Hi Everybody,
I have been working with a Tesla870 during a fellowship. Now I need to get a card for implementing some image reconstruction algorithms for tomography in our lab. As the tesla is quite expensive, I was planning to buy a GTX285 or a GTX295. But I can´t decide which one, a GTX285 with 2gb ram it has almost the same price than a GTX295. The GTX295 it has twice the cores than the GTX285, but on the other hand I have less memory and also for programming I need to split my algorithms between the two cards inside the GTX295.
Which of these two cards would you suggest me to use for implementing scientific algorithms?
Regards,
MartÃn Belzunce
The question is, are you willing to modify your code to take advantage of multiple GPUs? Or will your memory needs exceed the 896MB offered by the GTX295 per GPU? If you do not have time for this or need more memory, then the GTX285 would be a better choice.
If you want a lot of memory and be cheap, there are some GTX260/GTX275 which have 1792MB of memory, with two of those you get the best of both worlds (but you need a mainboard with two 16xPCI express).
Something to consider also is whether the big-ram cards have equally fast RAM as the smaller cards. I have no idea if it is, I don’t want to spready any misinformation, just be sure to at least have a glance at the memory bus info / memory speed data btw. the different cards.
Do you know where the bottlenecks are in the tomography stuff you’re doing? FLOPS or memory bandwidth or … ?