The New and Improved "G80 Rumours Thread" *DailyTech specs at #802*

Ouch!! Sorry CJ.

So no FP32 filtering and pretty much confirmation of stuff we "know" already.

 

Well this was understood from the standpoint of when the g80 was ready , MS seems to like to play in the middle of ATi and nV, they have been doing it for the last 4-5 years now. I don't remember if I made teh comment here or on Rage, no one had the r600 outside of ATi last week and before.

 

I suspect fp32 texture filtering has been omitted from the description, rather than being a feature that's left out of G80. We need to see a complete specification table to be sure, though.

As a minimum there needs to be support for fp32 unfiltered textures, anyway.

Jawed

 

No prob Geo.

People will just have to wait until 15:00CET.

I'll try to be more 'careful' when leaking stuff the next time like I usually do. I could just bang my head for not checking my hotmail for the last couple of days to see if I had gotten the documents.

 

R600's going out to devs, as spoilers, this week

Jawed

 

Well noone that I knew lol, and if dev's don't have them in hand most likely very few others will, actually in hand as in not supervised by ATi guys

Jawed some guys already are touching and feeling em I made a comment on the r600 on Rage, the info I have is quite rudimentary, so I didn't bother posting it here, and yeah these guys are keeping it under lock and key for the moment.

Once I get more "credible" and more infomrative info, I'll post in the r600 thread.

 

Awesome. ^ wizard rocks!

 

Not that it's the first time I read something similar; a capital B coupled with a capital S to put it in cliff notes.

 

http://www.pconline.com.cn/diy/graphics/reviews/0611/898744_26.html

Someone should correct that graph and results; what's 4x and what's 16xAA LOL

 

Based on the SS2 results? I'm willing to bet good money that those 38.1 fps are with one of the two 16xAA modes. There's no chance in hell that a G80 loses 70% from 1x to 4xAA; way less and in a couple of hours we'll be seeing hopefully more carefully conducted reviews than that one.

 

Well I think the 70% drop in this case is just indicative of the much higher performance ceiling in non-AA modes.

 

We'll see won't we?

***edit:

what's so inherently different in SS2 compared to other games that suggests such a huge drop I haven't been able to replicate yet on current cards? Measure the differences from 1x to 4xAA in all other games and 1+1 does make 2.

 

No problem, I always enjoy butting into a conversation. I think Jawed was talking about D3 #s from a B3D review at some point, wasn't he? Here we go. I actually assumed they were synthetic until Jawed mentioned D3 in a later post. But I get your point about theoretical/synthetic #s staying the same, in which case these 3DM #s line up.

Ah, so in resolutions and/or with AA lower than the max they tested, right? Those are usually gimmes for unoptimized drivers, aren't they? I mean, moreso than high res + AA, which tends to hit hardware limits along with the software ones. Generically unoptimized drivers may be more relevant with unified and hence more programmable shaders, especially with all scalars all the time.

Hopefully?! Rys knows we're comin' after him if it isn't!

 

What does that SS2 graph even mean? :S

 

Few cards popping up in froogle already. GTS isnt gouged too badly. GTX are $650-$700

 

Well Doom3 follows a similiar pattern. I'm really interested in the independent impacts of AF and AA on these G80 boards.

 

Yep, that was the most suspect phenomenon in those results. There's no valid reason for 7900GTX doing anything better than the GTS. Actually I take that back, texture address calc on the GTX would be faster.

 

I dunno, judging from the rest of the review (which BTW NV really should crack down on as it's not very nice to the other NDA abiders, though I guess it's hard to do much with sites so close to the manufacturing), that looks like they did the usual 1xAA 1xAF and 4xAA 16xAF testing, they just typoed the 16xAF. It is an absurdly huge drop, though. We also have to consider, in light of the GTS' much higher 1x1 score and the 7900's DNF and ATI's recent admission, that the X1950 might not be giving full 4xAA due to an ATI driver bug and that SS2 really is beating the cards that badly with just 4xAA. Maybe it's 4xAA with HDR?

 

The AA issue is only at 2560x1600 and only in OpenGL.