First REAL FX Benchmarks

[Doomtrooper]

Thanks to nggalai for his link, we have some early benchmarks using the UT 2003 Platform limited engine by Thilo @ PCGH.


http://www.forum-3dcenter.org/vbull...hreadid=51204&perpage=20&pagenumber=1

UT 2003 Version 2166 Flyby
9700 Pro     Ti-4600     FX Ultra
201            170          196          @ 1.024x768, 32 Bit
110            65           143          @ 2X FSAA + 4:1 AF

UT 2003 Version 2166 Botmach

9700 Pro     Ti-4600     FX Ultra
71              71            68           @ 1.024x768, 32 Bit
65              65            65           @ 2X FSAA + 4:1 AF

 

[Slides]

Ouch

 

[jb]

thanks,

can not read German but interesting...btw any one notice LovesuckZ avatar?

 

[Ante P]

CPU limited botmatch results are pretty worthless
however it surely shows that it has some impressive AA/Aniso power
though I have a feeling that it won't do as good in higher AA modes since nVidia only have allowed the two sources who've published results to post 2x FSAA benchmarks

2x FSAA + 4x Aniso is pretty much what my Ti4600 can do in any game so those scores are just rubbish and uninteresting unless the higher modes follow the same trend of beating the R300

 

[Arun]

And for everyone's who's too lazy to conclude % between the Radeon 9700 Pro & the GFFX Ultra:

Flyby
1024x768, 32 Bit: -2.5%
2X FSAA + 4:1 AF: +30%


Botmatch: Err, what the heck? I'm sorry but the 2X FSAA scores are simply impossible. The Ti4600 cannot be that fast with 2X FSAA & 4:1 AF, it does more like 45FPS AFAIK... It simply gotta be a typo. The FX Ultra & 9700 Pro being on par there, however, seems possible.


Uttar

 

[Slides]

CPU limited botmatch results are pretty worthless

I know, I was commenting on the flyby. The FX better perform at higher res.

 

[Doomtrooper]

Not great at German myself..spent a couple of years over there in the Black Forest Area..very nice..only words I still remember are:

Zwei Hefewiezen

Hope the spelling is correct

 

[nggalai]

Hi there,

here's a copy/paste from my earlier post in the "GFFX pictures" thread:

***

Yup, that's the results. Test system is XP 2600+, 512 MB RAM, Nforce2 chipset.

Another interesting bit: BLOODY noisy fan, even in final production revision. Card clocks at 300/300MHz in 2D applications and 500/500MHz for 3D.

Again only 2xFSAA

For now. That was Thilo's very first benchmark run, he promised higher settings for both AF and FSAA for the review, as well as higher resolutions.

***

Thilo's off benchmarking and probably won't be back in the thread DT linked to for some time still. But feel free to leave your suggestions anyway.

ta,
-Sascha.rb

 

[Iceman]

@Doom

nah, it should be:

Zwei Hefeweizen

 

[nggalai]

Not great at German myself..spent a couple of years over there in the Black Forest Area..very nice..only words I still remember are:

Zwei Hefewiezen

Hope the spelling is correct

Close--it's Hefeweizen. But I observe you have gathered the most important piece of German there is, so well done.

ta,
-Sascha.rb

 

[Doomtrooper]

Bummer...I was close though.

 

[Doomtrooper]

however it surely shows that it has some impressive AA/Aniso power

At low sampling levels..8X AF and 4 X + AA is what I want to see.

 

[nggalai]

however it surely shows that it has some impressive AA/Aniso power

At low sampling levels..8X AF and 4 X + AA is what I want to see.

*nod*

I agree. After all, DK mentioned the GFFX was "optimised" for 4xMSAA performance.

What I really do hope for, mind, is sensible image quality comparisons. There's not much sense comparing 6x to 6x FSAA between the GFFX and Radeon9700 without losing a word on the quality differences, for example. Traditionally, PCGH has produced some of the more sensible reviews in German print mag history, but Thilo has to hand in his review by Monday, Tuesday morning the latest, poor sod.

Oh, well. We'll see soon enough, eh?

ta,
-Sascha.rb

 

[Prometheus]

Radeon 9700 has a 46% hit with 2xfsaa & 4xaniso!!!
Isn't that way large?
What kind of aniso is GFFX using?

 

[Doomtrooper]

UT brings a unusally high hit for AF....without IQ comparisons though and if the NV 30 is using OGMS vs RGMS..comparing 4X to 4X may not be accurate.

 

[Fred]

The relative performance there seems fairly reasonable, though im not sure if the actual scores are. One would expect driver revisions to correct CPU limited cases, and drive them closer to parity.

Likewise, moving to higher AA, might help the GF FX assuming its caches are optimized for 4*. Otoh, moving to higher res, will probably pack a bandwidth hit, so the 30% difference will diminish there in principle.

Higher rez + 4* AA?.. who knows. Wait a few weeks.

I'm wondering about UT as a benchmark as well. One might ask at what point are we going to start seeing CPU saturation take hold with the new card generation. O/C a R9700 might give an indication.

 

[T2k]

And for everyone's who's too lazy to conclude % between the Radeon 9700 Pro & the GFFX Ultra:

Flyby
1024x768, 32 Bit: -2.5%
2X FSAA + 4:1 AF: +30%


Botmatch: Err, what the heck? I'm sorry but the 2X FSAA scores are simply impossible. The Ti4600 cannot be that fast with 2X FSAA & 4:1 AF, it does more like 45FPS AFAIK... It simply gotta be a typo. The FX Ultra & 9700 Pro being on par there, however, seems possible.


Uttar

Uttar, start praying!

 

[KimB]

Btw, the botmatch scores are not only useless because they are CPU-limited, but they also are not anywhere close to as reliable as the flyby scores. Every time they run, they're just a little bit different. So, you really need to run multiple results and average to get good results, but they're not going to be CPU-limited, so it is of little consequence.

Regardless, I still want to see the 4x FSAA + 8-degree anisotropic results.

 

[Tahir2]

Botmatch scores may not be representative of online play as obviously the CPU is no longer crunching as much AI code as it is when playing in single player mode. Correct me if I am wrong.

 

[Arun]

Uttar, start praying!

Err, what for? Sorry, I just didn't understand


Uttar