3 ms·
Yeah, I’m not sure where that figure came from either. My experience on an 8GB card is that it is always nudging up close to the limit and you don’t want to tr
by jandrese 3y ago
Yeah, I’m not sure where that figure came from either. My experience on an 8GB card is that it is always nudging up close to the limit and you don’t want to try to turn the knobs up much past the default.
SDXL is a total nonstarter as well on 8GB.
- ColonelPhantom 3y agoUsing vladmandic sd.next with v1-5 works fine* on my RX 580 4GB at around 1s/it on ROCm and 1.5s/it on DirectML. fine = requires --medvram or --lowvram in order to not run out of memory. The speeds above are with medvram, which is just about on the edge of what's possible. These options compromise speed but are still better than running on the CPU generally.
- brucethemoose2 3y agoVoltaML is also quite good on low vram cards.
- ColonelPhantom 3y agoDoes VoltaML have any support for AMD/Intel GPUs or CPU inference? e.g. I think Intel's openVINO supports ONNX, but the Github page only mentions Nvidia and CUDA.
- brucethemoose2 3y agoThe docs are somewhat out of date, and TBH I'm not sure about openVINO I know some are running it with ROCm.