9800 GX2 Reviews

No one has time for them anymore.

 

The vast majority of the added value in B3D's reviews is in the architectural analysis. We don't need a B3D review for rehashes like the GX2.

 

We dont need it but it would be nice to see some over the top scaling analysis

 

Well any site can do that....

 

I wish Mark would do reviews for B3D as he did for R3D. I know he doesn't have the time for it but he seriously had the best review format on the web. Always loved the IQ comparisons.

 

He'll be doing even much better than that within 3 months and pre-GT2xx/R7xx if everything goes as expected, don't worry. I firmly believe we can add a truckload of value to the benchmark side of things - just not by doing it the same way everyone else is doing it, though. And that can actually help the architectural analysis side of things rather than reduce the time we have for it.

Sorry for teasing, but this is still very much in the concept stage although it's advancing quite well... Doesn't mean we won't have any review before then, we hopefully will, but certainly we won't have any launch-day reviews before then as far as I can tell. And let's try not to derail this thread too much...

 

Glad to know you guys are still working on editorial content. Not that we don't appreciate your forum and news analysis work, but the reviews is where it's at.

 

Well I'm not sure "reviews" is the correct word... I mean, there will be that, but that's not where the value add will be... Anyway, I need to stop teasing about things that are far from ready, that's not going to get us too far!

 

http://en.expreview.com/2008/03/19/quad-crossfirex-smasher-geforce-9800gx2/

 

Wow, GX2 is faster then Quad CF?!

 

Considering some titles don't scale at all to Quad, why not?Since the GX2 is faster than the X2.

Also, in OpenGL there is no Quad support for the moment, it's going to be included in a future driver. His DiRT numbers are low(does he say how and where he tested?)-also, I think the number are generally futzed there because with patch 1.1 you can't have Ultra shadows with 512MB cards, so the X2 and GX2 will/should be using High Shadows, whilst the Ultra will be using...ehm, Ultra. IME, DiRT is one of the titles that actually scales wonderfully to QuadCF.

 

http://www.rage3d.com/previews/video/nv9800gx2/

For what its worth. Here's my analysis of the 9800GX2. Yes I know I'm am newbie reviewer so dont tear me apart please

Chris

 

Yeah I know, but it does also quite well where QCF works. (COD4, they propably tested with water on med, there is a bug with it with ATI cards and CF)
I hope this is not the end of QCF perf evolvement cause there is still much to improve (UT3 QCF is still far from optimal and renaming the exe to FEAR.exe does better), but the potential is there. (COJdx10, Timeshift and Oblivion are screaming)

 

Timeshift?As a curiosity, how is it doing in the very first level, the one called "Arrival"...more precisely, when the scene with the big-robot happens and then when the dude who's leading you has a head-on with a "helo"...and at which settings. BTW, PM me the answer so we don't go very OT here. Cheers.

 

It craws, at the beginning 30-50 and quickly to 2 fps. I don't know what's goiing on there.

BTW I'm running a 9800GX2 now and I'm suprized how wide spread SLI is working.(GX2 is dooing a steady 85+(running V-sync and TB now) in those scenes) It bought for 489 euro's and its really a good performing part, in fact it's as fast and much faster (where (Q(CF)) fails as my quad CF.
I was amazed its as fast in Oblivion!

Crysis isn't much faster, but it don't stutter at least.

 

Thought as much.

Now you only need to get a 790i SLi mobo and go QuadSLi

 

I never say never, but I'll wait for the next single GPU generation propably

 

Same here, I'll keep my 8800 GTX until then.

 

Does anyone find the look of the card's covering resembling a taser of sorts? A really big taser, but a taser none of the less.

 

Great review ChrisRay. Finally a high quality comparison against lots of other SLI datapoints! Happy to so the 16AA and 16Q also tested.