(www.tomshardware.com)

521 points hd4 | 2 comments | 20 Oct 25 12:31 UTC | HN request time: 0.426s | source

1. mighmi ◴[20 Oct 25 17:42 UTC] No.45646800[source]▶

To what extent is this practice applicable to other loads?

replies(1): >>45652382 #

2. wsfung2008 ◴[21 Oct 25 04:06 UTC] No.45652382[source]▶

This is for platforms that serve many different models, most of which have very low usage. e.g. huggingface, civitai

Alibaba Cloud says it cut Nvidia AI GPU use by 82% with new pooling system