Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

On a related-different axis, I've consistently seen on-prem GPUs running identical workloads ~35% faster than the same workloads on the same cloud hardware, regardless of intermediate infra stack layering/versioning choices. Weird but I'm not complaining!
 help



They might be undervolting and underclocking the GPUs to improve longevity?



Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: