Zeitpunkt Nutzer Delta Tröts TNR Titel Version maxTL Mi 03.07.2024 00:02:08 2.880 0 374.021 129,9 Toot Café 4.2.9 500 Di 02.07.2024 00:00:32 2.880 0 373.842 129,8 Toot Café 4.2.9 500 Mo 01.07.2024 00:02:14 2.880 0 373.716 129,8 Toot Café 4.2.9 500 So 30.06.2024 00:01:10 2.880 0 373.617 129,7 Toot Café 4.2.9 500 Sa 29.06.2024 00:01:28 2.880 0 373.506 129,7 Toot Café 4.2.9 500 Fr 28.06.2024 00:01:15 2.880 0 373.331 129,6 Toot Café 4.2.9 500 Do 27.06.2024 00:01:23 2.880 0 373.178 129,6 Toot Café 4.2.9 500 Mi 26.06.2024 00:01:13 2.880 0 373.014 129,5 Toot Café 4.2.9 500 Di 25.06.2024 00:00:08 2.880 0 372.803 129,4 Toot Café 4.2.9 500 Mo 24.06.2024 00:00:14 2.880 0 372.619 129,4 Toot Café 4.2.9 500
Baldur Bjarnason (@baldur) · 04/2017 · Tröts: 25.869 · Folger: 6.044
Mi 03.07.2024 15:59
For some context, the study uses AWS-hosted NVIDIA A100-SXM4-80GB GPUs, just like what’s used in most cloud-hosted generative model services. The power usage for a single GPU like that is around 400w (the instance has 8). For just one of the associated Intel CPUs the draw is around 200-240w AFAICT.
The biggest Macbook Pro power supply, for the 16 inch, is a 140w one, to give you some comparison.
[Öffentlich] Antw.: 0 Wtrl.: 0 Fav.: 0 · via Toot!