Search Results
8/9/2025, 4:07:37 PM
>>106198896
I got a slight improvement from the q8 GGUF in one test but the perf is worse and it's almost crashing my computer. so I'm giving up on it, will have to stick to fp8.
also, I tried q6 and it took just as much VRAM as q8 and was just as crash-prone? what the fuck is the point of lower quants then? is this a problem with rocm and/or the 7900 xtx architecture?
I got a slight improvement from the q8 GGUF in one test but the perf is worse and it's almost crashing my computer. so I'm giving up on it, will have to stick to fp8.
also, I tried q6 and it took just as much VRAM as q8 and was just as crash-prone? what the fuck is the point of lower quants then? is this a problem with rocm and/or the 7900 xtx architecture?
Page 1