I'm about to purchase one of the new Fermi line graphics card from nVidia, and of course would like the investment to be future proof. The entry level card is the Quadro 4000 (2GB buffer, 250 GPU cores); its price/performance ratio is good but it's not designed to work with a second or third card of the same type, to form the scalable CUDA platform. For that, the much more expensive Quadro 5000 or 6000 are required (see here : http://www.nvidia.pl/object/quadro_sli_pl.html).
Now, can Moldflow reveal some details about the planned scalability of CUDA support in the future MPI releases? Will the solution described in the above link be mandatory to expand the GPU computing power with consecutive cards, or will it also be scalable with the Quadro 4000 model, by simply adding another card of the same type?
TIA
Piotr
Dear Yannick,
Thanks so much for your answer. For my main workstation, I've already settled down with the Tesla C2050 - AMI 2012 Beta uses it perfectly well (so does AMA 2012, BTW). In my next machine, it will be complemented with the Quadro 6000 (or equivalent)...
But, since I'm shopping for a laptop anyway (to prepare presentations, and even run ad hoc analyses at my customers'), I'd like to take advantage of CUDA as well. The problem with Quadro 5000M (which is only slowly becoming available in the highest, and VERY expensive, flagship models from HP and Dell) is the price, so I'd like to buy a fast "consumer" type machine. The Dell XPS 17 features the GeForce 445M GT card with as much as 3GB frame buffer - but neither Dell, nor nVidia support have answered my inquiry whether the card fully support the 64-bit FP computing....
Could you help with this, please? Again - it's the 3GB GT 445M card.
TIA
Piotr
Sorry...It is out of our range. Those cards and hardware are brand new, and we haven't been able to test GEforce from this generation.
Note that the GeForce cards are not manufactured or supported by Nvidia(they only manufacture the chips). So you should pester the card manufacturer or the laptop reseller.
Our developper tried to look and there is a possibility that the card you mentionned has a compute capability of 2.1.
This value of 2.1 is new to us, and we haven't tested that card, so we cannot guarantee anything at this stage.