so my question: how important is the onboard memory for the applications I'm going to be using? how about speed?
is the 580 going to be worth that couple hundred extra? I've read that complex 3D scenes can benefit from higher onboard memory, which the cheaper card has...I know there's a lot more involved than just memory, so I need some advice
You're better off with a Quadro or FirePro card. These cards have certain on-board functions for real-time rendering disabled. The speed can be important as about as much as the memory, but what is more critical is the clock of the memory. Even then, that's not very important when you have hardware-specialized cards for these things. Both the Quadro and FirePro are very adept at what you're doing.
Now, if you're going to game too, then they are useless. If you are trying to do both, the GTX-580 is better if you can use CUDA, notwithstanding the more important higher memory clock. The GTX-580 is about worth the increase in price.
VRAM is fairly important when it comes to rendering complex 3D scenes, and professional workstation cards can have up to 6gb of onboard memory. However, desktop cards offer great value in comparison to the much more expensive Quadro and Firepro cards in 3D applications. Desktop models may not be quite as fast or as stable, but they are pretty close, and are much better for other tasks like gaming and what-not than the pro cards.
thanks everyone for the input; cost is definitely an issue here, so that is why I was leaning toward the 560 (with the 580 pushing the budget)
Given that, and the fact that I won't be using "real-time" rendering, such as VrayRT, does it still make sense to splurge on the 580? I guess I don't understand what the GPU can actually do for me come production time. Is it actually going to work alongside my CPU to speed up a 3D animation batch render, for instance?
I know I need to do some more googleing, but I really appreciate your advice
this is interesting, could you elaborate on this just a bit for me? I'm starting to understand that clock speed relates to redraw/refresh rates, while memory relates to the size/complexity of the scene that can be accomodated? meaning 3GB would let me render extra-hairy stuff?