Tomb Raider: AOD DX9 Benchmarks

Dave Baumann

Dave Baumann

Gamerscore Wh...
Moderator
Legend
Reverend introduced a new benchmark to B3D recently, being Tomb Raider: Angel Of Darkness. The importance of this benchmark is that it is the first to full utilise PS2.0 DX9 functionality within a gaming environment.

To gain an understanding of the relative performances of DX9 boards we have a short article comparing the performances of the 5200 Ultra, 5600 Ultra, 9500, 9600 PRO and at the high end the 5800 Ultra, 5900 Ultra, 9700 PRO and 9800 PRO.

http://www.beyond3d.com/misc/traod_dx9perf/
 
err that was me that asked for benchies at same clock... oh and it might be intereesting to see how the nvidia cards fare at the same clock as the ati's, but i dont really care about them because i'm NOT bying another nV card til they get their act together o_O
 
If that isn't devastating for nvidia, I don't know what is...
Of course this confirms all (synthetic) ps2.0 tests we've seen, there's no denial the GFFX is just slow (compared to R3xx) at executing PS2.0 shaders, even if the code is optimized for its architecture. Definitely interesting to see the benchmark is indeed limited by PS2.0 performance (as evidenced by the 5200U vs. 5600 score when the former isn't running the PS2.0 shaders).
Though what I'm missing is does it actually look better if all the features are switched on? It would be nice to see some screenshots.
And I'd have liked if the 9600 would have been included instead of the 9500 - it has the advantages it's widely available, is cheaper, doesn't make any noise and it doesn't suck up a lot of power... And a 9000 included would also be nice - of course no full-featured mode, but it looks like you have to switch PS2.0 off anyway with the 5200 (even Ultra) to get playable performance.
Oh, and there is an error on page 2:
We can see that the 5800 Ultra, despite the clock speed advantage, is outperforming the 5900 Ultra, which is likely a factor of the bandwidth difference.
Click to expand...
 
mczak said:
Though what I'm missing is does it actually look better if all the features are switched on?
Click to expand...

The most noticable of all the effects is the PS2.0 DoF effect - you are getting to the point of diminishing returns woith many of the others, and I wouldn't know where to look to get differences for all the settings.

And I'd have liked if the 9600 would have been included instead of the 9500 - it has the advantages it's widely available, is cheaper, doesn't make any noise and it doesn't suck up a lot of power...
Click to expand...

I was looking for the "lowest of the low" DX9 board configuration from ATI, and the 9500 is it really - although the 9600 SE will soon take that place probably, but I don't have one.

And a 9000 included would also be nice - of course no full-featured mode, but it looks like you have to switch PS2.0 off anyway with the 5200 (even Ultra) to get playable performance.
Click to expand...

The point of the test was to limit it to DX9 boards.
 
In the forthcoming Albatron GFFX5900 (128MB, non-ultra) review scheduled for release tomorrow (Tuesday), there will be a couple of additional pages providing R9600PRO and R9700PRO benchies. This coming review will provide IQ fidelity of the 5900 vs refrast, Cg vs DX9 HLSL compiler (quite the difference) in terms of both performance and IQ and... er, well, just wait for the review.
 
Great job, Rev...

Great article, Reverend. :D I did not think that PS 2.0 would lack that badly on the 5800/5900 in a game. Anyone here at B3D would have known that after the synthetic tests, PS 2.0 was lacking on nV. However, I thought that some improvement (without cheating) would be made by nV. Consider this pure speculation, but I hope you don't make the list of "nV optimization target" if you know what I mean. ;) Once again, good job...and I look forward to your next article.
 
Fantastic article. Really shows beyond any sort of doubt how shoddy NVIDIA's PS support is.
 
Great work. Another example of why B3D is the best site of its kind. BTW any screens, forthcoming, to compare the default I.Q.'s ? Also is there a way to use non CG compiled shaders for nVidia cards. Curious to know how much benefit CG gives to them?
 
nelg said:
Also is there a way to use non CG compiled shaders for nVidia cards. Curious to know how much benefit CG gives to them?
Click to expand...
In the settings panel, you just uncheck the "Use Cg Compiler" box to use non-Cg compiled shaders. As far as I can tell, the check box is greyed out unless you have the Cg compiler installed anyway. I'm not 100% certain of this - I'm currently testing out some Asus V9950s and the option to select Cg compiled shaders is not available for me. I'll install the Cg SDK again in a bit once I've finished some tests that are currently running.
 
A number of questions posed here will be answered in the aforementioned Albatron GFFX5900 review mentioned above, including IQ comparisons questions and Cg.

Neeyik, you don't need to install anything from NVIDIA (SDK, toolkits, whatnots) wrt Cg to enable usage of the Cg compiler.
 
Reverend said:
Neeyik, you don't need to install anything from NVIDIA (SDK, toolkits, whatnots) wrt Cg to enable usage of the Cg compiler.
Click to expand...
Well that's odd then. Any thoughts as to why the checkbox is blanked out for me?
 
Neeyik said:
Reverend said:
Neeyik, you don't need to install anything from NVIDIA (SDK, toolkits, whatnots) wrt Cg to enable usage of the Cg compiler.
Click to expand...
Well that's odd then. Any thoughts as to why the checkbox is blanked out for me?
Click to expand...

because the game just shipped with a DX8 compiler
so you'll need to download Cg compiler 1.1 from www.nvidia.com
 
I thought that would be the case - fortunately I've already got the whole Cg caboodle on a disc buried somewhere. Even more fortunately is that I won't have to do the tests again; man, they take forever! ;)
 
Well it's really interesting to see what's going on between the top end cards where DX9 is concerned but I wasn't really suprised. What did suprise me though was the FX5600 which by all accounts is the product thats being portrayed as the card thats taking the DX9 market share and doesn't even perform any DX9 functions.

Seems more a job for trading standards!
 
You must log in or register to reply here.

Similar threads

Kaotik
Rise of the Tomb Raider, sketchy benchmarks
2
Replies
30
Views
6K
Clukos
Clukos
M
tomb raider benchmarks
Replies
1
Views
2K
muppy
M
T
HL2 perf. from anand
2
Replies
21
Views
3K
Dio
D
E
Valve is a liar. Cheating on Halflife2
2 3
Replies
59
Views
8K
Chris123234
C
D
It was only a matter of time
4 5 6
Replies
110
Views
15K
parhelia
P
Back
Top