HFR review up

[Evildeus]

Interesting, the X800XT is 20% faster than the 6800U Extreme. All the bench with 6800 GT/U/UE and X800pro/XT.
http://www.hardware.fr/articles/491/page1.html
2 charts interesting from my point of view:
Ati is clearly in the lead in VS. How come the 6800 is faster in dynamic branching than static???

Ati is also faster in PS2.0, except procedural

 

[Pete]

How come the 6800 is faster in dynamic branching than static???

My guesses are probably the same as yours:
a) nV's drivers are just not as efficient at unrolling shaders as ATi's drivers are, for now?
b) if the shader is long, perhaps nV still has more limited register usage compared to ATi?

 

[AlphaWolf]

ED, you have a . missing in your url.

 

[Evildeus]

ED, you have a . missing in your url.

Yeah, thanks. Weird with my copy/paste

 

[Evildeus]

HFR says that there's probably a long lasting bug in the drivers.

How come the 6800 is faster in dynamic branching than static???

My guesses are probably the same as yours:
a) nV's drivers are just not as efficient at unrolling shaders as ATi's drivers are, for now?
b) if the shader is long, perhaps nV still has more limited register usage compared to ATi?

 

[Ylandro]

How does the benchmark work?
I notice that the Radeon are running "V.S. 2.0" and "V.S. 2.0 static flow control" at exactly the same speed. Is that coincidence, or is it designed that way?

And how does that compare to the 6900's that run the static flow control at almost half speed compared to 'normal'.

 

[Stryyder]

So who is going to translate the whole thing for me???

 

[Pete]

Google?

 

[Evildeus]

Not me, too long

 

[vrecan]

whats really interesting to me is seeing how fast the GT is. If this card hits the right price it might be a steal! Almost twice the speed of a 9800xt in a lot of things.

 

[cthellis42]

How come the 6800 is faster in dynamic branching than static???

My guesses are probably the same as yours:
a) nV's drivers are just not as efficient at unrolling shaders as ATi's drivers are, for now?
b) if the shader is long, perhaps nV still has more limited register usage compared to ATi?

He's talking about the 6800 compared to itself, not ATi. That comes down to the differences between how the card and drivers handle static and dynamic branching, for which I'd personally have no idea. Heh...

 

[Pete]

Sorry if I wasn't clear, but I was referring to the 6800 alone. My point is ATi loses virtually no performance with branching, whereas nV does. So the culprits would be either drivers or hardware. Either ATi is better at dealing with static branches at the driver level, or their hardware (cache / memory switch / ?) allows them to minimize that performance hit.