nVIDIA november assault

Page 4 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

alcoholbob

Diamond Member
May 24, 2005
6,271
323
126
Originally posted by: lopri
Originally posted by: Ares202
when is the 2950 due because in initial benchmarks its beating the 8800 ultra comprehensively
Can we see those initial benchmarks?

I'm guessing it does 33fps in Crysis instead of 31fps.
 

Cookie Monster

Diamond Member
May 7, 2005
5,161
32
86
Lets take into account that the G92 has a die size of 17x17 (mm). Thats 289mm^2 on 65nm. (G80 is 484mm^2) Now the current rumours point to ATi's RV670 having a die size of 196mm^2 at 55nm. IF G92 is being built around 55nm process, than the die size of 234mm^2. Now RV670 packs everything that the R600 has except the 512bit memory bus (has 256bit bus instead but also includes UVD).

So G92 having 112SPs doesn't sound too far fetched.
 

bryanW1995

Lifer
May 22, 2007
11,144
32
91
g92 is going to be clocked higher than current 8800 gts, so it if also has 112 sp's...
 

superbooga

Senior member
Jun 16, 2001
333
0
0
Originally posted by: s44
If the GT is 112, what room does that leave for even the revamped GTS?

None. Why bother with G80 when it costs a lot more to make than the G92? G80 can be reserved for the GTX and Ultra.

The G92 will be at least as fast as the current 8800GTS, because tt needs to be faster than the 2950Pro.
 

superbooga

Senior member
Jun 16, 2001
333
0
0
Originally posted by: keysplayr2003
There is a missing link here that we can't seem to get our hands on. The current 8800GTS is supposed to get an upgrade of 16 more scalar shaders for a total of 112, and all else remains the same. The 8800GT, if said to encroach on a 96 shader 8800GTS, is very likely more than 64 shaders especially with a cut down bus (256bit).

The missing link is that the card must beat the 2950Pro decisively. It's not going to with 64 shaders, and the 8800GTS isn't going to be sold at $199-$249.
 

R3MF

Senior member
Oct 19, 2004
656
0
0
i think nvidia are playing us all for fools, there have been so many contradictory and insane cross pollinations of different rumours that their PR department should win an award for disruptive FUD after the release of whatever the G92 is.

First the G92 was high-end
Then it was perf-mid
Then it was 256bit with 64 SPU's

Then the 8800GTS gets a revamp with 112 SPU's
Then the 8800GT comes along with 112 SPU's (rumoured to be the G92)

How about this:
8800GT is a 8800GTS with a few more SPU's (thereby accounting for the 112 rumour)
Everybody knows the GTS 320 doesn't have enough memory, so the arrival of the 8800GT allows the GTS640 to replace the GTS320
8700GT is in fact the 64 SPU 256bit G92 chip and sits below the 8800GTS 640

On the other hand:
A 289mm^2 chip at .65u is still big enough to be a die-shrink 8800.
The NVIO chip is still separate so PCIe2 is easily added to whatever SKU of the G92 (high) nvidia fancies.
So there is a .65u 8800GTX - 128 SPU 384bit (maybe with a different name)
And a .65u 8800GTS - 96 SPU 320bit (maybe with a different name)
And in the middle a .65u 8800GT - 112 SPU 320bit (maybe with a different name)

Or it could be something completely different, because nvidia have played a devious game of FUD this time around!