FX 5200 benchmark revised

[DOOM III]

this is not a official benchmark. it's done by one of my friend,just have a look:

3d mark2001 SE

under ABIT BH7 3.06GHz 1GB PC2700

http://bbs.gzeasy.com/uploads/post-1-1048748994.png

test under 3dmark03 has also be done,the score is around 1400.

more accurate data is coming soon

 

[binmaze]

Driver version?

What's the driver version?

 

[DOOM III]

43.XX,he didn't tell me the accurate version

 

[Sabastian]

It still loses out to the Geforce MX on most test with the exception of course where the DX8 features kik in. lol, I thought that the Geforce MX supported DX8. I really am less interested in how the card performs compared to the MX I would rather see it compared with the Radeon 9200 or the lower priced version of the Radeon 9600.

 

[Evildeus]

Well the 43.45 just get out hope to fid some review with them in the neer future

http://www.nvidia.com/view.asp?IO=winxp-2k_43.45

 

[Joe DeFuria]

And idea if these are WHQL certified?

In any case, it's good to see SOME new publically available drivers from nVidia....

 

[Evildeus]

No idea :?

*edit* Doesn't seem so

 

[BoardBonobo]

The 43.45 dets have trebled most PS2.0 scores (~14fps to ~45fps) and they appear to render everything correctly in 3DMark03.

 

[Arun]

May I make everyone note that the FX5200 got many software hacks to reduce transistor counts, even though it does have some HW VS support?

Thus, benching on a 3Ghz machine is kinda insane IMO: no one is going to use a FX5200 on a 3.06Ghz CPU!

Thanks for the number, though!

Oh, and BoardBonobo: Woah! Interesting. But I'd guess that nVidia is *still* using FP16 ( maybe even INT12 ) where they're asked for full precision. But if they can get away with it perfectly, it just means FutureMark didn't use "half" correctly - and a serious game engine, such as Doom 3 or Unreal, certainly would be using it much more correctly.

So, it might be more representative of real game performance. But the real question is not that, though: the real question is whether they are truly cheating by doing this, IMO.


Uttar

 

[LeStoffer]

Oh, and BoardBonobo: Woah! Interesting. But I'd guess that nVidia is *still* using FP16 ( maybe even INT12 ) where they're asked for full precision.

I don't see how anyone could hack it to use int12, but the performance increase is so great that I can't blame you for thinking that is the case!

No, really, I think that some of the performance increase is down to nVidia getting the fragment processor and register combiners to work better together on shaders ops. Those register combiners are plenty fast, so maybe they have found a workaround to get away with using them on some shader ops not needing FP precision e.g. during final blend states. Just an idea...

 

[binmaze]

Any IQ comparison on GT4 Mother Nature?

So how good is the IQ on GT4 'Mother Nature'?
Are skies, leaves, and water rendered & lighted correctly?

 

[cho]

the gf fx 5200 was clocked at 250MHz(2X2)/400MT/s(128bits), and the gf mx440 was clocked at 275MHz(2X2)/513MT/s(128bits) . gf4 mx440 have more raw power(if not consider the shading power) compare with the gf fx5200. so in this test , gf fx5200 is ok (just) .