Question GPU for mostly productivity

[anvoice]

Contemplating which GPU to get for a new Ryzen AM5 build. I will not be gaming, prompting the question: what GPU would serve me best for things like regular productivity tasks, programming, CAD modeling, possibly machine learning as well as future proofing?

I'm considering a used RTX 3090 since my budget allows for it, but what do you think? Most opinions I read favor the 4070 ti over the 3090 for gaming, but again, not doing that and don't want to take the bus width and memory hit (although the better thermals on the newer card are nice). I also don't think specialized CAD or machine learning cards are the best idea as they tend to do only that one task well and nothing else, on top of being expensive.

 

[kanewolf]

Contemplating which GPU to get for a new Ryzen AM5 build. I will not be gaming, prompting the question: what GPU would serve me best for things like regular productivity tasks, programming, CAD modeling, possibly machine learning as well as future proofing?

I'm considering a used RTX 3090 since my budget allows for it, but what do you think? Most opinions I read favor the 4070 ti over the 3090 for gaming, but again, not doing that and don't want to take the bus width and memory hit (although the better thermals on the newer card are nice). I also don't think specialized CAD or machine learning cards are the best idea as they tend to do only that one task well and nothing else, on top of being expensive.

If you want CAD and productivity, get a Quadro. The drivers are optimized for CAD not gaming.

 

[anvoice]

If you want CAD and productivity, get a Quadro. The drivers are optimized for CAD not gaming.

A good Quadro is a bit over my budget, and probably not good for AI training.

 

[hotaru.hino]

Do you do any of this stuff now? If yes, figure out where at now, where you'd like to be, figure out if consumer cards even work with the tools you use, and then go from there. If no, at least figure out the last part.

I don't find spending top dollar for a halo card worthwhile unless your time is literally money. And even then, it probably isn't worth while.

 

[anvoice]

Do you do any of this stuff now? If yes, figure out where at now, where you'd like to be, figure out if consumer cards even work with the tools you use, and then go from there. If no, at least figure out the last part.

Thanks for the suggestion. I do programming and CAD work. I haven't jumped into machine learning yet but it's a likely option in the future. Not sure how high level it will be, but definitely doing some deep learning, etc. I do realize that's a bit vague but it's my best approximation.

I do have the option of using an older GTX 1070 for a while and deciding later. Not sure if there'll still be a good supply of 3090 cards by then though...