3.0
@hn_d7840b
25 days ago
Yeah, I was actually wanting to pick up a second 3090 recently to run Stable Diffusion faster. But it apparently doesn't support multi-GPU.
Open-source latent diffusion text-to-image model that runs locally on consumer GPUs; released under a permissive license by Stability AI with academic collaborators.
@hn_d7840b
25 days ago
Yeah, I was actually wanting to pick up a second 3090 recently to run Stable Diffusion faster. But it apparently doesn't support multi-GPU.
@hn_21f8af
26 days ago
FWIW, you can definitely run stable diffusion on your 3090. I played around with it on my 3060m (laptop) and it works surprisingly well. The models are pretty tiny as well, a few GB. I think you could get pretty far with your setup if you wanted too
@hn_b2ba07
26 days ago
I have a dated 1070 with 8gb of vram, some of which also renders my desktop. I was able to obtain 256x512 images with this card using the standard model, but ran into OOM issues. I don't mind waiting, so now I am using the "fast" repo: https://github.com/basujindal/stable-diffusion With this, it takes 30s to generate a 768x512 image (any larger and I am experiencing OOM issues again). I think you should expect a bit faster at the same resolution with your 3060 because it's a faster card with the same amount of memory.
@hn_295402
26 days ago
LLMs perhaps (I'm not sure either way, everything moves too quickly), but SDXL 1.0 (July 26, 2023) was a lot better than DALL•E 2 (6 April, 2022). I think DALL•E 3 (August 10, 2023) is a bit better than SDXL, but other than text generation their quality seems very close to me. (That said, perhaps I'm Clever Hands-ing myself by only using SDXL for what it's good at. It's terrible at dragons every time I've tried that…)
@hn_fbed32
26 days ago
The analogy falls apart when you consider that for instance stable diffusion was trained on 2.3 billion images, but the model itself is only around 10GB. That only works out to be 4 bytes per image. Clearly it's not the same as pointing a camcorder at a movie screen.