10-15-2026, 11:40 AM
Tried to reproduce it just now. On dual 3090 with SGLang, I saw basically the same behavior. The surprise was that memory pressure showed up before throughput collapsed.
|
first LoRA that actually helped on B200
|
|
« Next Oldest | Next Newest »
|
| Messages In This Thread |
|
first LoRA that actually helped on B200 - by fusedhall - 10-08-2026, 11:58 PM
Re: first LoRA that actually helped on B200 - by cacheline - 10-09-2026, 02:39 AM
Re: first LoRA that actually helped on B200 - by batchsize - 10-09-2026, 02:51 AM
Re: first LoRA that actually helped on B200 - by batchfield55 - 10-12-2026, 02:58 AM
Re: first LoRA that actually helped on B200 - by quiethouse - 10-12-2026, 03:34 AM
Re: first LoRA that actually helped on B200 - by batchsize - 10-15-2026, 08:24 AM
Re: first LoRA that actually helped on B200 - by batchsize - 10-12-2026, 04:38 AM
Re: first LoRA that actually helped on B200 - by benchplease - 10-13-2026, 05:17 PM
Re: first LoRA that actually helped on B200 - by northgrid33 - 10-13-2026, 07:46 PM
Re: first LoRA that actually helped on B200 - by cudascope - 10-15-2026, 11:13 AM
Re: first LoRA that actually helped on B200 - by quiethouse - 10-15-2026, 11:40 AM
|
Users browsing this thread: 1 Guest(s)