[B3D Analysis] R600 has been unleashed upon an unsuspecting enthusiast community

8.37.4.2 seems to be more recent than 8.37.4.3 (11th of May vs 10th of May)

 

That is pretty cool.

 

frankly ? no.

one DX10 bench : Lost Planet

 

I think the only thing that may save this card is the price point at which it enters the market. Its late. There are other cards out that already show good performance. And honestly - who cares about DX10 right now when it is Vista only so far. Most hard core gamers are not running on Vista yet due to poor performance (regardless of gpu manufacture).

Buyers do not typically buy a card for tomorrow, they buy for today.

And if the R600 spends time updating drivers for either games or direct x version, who says the other manufacture won't do the same? I'm not sure who to blame more - the people/buyers for generating the hype and setting false expectations or AMD for not correcting them.

What games are coming out this year that say they are DX10? Unreal Tournament 3, others? And of those games - what sort of shaders or other processing tricks do they use within the engine.

Maybe that then will indicate what the 2900's will do.

 

This is from the dinox item..

I agree about the blurring, I don't like it. But the detail that is actually kept after applying 8xAA is amazing.

8xAA+WideTenT
http://www.dinoxpc.com/Tests/articoli/articolo/video_jetway_hd2900xt_images/aa8x+wt.jpg

16xQAA
http://www.dinoxpc.com/Tests/artico...d2900xt_images/AA16xQ_pic_785472_25_small.jpg

 

Downside to that is questionable image quality, which is worse than GTX's.

 

Forget that..

real DX10 benchmarks:
http://www.legitreviews.com/article/504/1/

 

as far as i know , Lost Planet is being optimized by nVidia

 

The edge quality is good, but the text is blurred in a way that it's messing with my eyes (well, brain).

 

I donb't see what's more or less real benchies. Both are legit :/

 

R6:LV benchmark has no AA ( <- that's a performance enhancer for the 2900, right there) due to issues. So image quality is equal in there

 

In one word: Underwhelmed.

Where is the R650?

 

Yeah, I'd be interested in further noise there, as it would seem to imply that G80's dynamic load balancing ... isn't.

-Dave

 

take a look at this review : http://www.ixbt.com/video3/r600-part3-d.shtml#p818

shows impressive numbers with and XT clearly beating GTS and appeared very late in the day ,maybe with better drivers ?

 

That is an interesting wrinkle.
I guess R600's compiler is capable of working through computational dependencies very well.

G80's scheduling method seems to be less efficient.
Perhaps its heuristics are targeted for a graphics workload that uses texturing.

It doesn't seem G80 suffers all that much from broken load balancing, which means the bottleneck must be elsewhere.

Considering R600's superior instruction issue rate on hand-picked instruction mixes indicates that its utilization on real code is way below peak on game code, I wonder if making much noise of the fact that R600's competitor is winning despite being hobbled (when it's not already hobbled as a GTS) is a good thing.

 

8.37 like most if not all, dunno which 8.37 though

 

CoJ benchmark numbers appear to be almost the same
the 2900 loses 5fps comparing the dx9 version to dx10(30 to 25) at 1600
the 8800gts loses 18fps (50%) 36 to 18 at that res.

 

"Hope is a dangerous thing" -The Shawshank Redemption

Nah, lets hope the driver gods will hear our prayers.

 

great writing Rys!!!

 

Is there anyone that tried disabling CatAI and hope for some... "results" ?

I can imagine it does some good to IQ higher res.