-1
$\begingroup$

Though the combination between gaming and deep learning might not sound "serious" enough still this is the case - on the one hand, I need a great GPU for my son to play games, and it will be great if I can make ML on the same GPU. Thus my question is as follows:

I know that the line of GPUs called RTX, such as 3060,3060Ti up to 3090, and 3090Ti should do the job, but. When I go to online stores such as KSP, I get many "subtypes" of RTX 3090: an example of rtx3090 "sub-types" available And I am afraid to purchase something that will only fit for gaming and not fit my needs. I am pretty new to ML and I know that a graphic card should support something called CUDA, are all of those cards in the image above support everything needed? Which one of them may be preferable?

Improve this question
$\endgroup$

1 Answer 1

Reset to default
3
$\begingroup$

All cards from this series support CUDA. In fact they even have special cores, designed for faster deep learning calculations called 'tensorcores'.

If you want to do some deep learning with big models (NLP, computer vision, GAN) you should also focus on amount of VRAM to fit such models. Nowadays I would say at least 12GB should suffice for some time.

So I would select cards with minimum 12GB and buy the best you can afford.

Personally, I would probably focus on 3090 and not 3090 ti, as the price increase is pretty significant and probably not worth the increase in computational power.

Also if you're new to ML/DL, probably you should first learn some of this stuff before deciding on spending money on equipement. Not all ML/DL models benefits from using GPU over CPU. Smaller/simpler models are trained faster on CPU. More on that matter: Is a GPU always faster than a CPU for training neural networks?

Improve this answer
$\endgroup$
3
  • $\begingroup$ Hello, thank you very much, sir. And yes I am doing CNN particularly YOLOs But, regarding my concern. Are all of the presented in imaged versions of 3090s are fit for deep learning? I do not totally understand differences between them $\endgroup$
    – Igor
    Commented Aug 21, 2022 at 15:16
  • 3
    $\begingroup$ Yes, they are all fine. They all should have the same specs that are relevant for DL such as memory, cuda cores, tensorcores, bandwiths, memory buses. The main difference between these 3090s are manufacturers of some of the external elemnts of the gpu (fans, sockets, board, cables, etc) but the main chip onboard of these GPUs, that is responsible for calculations is always from NVIDIA and has all these parameters mentioned ealier the same across different manufacturers. $\endgroup$
    – GKozinski
    Commented Aug 21, 2022 at 16:35
  • $\begingroup$ Thanks you very much, sir $\endgroup$
    – Igor
    Commented Aug 22, 2022 at 11:13

Not the answer you're looking for? Browse other questions tagged .