nVidia vs ATi with 8xAA

[lopri]

Originally posted by: n7

Originally posted by: BFG10K
Benchmarks with 16xAF combined with 4xAA and 8xAA should be mandatory. It's 2008 for heaven's sake.

I'd agree, but too many games are using deferred shading these days...

Yup. That's the problem. Current cream of the crop games, thanks to their console origin, just don't like traditional AA it seems. Either AA doesn't work very well (visually) or it cripples the performance by some absurd margins. Crysis is an exception but that thing doesn't run well on anything on earth anyway.

All of sudden benchmarks with no AA became acceptable. And I'm afraid this trend will continue for a little longer. (At least until next gen hardware hits the market)

 

[BFG10K]

BFG, I actually believe that those with 8 series video cards can force AA with S.T.A.L.K.E.R., so...yay?

Do you have more details about this?

For me forcing driver AA with the static path results in no AA while forcing it with the dynamic path results in a yellow screen.

 

[BFG10K]

Okay, I have some results from Stalker + driver AA:

[*]2xAA works and screenshots confirm reduced jagged edges over no AA.

[*]AA levels other than 2xAA range from severe artifacts to a BSOD.

[*]TrAA does not appear to function in any form.

 

[Munky]

8x AA would make more sense when testing dual card setups, because in many modern games no single card (except the 3800x2 maybe) has the power to run it smoothly at resolutions I care about. As far as single cards go, I'd rather they test with TRAA/AAA enabled to get rid of the jaggies in objects using alpha textures. I play with TRAA enabled whenever possible; we've had that feature for several years now from both camps, and yet most review sites blissfully ignore it.

 

[nitromullet]

Originally posted by: lopri
All of sudden benchmarks with no AA became acceptable. And I'm afraid this trend will continue for a little longer. (At least until next gen hardware hits the market)

Doesn't DX10.1 require a capability of at least 4xAA?

Originally posted by: munky
8x AA would make more sense when testing dual card setups, because in many modern games no single card (except the 3800x2 maybe) has the power to run it smoothly at resolutions I care about. As far as single cards go, I'd rather they test with TRAA/AAA enabled to get rid of the jaggies in objects using alpha textures. I play with TRAA enabled whenever possible; we've had that feature for several years now from both camps, and yet most review sites blissfully ignore it.

That is a good point. I generally don't just run with AA, but with transparency, super sampling AA whenever possible.

 

[Ackmed]

How about they show the power draw from two 8800GT's in SLI, and not just one 8800GT. They show frames based on SLI, but only the power consumption of a single card? That doesnt make any sense.

 

[Cookie Monster]

But what about CSAA modes specifically designed to use bandwidth and framebuffer more efficently than traditional MSAA while producing respectable IQ? e.g why use 8xQAA when 16xCSAA would most likely be similiar to 8xQAA while maintaining a good performance? If this was the case, how would one judge which card is faster?

One could use 8xQAA (NV) vs 8xAA (ATi) following the apple to apple benchmarking procedure BUT why not use modes like 16xCSAA where it was designed for those exact situations (bandwidth/framebuffer limited). Computerbase.de shows the HD3870 faster using traditional 8xMSAA compared to the 8800GT BUT would this be true if the GT used 16xCSAA? We all know the GT lacks bandwidth alot for high res AA/AF. So why not use those AA modes that counter the problem?

Unless someones willing to deeply investigate higher AA modes (where there are MANY), id say just stick with TRAA/AAA with 4xAA for single cards. For dual GPU setups, im not too sure but i do agree with munky.

 

[themisfit610]

Portal was a blast at 1920x1200 8xAA. The 8800gt ripped through it.

 

[Blacklash]

Very well. Poll added.