boost" turned out to be nothing?

If they do introduce a test with partial precison, they should just give it a weight of zero so it wont corrupt anything...
 
I thought it was pretty well accepted that the z-reject and of the GF4 was more efficient than the R200?
 
Anon Lamer said:
Radeon 9100 at 250/250 = 1100 3dmarks with 148 cpu score.
Gf 4200 at 250/250 = 1110 3dmarks with 210 cpu score.

Radeon 9100 at 250/250 in Villagemark at 1024x768x32 = 83 fps
GF 4200 at 250/250 in Villagemark at 1024x768x32 = 99 fps WTF?

Villagemark does not appear to be very cpu intensive..... My guess is that
Nivida has corrupted *all* major and minor benchmarks.
Click to expand...
How do you get to that conclusion? Because R9100 and Ti4200 are very close in 3DMark03, they should be very close in any other benchmark, and if not, NVidia is corrupting benchmarks???

GF4 Ti4200 has always been faster than R9100 at equal clock speeds in most benchmarks (except those with AF only), but R9100 profits from PS1.4 support in 3DMark03.
 
I'm sure some NV25 user can make a simple experiment and run Villagemark with and w/o T&L 8)
 
Ailuros said:
I'm sure some NV25 user can make a simple experiment and run Villagemark with and w/o T&L 8)
Click to expand...

What would that change? The R200 has T&L too.

I guess 3dmark2003 has no test that checks for z occlusion ability - would it not be a good idea to have such a test included?
 
Anon Lamer said:
I guess 3dmark2003 has no test that checks for z occlusion ability - would it not be a good idea to have such a test included?
Click to expand...
You don't need a specific test to test for occlusion culling: The tests that are there can benefit from such features already.
 
A specific synthetic feature test for hidden surface removal would be nice though... IMHO 8) Afterall Villagemark is really starting to show its age but suprisingly still does an excellent job.
 
Kristof said:
A specific synthetic feature test for hidden surface removal would be nice though... IMHO 8) Afterall Villagemark is really starting to show its age but suprisingly still does an excellent job.
Click to expand...
I don't see it as that big of a deal. Either you have it, or you don't. Either way, your benchmark scores will be affected accordingly. If card A is faster than card B, does it matter whether or not card A's occlusion culling (if it has any) is better than card B's?
 
Well, it's an interesting piece of information, like other syntetic tests measuring for instance transform rate or fillrate.
 
Interesting only for a geek, Humus... ;)

Like Kyro2 kicked GF2 butt, it had less geometry processing capability and fillrate, but better real-world performance. What does it really matter if card X do Y better if it's still slower than card Z? Well the fanbois can feel better about something, but nothing real changes...


*G*
 
What would that change? The R200 has T&L too.
Click to expand...

I´m not sure if the scores change when disabling T&L on Radeons, they do though on the NV25, hence the remark.

OGLguy,

Agreed on your point.

3DCenter played a bit around with theoretical benchmarks. I´d say that the R3xx´s do exeptionally well in the Overdraw Z Reject Tester:

http://www.3dcenter.de/artikel/2003/06-05_b.php

;)
 
You must log in or register to reply here.
Back
Top