Friday, November 19th 2010
NVIDIA Readying Dual-GF110 Graphics Accelerator, Eying Total Performance Leadership
NVIDIA stunned the computing world with a speedy launch of the GeForce GTX 580. The GPU was able to increase NVIDIA's single-GPU performance leadership, and also iron-out some serious issues with the power-draw and thermal characteristics of previous generation GeForce GTX 480. It is now that a dual-GPU implementation of the GF110 graphics processor, on which the GTX 580 is based, looks inevitable. NVIDIA seems to be ready with a prototype of such a dual-GPU accelerator, which the Chinese media is referring to as the "GTX 595".
The reference design PCB of the dual-GF110 accelerator (which still needs some components fitted) reveals quite a lot about the card taking shape. First, it's a single PCB card, both the GPU systems are located on the same PCB. Second, there are slots for three DVI output connectors present, indicating that the card with be 3D Vision Surround ready in a single card. You just have to get one of these, plug in three displays over standard DVI, and you're ready with a large display head spanning three physical displays.Third, it could feature a total of 3 GB of video memory (or 1.5 GB per GPU system). Each GPU system has six memory chips on the obverse side of the PCB. At this point we can't comment on the memory bus width of each GPU. The core configuration of the GPUs are also unknown. Fourth, power is drawn in from two 8-pin PCI-E power connectors. The card is 2-way SLI capable with another of its kind.
Source:
enet.com.cn
The reference design PCB of the dual-GF110 accelerator (which still needs some components fitted) reveals quite a lot about the card taking shape. First, it's a single PCB card, both the GPU systems are located on the same PCB. Second, there are slots for three DVI output connectors present, indicating that the card with be 3D Vision Surround ready in a single card. You just have to get one of these, plug in three displays over standard DVI, and you're ready with a large display head spanning three physical displays.Third, it could feature a total of 3 GB of video memory (or 1.5 GB per GPU system). Each GPU system has six memory chips on the obverse side of the PCB. At this point we can't comment on the memory bus width of each GPU. The core configuration of the GPUs are also unknown. Fourth, power is drawn in from two 8-pin PCI-E power connectors. The card is 2-way SLI capable with another of its kind.
153 Comments on NVIDIA Readying Dual-GF110 Graphics Accelerator, Eying Total Performance Leadership
EDIT - Your username pretty much explains everything lol.
You are to young and innocent to remember that back on the time, the best 3D card was 3dfx VooDoo with 2 GPUs, than Voodoo 2 with 3 GPUs, than the prototype Voodoo 5 with 4 GPUs, etc. Everybody was talking about the card, not about how many GPUs had.
The same thing with this card. If they will sale this, probably nvidia will became the producer with the fastest graphics card in the market. That is all that matters. ;)
PCI-E bandwidth doesn't have anything to do with the limiter, and keep in mind, PCI-E 3.0 isn't out and we still have barely saturated 1.0/1.1 anyways.
But just look how true the "SLI scales better" argument is, generally, that without any comment, by only posting some charts, some empirical data collected by W1zz, so many people in this thread suddenly thought I was implying SLI is better. Sorry guys that's only your own subconscious betraying yourselves and indirectly making you agree with something you would never admit... First of all, read above. :roll:
Second, of course only max resolution matters on this particular SLI/Crossfire debate (GF110 vs Cayman). I refuse to judge $800-1000 graphics setups based on low resolutions, that's stupid. I can go even farther, anyone who does $500-1000 SLI/Crossfire in order to play on anything below 1920x1200 8xAA or 2560x1600 4xAA is just stupid, let alone at the lowest 3 out of 5 of the resolutions that W1zz uses in his reviews. If you want to game at a max res of 1680x1050, buy a single GTX460 or HD6850 and that's it.
A15: The PCIe Card Electromechanical (CEM) 3.0 specification consolidates all previous form factor power delivery specifications, including the 150W and the 300W specifications.", I'm wondering if this GTX 595 won't be then what cards will be using the new specification.
As for the position of a dual GF110 for total performance leadership, it becomes very relevant for high end scaling as it will be two near top gpu's from both camps slugging it out. Fudzilla suggests it will be out soon, others say it will wait for the 6990.
I think these cards will be crippled by their own siblings. Does putting two gpu's onto one core (and surely downclocking or underpowering them) not give less performance than two sli or crossfired cards?
i.e lets say the green dualie is two GF104's. (i can't believe it will be two fully operational 580's.) Therefore to 580's in sli would be better. So cost will have to be carefully considered.
Likewise for AMD. Two fully operational 6970's seems too much (though we dont know how they are yet!).
Like i've said before in this thread somewhere - not for me. I'd rather have the one powerful gpu from now on - and i'm leaning green right now. Maybe Dec 13th will change my mind?
I have 2 GTX260's that for the most part are on par with the 295 witch is why I have yet to upgrade.
The GTX 595 should be good a great card, I will more then likely pick on up due to it having NV surround on one card, I am running 2d surround on my 260s now and for the most part is great, I do run out of Vram on some games.