Most likely it's the score of only one of the onboard GPU's due to driver limitation, so 3DMark'06 can't "see" the two of them...
nvidia "D8E" High End solution, what can we expect in 2008?
- Thread starter 2900guy
- Start date
Most likely it's the score of only one of the onboard GPU's due to driver limitation, so 3DMark'06 can't "see" the two of them...
VR-Zone also posted some supposed gamesbenchmarks of the GF9800GX2: http://forums.vr-zone.com/showthread.php?t=229127
In that thread it's mentioned that the GF9800GX2 uses 2 G92GTS GPUs clocked at 550Mhz.
In that thread it's mentioned that the GF9800GX2 uses 2 G92GTS GPUs clocked at 550Mhz.
It is "fake" - http://www.xtremesystems.org/forums/showthread.php?t=171932
But who knows, it might be a good approximation.
But who knows, it might be a good approximation.
These are not fake: http://forums.vr-zone.com/showthread.php?t=229189
About the "benchmark", someone recycled old "simulations" on xtremesystems and reposted them on vr-zone today, hence the confusion...
About the "benchmark", someone recycled old "simulations" on xtremesystems and reposted them on vr-zone today, hence the confusion...
The cooling solution, and the way GX2 is configured is pretty interesting. Unlike its previous incantation aka 7950GX2, this one has the GPU side of the PCBs facing inwards, so they can utilize a dual slot HSF to cool both GPUs, instead of having one single slot HSF per GPU. Theres even holes at the backend of the PCBs for the fan.
Looks like its connected by a SLi bridge as seen by the pics at vr-zone.
Looks like its connected by a SLi bridge as seen by the pics at vr-zone.
How do you know this? is the mounting holes same as current cards?
Wait just how long is the GX2? the 3870X2 is about the same length as the Ultra.
Many aspects might have improved since the Napalm/V5 with multi-chip/-GPU configs, but there's still way too much redundancy involved for me personally to consider them efficient sollutions. This should of course not mean that they're useless or any of the sorts.
AMD is obviously concentrating in that direction and their drivers for such configs are constantly improving, but the path is still long enough for further necessary improvements. After those few slides that had leaked out, it looks like AMD will do something for the memory problem in the foreseeable future.
At least one step in the right direction, since I'd personally expect when you drive such a config to high resolutions with X amount of AA samples that you won't run out of memory. For the time being the only thing IHVs could do is to dedicate 1GB ram to each chip/GPU, which is the first redundancy point I was targetting above.
As for the 9800GX2 thingy, if there's going to be shortly after a higher clocked G92 (9800GTX or whatever it's gonna be called), I can't get rid of the suspicion that if one would pull them through a long set of benchmarks, including those where drivers can improve little to nothing, that the latter won't be too far apart from the first.
IMHLO until process technology and/or other factors truly hit a dead end, at least for NV dual-whatever configs could be mostly gap-fillers. The highly touted performance/Watt ratio when it comes to GPUs with high power consumption points rather in the oxymoron direction with such configs.
AMD is obviously concentrating in that direction and their drivers for such configs are constantly improving, but the path is still long enough for further necessary improvements. After those few slides that had leaked out, it looks like AMD will do something for the memory problem in the foreseeable future.
At least one step in the right direction, since I'd personally expect when you drive such a config to high resolutions with X amount of AA samples that you won't run out of memory. For the time being the only thing IHVs could do is to dedicate 1GB ram to each chip/GPU, which is the first redundancy point I was targetting above.
As for the 9800GX2 thingy, if there's going to be shortly after a higher clocked G92 (9800GTX or whatever it's gonna be called), I can't get rid of the suspicion that if one would pull them through a long set of benchmarks, including those where drivers can improve little to nothing, that the latter won't be too far apart from the first.
IMHLO until process technology and/or other factors truly hit a dead end, at least for NV dual-whatever configs could be mostly gap-fillers. The highly touted performance/Watt ratio when it comes to GPUs with high power consumption points rather in the oxymoron direction with such configs.
To tell you the truth, when I first saw that I hit up this thread to see if it was something that had been deduced or found out by a member...perhaps one on a patent expedition.
While the statement seems obvious enough (comparing G80 to G92), the last part of that statement makes me think a certain gfx card reviewer got their shiny new G94 and/or G96 review kit.
Probably purely and simply because it's coming much later, and noone knows about the specifics outside of Nvidia/TSMC. AFWIW D9E is probably in the same stage as RV770 (and vicariously R700) atm. G94 is out and about in (almost) public hands.
Yes but it seems that rv770 is next-gen AMD/ATI product. GF9800 is rumoured to be based on G92 GPU with only higher clocks. So could it be possible NVIDIA want to push another G92 SKU against rv770 i Q2?? Hmm another thing is if GF9800 will be based on G100 or GT200.
Hmm another thing is if GF9800 will be based on G100 or GT200.
Unlikely, considering 9800GX2 is G92x2
More Pics at vr-zone (from chiphell)
http://forums.vr-zone.com/showthread.php?t=230836
http://forums.vr-zone.com/showthread.php?t=230836
