3dmark 2003 9700 vs. GFFX actual results here

[***CENSORED***]

From [H]ard OCP:

http://www.hardocp.com/images/articles/1044928685ymTd7LNuhM_4_4_l.gif

[intercede007]

Wrong forum. Volt? :D

[poursoul]

WOW!

wow.

what cpu were they using for the gffx system? *notes 666 marks*

low ps perfomance. hmm.

[intercede007]

Good point. Not much of a comparison if they aren't using the same setup, is it?

[volt]

CENSORED: some software are benchmarks :D

[ReDeeMeR]

OMG, PS and VS are slower on FX!!! WTF!?
And P&VS arent limited by other HW are they?

Man this is gettin worse by day.

[ALobpreis]

According to this, they are both using the same CPU!! :eek:

Asus A7N8X, AthlonXP 3000+ @ 2.5Ghz (14x180Mhz), 512MB Kingston HyperX PC3500, Maxtor 40GB HDD, Windows XP w/SP1, DirectX 9.0

:confused: :confused:

[Nutty]

Still came out top tho! Even with its broken/unoptimized pixel shaders.

[Aglarond]

As you may have noticed the the GFFX got a big boost in preformance with the new drivers but no boost in the synthetic tests. It is therefore possible t´hat Nvidia could hawe done some optimizing (ie usefp processing quality for the benchmark)Yust like Cormack sad he will do For the FX with doom. The FX is usseless at high internal precission acording to Carmack its half of what the Radeon Currently has at 24.So maybe Nvidia forced 16 for older PS progs and uses 32 only for the pS 2.0 things.It would be great since not all shaders need the highest percision.The ATI card for example always uses 24 so this could be a good sign for the FX also I dont think that things will improve that much for the fx well back to my DX7 card!!:(

[gmontem]

Wow. The R9700 got a beating in the multi-texturing fill rate test but aced the Pixel Shader 2.0 test big time. Do they tell you how the 3DMark score is calculated at all?

[jnd3]

Note that the multitexturing fillrate on the FX is 44% higher than the 9700. Too bad the rest of it (PS/VS) seems to be hobbled. They've certainly got something in the drivers that will enable a big boost at a later time. Certainly... :rolleyes:

Cheers,
JND

[Sazar]

check this out...

"old" drivers

http://www.hardocp.com/image.html?image=MTA0NDkyODY4NXltVGQ3TE51aE1fNF8xX 2wuZ2lm

"new" drivers

http://www.hardocp.com/image.html?image=MTA0NDkyODY4NXltVGQ3TE51aE1fNF80X 2wuZ2lm

notice the optimizations from the "new" drivers mainly affect the scores in the games that are being counted towards the 3dmark total ?

:)

I'll let the driver "optimizations" speak for themselves...

[The Baron]

Saz, stop with your conspiracy theories--it's because the shaders on the NV30 were completely unoptimized (like, DUH--new architecture, COMPLETELY new shader architecture)... so yes, fixing the shaders IS going to cause a rather absurd improvement in a benchmark where half of the tests are shader tests. ;)

[Sazar]

conspiracy theories ?

when have I started any conspiracy theories :)

thats pretty much the point Kyle was raising as well when this was posted on the hardocp site...

btw the hit on the cpu seems to be higher when using the gf FX than with the 9700pro... hence the lower cpu score for the gf FX...

the test for cpu still has some links to the gpu apparently...

per what I been reading...

[Solomon]

Does anyone really care about 03? I would like to see the driver optmizations in the games that are sitting on store shevles waiting for my money? I can careless about "optimizing" for a stand alone benchmark that you don't play anyways. Am I alone on this situation?

It's nice to see that the drivers are improving the scores. But are they improving in the games? It would be interesting to see [H] re-run some games with these new drivers (or are they the same used in the previews of the FX) to see if the drivers did anything to the games we play?

I for one don't care if the optimizations show up for only a benchmark program. I guess that's my two cents on that one! Hehe.

Regards,
D. Solomon Jr.
*********.com

[Gator]

3dmark performance boost from driver update! Woohoo! Now if only the driver update would fix everything else and make the GFFX a magically happy card form a afar away land
:rolleyes: :lol2:

[Pafet]

nVidia is going DOWn!!!

j/k :D

[The_KELRaTH]

I don't really think that forcing 12/16bit integer precision to get a benchmark boost is really doing Nvidia much of a favour. After all, it's Nvidia thats been promoting the NV30 as "CineFX" for the last year yet in reality it runs awful in 32bit precision and PS2.0

The multitexture fill rate is just showing what is known already - the NV30 Ultra runs at 500Mhz v 325.
MT is much less important though, as the more Pixel Shaders are used the more single texture performance becomes more important.

[ragejg]

What may help in this confused situation is another series of benchmark tests, this time with both cardz on an ordinary XP2000+ on a kt266a/kt333 mobo with un-optimized PC2100 RAM and AGP4X... Benchmark results for:

3DMark2000
3DMark2001
3DMark2003

...should be posted and compared... Yes 3dMark2000 has relevance here because of its *other* algorythyms(sp) for fillrate and texturing... I'd almost go as far to say that an ideal set of test results would be an AVERAGE of the 3 tests for both cardz...

Games nowadays are still based in DX7 code... and now (finally) use quite a bit of DX8... And it's not known yet how well this card (or the ATI DX9 cardz) will perform in DX9-heavy games... It could be different than the first DX8 card, the GF3, which still can run heavy DX8 games without much problem (esp. on a smokin system)...

An average of those tests would benifit many, IMO. And runnin @ agp4X/266fsb would keep the cards churnin a little more, being less cpu dependant...

now, I'm only on my first cup o java yet, so if something's not relevant in this post, be a nice guy and just poke me with a stick... I don't have my flame suit on *tee hee*...

:D

[Smokey]

Originally posted by ragejg
What may help in this confused situation is another series of benchmark tests, this time with both cardz on an ordinary XP2000+ on a kt266a/kt333 mobo with un-optimized PC2100 RAM and AGP4X... Benchmark results for:

3DMark2000
3DMark2001
3DMark2003

...should be posted and compared... Yes 3dMark2000 has relevance here because of its *other* algorythyms(sp) for fillrate and texturing... I'd almost go as far to say that an ideal set of test results would be an AVERAGE of the 3 tests for both cardz...

Games nowadays are still based in DX7 code... and now (finally) use quite a bit of DX8... And it's not known yet how well this card (or the ATI DX9 cardz) will perform in DX9-heavy games... It could be different than the first DX8 card, the GF3, which still can run heavy DX8 games without much problem (esp. on a smokin system)...

An average of those tests would benifit many, IMO. And runnin @ agp4X/266fsb would keep the cards churnin a little more, being less cpu dependant...

now, I'm only on my first cup o java yet, so if something's not relevant in this post, be a nice guy and just poke me with a stick... I don't have my flame suit on *tee hee*...

:D

I cant run 3dmark2000 anymore with newer drivers :(