Nvidia Tesla K40 for lab?
-
I know nothing of machine learning or GPU processing in general. But I'd like to learn.
To get started I assume you need a GPU card of some kind and GPUs like Nvidia Tesla K40 are showing up refurbished for reasonable prices. Would that card be sufficient to get started or is it too old?
Also when looking at desktop virtualization would the Tesla K40 also be something you could use for this? Or does that require something else? I've seen Grid K1 & K2 in some applications but I don't know the difference between those cards and something like the K40.
-
NVidia recommends K40 for machine learning developers. So I think it's plenty for that kind of stuff.
-
-
-
I think you are looking at cards meant more for deployment of machine learning, rather than learning it. Even in production, lots of people aren't going for that kind of hardware. I'd think for learning that a 1050 or something would be plenty.
https://blog.slavv.com/picking-a-gpu-for-deep-learning-3d4795c273b9
-
@pete-s said in Nvidia Tesla K40 for lab?:
I know nothing of machine learning or GPU processing in general. But I'd like to learn.
To get started I assume you need a GPU card of some kind and GPUs like Nvidia Tesla K40 are showing up refurbished for reasonable prices. Would that card be sufficient to get started or is it too old?
Also when looking at desktop virtualization would the Tesla K40 also be something you could use for this? Or does that require something else? I've seen Grid K1 & K2 in some applications but I don't know the difference between those cards and something like the K40.
Those special purpose cards are really only for after you have a very specific need. Machine learning and GPU compute can be done on just about any graphics card of reasonable age. The add in cards make more sense for servers where the standard GPU is just enough to run a display, and nothing more.
