3 ms·
I use CUDA for scientific calculations and I wonder how this will affect GPU performance since any detection algorithm may have fault positives. The only proble
by nuccy 6y ago
I use CUDA for scientific calculations and I wonder how this will affect GPU performance since any detection algorithm may have fault positives. The only problem with such artificial limitation is that usually people who mine crypto are very inventive (if you have doubts about that just check laptop mining farms [1]), since they try to squeeze out as many H/s as possible. It is just a matter of time when the 'right' driver will be available.
1. https://www.notebookcheck.net/Nvidia-GeForce-RTX-3060-laptops-from-Hasee-snapped-up-by-the-hundreds-for-a-single-crypto-mining-farm-in-China-as-business-booms.519605.0.html https://www.notebookcheck.net/Nvidia-GeForce-RTX-3060-laptop...
- jbnorth 6y agoHonestly if it's an issue they'd probably tell you you're also using their GeForce cards wrong and you need to move up to Quadro.
- notum 6y ago"No no sir, I see that you're using Quadro TensorFlow, you need to use a Quadro PyTorch card."
- hobofan 6y ago> since any detection algorithm may have fault positives Though not impossible, NVIDIA has been doing application detection for ages for the purpose of tuning the performance of the latest popular games. So as long as your scientific calculations don't calculate too much Keccak-256 hashes, they should probably be fine.
- terramex 6y agoTheir application detection is based on executable name. Fairly popular game I worked on was misdetected because apparently we used executable name of some decade old FPS. They fixed it in NVidia Control Panel, but Shadowplay continues to put gameplay recordings in wrong folder to this day.