very detailed 9700 preview at extremetech

You get the most complete pocture of the card from this preview. They also go into more detail explaining the new Hyper Z and Smoothvision. They show benchmarks at 1024x768 and 1280x1024. There are also some clear *weaknesses* of the card. Like for instance apparently Smoothvision will not work with pure 16 bit games... Not good. On the other hand i leared that each pixel pipe has 2 Stencil registers. Something that will clearly benefit doom 3 and its engine licencees.

Anyway.. here is is.

http://www.extremetech.com/article2/0,3973,388801,00.asp
 
Great read, thanks for the link, Hellbinder.

Love the way they present their benchmarks. I wonder how long it will take until somebody just browses the review and simply links to the JKii benches . . . 414fps indeed . . . ;)

ta,
-Sascha.rb
 
unfortunatly i was a little let down by the numbers...

Looking at 1024x768x32 and 1280x1024x32 at some of the common games of the day the 9700 just does not outperform the 4600 that much. Of course it was a little disturbing to me that they did not run with Aniso in any of their test...

I wonder if that is an indication of underlying bias? As clearly from other information the Lead of the 9700 becomes extreme when both use Aniso and FSAA...

hmmm...

Not sure what I feel about this. it makes you wonder what will happen when the Nv30 is previewed.. will it whipe the floor with the 9700? or will it have similar tendencies...
 
I think wihout AA and AF enabled or even just one of the features you are going to be more CPU bound.... Im sure the 9700 has a lot of muscle it just cant flex at 'standard' settings.

Only thing that disturbed me was the fact that the Radeon 9700 has some sort of problem with 16bit textures/resolution as witnessed in NHL.

Anyways.. it still looks good for the 9700 IMO.
 
wtf

uhh...wtf is wrong with that review. i think somebody forgot to turn off vsync. every fuckin score for geforce4 is 100. wtf lol
 
you shouldnt expect Radeon9700 to smash the GF4 Ti4600 in 1024x768 and resolutions like that without all the eye candy turned up... ESPECIALLY in simple dual texture games: the R300 and NV25 have very similar dual texture fillrate. The difference grows when you turn on eye candy, and crank the resolutions. Frankly, were seeing 1024x768 really being maxed out and no matter how fast the R300, or even the NV30, I don't expect much improvement at this resolution... without turning on eye candy.

and by eye candy, i mean max anisotropic and fsaa (4 samples or more.)
 
A big concern is the ability to properly play 16-bit games. This card looks fantastic, but it'd certainly look worse if we'd have to use another card to play older games. I suppose PCI V5's will be worth even more, then. ;)
 
i wouldn't be too concerned with the jk and q3 numbers
since their basically from an engine designed 3 years ago...their limits have been reached
 
nascar 2002 is very computer limited . i play the game all the time and it has to be tweaked to run right
 
phynicle said:
i wouldn't be too concerned with the jk and q3 numbers
since their basically from an engine designed 3 years ago...their limits have been reached

Mmmkay...its about the only engine that has scaled to 300 fps, its still being used in recently released and upcoming games, and it has rarely failed to impress even almost 3 years later...so how has its limits been reached?
 
scaled to 300 fps??vea knew that
i jus thought it was old and....
btw what do you guys consider the ut2003 as a benchmark?
 
I think it's suffered in some of those tests due to early drivers (even if they are solid, there are probably still kinks to work out). Not to mention some of those games (Commanche 4, what are they thinking) are probably a bit CPU limited.
It also looks like there might be some issues (driver maybe) with 16-bit rendering. That I don't really care too much about.
 
Hellbinder[CE said:
]unfortunatly i was a little let down by the numbers...

Looking at 1024x768x32 and 1280x1024x32 at some of the common games of the day the 9700 just does not outperform the 4600 that much.
These are CPU limited cases, what do you expect? Turn on some features like AA and aniso and you won't be disappointed :)
 
It would probably be more eye-opening if we could see real scores on these tests, especially since the close marks are very CPU bound tests as fps numbers would have shown. Obviously if card A is like 43 fps, it would be pretty apparently a CPU bound test. :)

I'm a bit floored at the JK-II @ 1600x1200 4xAA at 414 vs 100. The reviewer did sneak in the hint that it was over 85fps at this resolution. 1600x1200x32 with 4xAA at "over 85 fps" is pretty damn impressive. I just wonder how much "AA" that 4xAA really does. :)
 
It seems a lot of people don't know that Commanche is VERY CPU bound
... on a 2.53Ghz P4 ?? .. is Intel lagging behind a technology curve ? :H
screw multichip gfx solutions.. gimme 64-proc cluster of P4-s instead for real-time duck-hunt instead :p
 
Back
Top