Nvidia sampled only a limited number of journalists and partners with boards, and much to our surprise, don't expect our Hardwarewibbling Roundup to feature more but a small number of controlled reviews.
But, like that it matters anyways - we have already disclosed that this product is coming to market just to keep the six month cycle intact, and that the 8800Ultra is not something revolutionary, rather a check-box launch.
If you are wondering how the product looks, here is a picture below:
This new monster costs $200 more than 8800GTX...is it really worth it?
And if you want to see the scores, we have 3DMark06 score when run on a following machine:
Intel Core 2 Extreme QX6800
EVGA 680i motherboard
2x1GB OCZ Flex XLC PC2-6400 CAS3
EVGA 8800GTX ACS3 (626/1450/2000)
Nvidia Reference GeForce 8800 GTX (575/1800)
Nvidia Reference GeForce 8800 Ultra (612/2160)
Dell 1901FP19" LCD
Seagate Barracuda ES 250GB
Sony BWU-100A
OCZ ProXstream 1000W PSU
Since we did not had the 8800Ultra in our hands, we worked with our source (let's call him GeSource) to create two identical machines: both hardware and software-wise. Even though we did our best, our GeSource was still located couple of thousand miles away, so some differences may occur. Our computer ran EVGA's 8800GTX and 8800GTX ACS3, while the computer of unnamed source ran the 8800 Ultra. Our GeSource ran the 8800GTX as well, and we have compared the scores. Since score difference was practically in single digits, we deemed this comparison valid.
We apologise for the lack of bars, but for 3DMark06, performance was so close that all bars looked identical, with only half a pixel or a pixel margin, so we'll just tell you the numbers.
3DMark06 score is the following: default 8800GTX scored 12.515, EVGA's ACS3 board scored 13.115, while the latest baby from Nvidia scored 13.191, with the main performance difference between eVGA and Nvidia being the clock of 128 scalar shaders (even though the eVGA has 14 MHz performance advantage, nV has higher shader clock - 1.456 compared to 1.500 MHz).
When divided by segments, SM2.0 test was 5131 for 8800GTX, 5354 for EVGA's ACS3 and 5438 for 8800Ultra. SM3.0 test yielded in 5.418 for both default 8800GTX and 8800Ultra, while 8800ACS3 scored 5.431, taking the lead here. CPU Score was practically equal, with 8800GTX and Ultra sharing the very same CPU score (4.556), and ACS3 botched this score (4.546).
Where it gets interesting is the fact that ACS3 board has higher fill-rate, when compared to 8800Ultra: ACS3 will churn out 7541.83 Mtexel/s in Single- and 19.499,91 MTexel/s in Multi-Texturing mode. 8800Ultra cannot count on higher shader-clock here, so the board churned out 7.370,04 MTexel/s in Single- and 19.085,74 Mtexel/s in Multi-Texturing mode.
Things also got heated in both simple (and complex Vertex Shader operations, where 8800GTX scored 107.29 fps, 8800ACS3 scored 118.12 and 8800Ultra trailed with 115.66 fps.
8800Ultra's muscle showed in Pixel Shader test with 9fps lead (518.54 vs. 507.70 - regular 8800GTX scored 478.25 fps), Shader Particles showed who is the boss (Ultra scored 180.90fps, compared to 166.67 fps on the ACS3 and 161.19fps for the default 8800GTX).
Sadly, my personal opinion is that the value of "Ultra" brand is a bit devalued with 8800 series. We have seen many 8800GTX boards that can be clocked higher than 612/2160, and if you really have $800 to spend - we would advise you to plunk them for upcoming R600 board(s) or get 8800GTX+680i board. ?