A promise is a promise :) there you go!

Gotta admit... it looks darn nice.. My little bias meter is kicking in. But I'd like to see Nvidia come back and pull the rug under ATI,


Mainly cuz of chipset troopers

 

All the fill rate in the world won't defeat being CPU-limited, you idiot (YES, I AM CALLING YOU AN IDIOT!). You said your results are for BOTMATCHES. If that's the case, HOW is that card getting TWICE the performance of a Radeon 9800XT on the same CPU - when overclocking the XT does SHIT for performance?

 

maybe the lower score we see in botmatch is a result of VS & CPU limited combined,and NV40 just free up the VS part.

But I also think OCing a 9800XT should give a higher score.

 

Okay okay there is a fair bit of concern about stuff being CPU limited atm can you give us the scores for UT2k4 ( I think thats what the game was ) on the same rig but with high end card that is currently aviable to the public and then rerun it with the card overclocked?


As for this

well this was clearly answered today actually

 

maybe NV40 has an "AI" co-accelerator built-in to the die

 

Ummm.... Maybe I'm missing something, but the Radeon 9800 XT wasn't overclocked, dimwit... The 9800 Pro was overclocked to 9800 XT speeds, or close to it, and gained 5 FPS.

 

Whoops, my mistake, I mis-read the graph labels. It still didn't bring performance very much higher than the XT - compare clock speeds.

 

actually dimwit there was a 9mhz speed increase over the xt . the pro and xt clocked at same speeds perform exactly the same in unreal 2k4

if you look at the full article you will see the 9800pro overclocked is getting about 1 fps more through out the whole test till what he posted.
Then its loosing by .2 fps . Basicly dead even. its most certianly cpu limited .

 

Given, the ram of the supposed test system, it is possible the person is overclocking the A64 3200+. If by some chance he maxed out his ram/proc speed it would be roughly a 37.5% increase in performance (maybe / probably more) over an A64 3000+. Don't know if this is correct or not, just thought it might be a possibility.

 

I really, really, really don't want to get involved in bickering but would like to offer a suggestion which might help ease justifiable suspicions in light of what happened prior to the nV30 being available..

How about posting a picture of the card?

Also, as suggested by stevem, "How about some shadermark output?".

 

[/quote]

That's new to me,I'm quite impressed by that they didn't use VS to do the character skelecton animation.

But if there's no dedicated TnL unit in the GPU,FFP still need to be emulated by VS,hence it can be called a VS application.

 

If anything I'm an angry aunt.

And you'd better check on those numbers. >_>

 

Can we cease the name calling? It's not really, uhm, useful here.

 

Yeah fillrate testing as well as UT2K/3dmark2k1 testing doesn't really show off the pixel shading power which is supposed to be 2.5X more powerful ( I'm hope that is for clock per clock and pipeline per pipeline ) then the NV30.

 

I'm sure the shader performance is up to par this go around.
Nvidia knew that was a big down fall of the NV3x.
It had to be fixed.

 

so 200mhz is supposed to get rid of a cpu bottleneck. Which is what i believe is happening (which i have seen happen with the botmatchs . 9800xt on a 2500+ athlon gained 3 fps by increasesing 100 mhz) .

Not sure .

Anyone that has more knowledge on this benchmark care to comment before blood is spilt