 |
February 14, 2021, 12:20:16 AM |
|
Can you add some stuff from Nvidia Developer that accelerates your GPU? There is a memory bus that allows you to get 2x the performance out of your GPU. Tensor Cores aren't utilized at all. You have to call up specific functions just to turn them on and it has to be through a daemon processor. Then you have to use something like ML studio to change their floating-point integers into linear computer processing. I use the A100 to train my data, but I have of course been curious. I spent about 6 full days looking into it. It's too much for me. But if you can, you should try to get Tesla, A100, etc. 8 of the A100s had an ETHash rate of 1398MH/s
|
