New 3DMark05 screenie

[Neeyik]

If I remember correctly my GF3 gives me 1052 without nv's cheats .. and the 5200 gives in the region of 1700(I could be wrong but I think this is what I remember seeing). That's 700 points SM2.0 gives to 3DMark03.
US

Not quite. That's 700 points from having PS1.4 support for GT2 & 3 (reducing passes), two sided stencilling for GT2 & 3, better vertex throughput too; as well as SM2.0 support. Just remove the points from GT4 from the 5200 result and you'll be able to see exactly how much that last test is producing.

 

[Unknown Soldier]

I don't know Neeyik .. looking at this seems to show that GT4 produces alot of the final score.

Look at the X700XT and the GT6600.

Final score:
X700XT - 8200
GT6600 - 7831

GT1:
X700XT - 214.2
GT6600 - 221.4

GT2:
X700XT - 54.8
GT6600 - 57.4

GT3:
X700XT - 46.3
GT6600 - 49.9

GT4:
X700XT - 55.5
GT6600 - 45

The X700XT loses all of the GT's except for GT4. Yet the result is the X700XT having almost 400 points over the GT6600.

I think GT4 has something like 40% of the final score. GT2&3 25% each and GT1 10%. (someone please correct these percentages).

BTW Neeyik .. does the 5200 have PS1.4 support or PS1.1? Or let me rephrase .. when the 5200 runs 3DMark03 .. does it run PS1.4 or PS1.1? As far as i've understood .. it runs PS1.1. The R3xx runs PS1.4.

 

[Scali]

The X700XT loses all of the GT's except for GT4. Yet the result is the X700XT having almost 400 points over the GT6600.

The difference in GT4 is also by far the largest, relatively speaking.
I believe the actual formula is in one of the whitepapers available from www.futuremark.com

BTW Neeyik .. does the 5200 have PS1.4 support or PS1.1? Or let me rephrase .. when the 5200 runs 3DMark03 .. does it run PS1.4 or PS1.1? As far as i've understood .. it runs PS1.1. The R3xx runs PS1.4.

The 5200 supports 1.4. D3D demands that all hardware is backward-compatible. So if you support 2.0, you have to support everything below.
In the case of the FX series this was unfortunate, since they re-used the integer pipeline of the GF4 series, which only supported up to ps1.3. ps1.4 requires extra precision, so it can't be made to run on that pipeline. Instead, NV had to run ps1.4 on the ps2.0 units. I believe it even requires 32 bit float precision, because 16 bit doesn't offer enough precision for ps1.4's spec.
At any rate, ps1.4 is very slow on the FX series, and when you use a program like 3D Analyzer, you can force 3DMark03 to run in ps1.1, and notice it's actually faster that way.
I believe that there was a cheat in the drivers at one point that forced ps1.1 during GT2 and GT3. So in that case, the 5200 would run ps1.1, although it's not supposed to.
3DMark03 now includes an option to force ps1.1 in GT2 and GT3 anyway.

 

[Neeyik]

US - certainly GT4 offers plenty of points to the final score; what I was pointing out is that the 5200 should (in theory at least) gain additional benefits from being a DX9-level card, as well as being able to run the final game test. I ought to dig out my 5200 and see what the current state of play is, with regards to drivers.

 

[Unknown Soldier]

k

btw .. with my GF3, I got 1244 with 6167 Drivers.

Please have a look what you get with the first 3 GT.

Thx
US

 

[Fodder]

what I was pointing out is that the 5200 should (in theory at least) gain additional benefits from being a DX9-level card, as well as being able to run the final game test.

If it was any good at PS1.4 it would, but as Scali mentioned it unfortunately isn't.