Skip to content

Latest commit

 

History

History
22 lines (17 loc) · 1.04 KB

README.md

File metadata and controls

22 lines (17 loc) · 1.04 KB

Stable Diffusion with 🤗 Diffusers

Inference Performance Comparison between GPU, CPU, and Apple M1/M2

Our results: Average time for one (diffusion) step (measured in seconds)

Hardware (512, 512) (768, 768)
Nvidia RTX 8000 0.23s 0.81
Apple M1 Pro (16‑Core GPU) 1.30s 6.67
Apple M1 Pro (10‑Core CPU) 4.77s 15.41

Average total time to generate one image using 200 diffusion steps:

  • Nvidia RTX 8000: ~45s / ~2.5 min
  • Apple M1 Pro (16‑Core GPU): ~4.3min / ~22min
  • Apple M1 Pro (10‑Core CPU): ~15min / ~51min

Details

  • We run the diffusion process for 50, 150 and 200 diffusion steps calculated the average time for one step
  • Each run was repeated 5 times and results were again averages
  • We did not use half-float precision, since it is not (yet) available for MPS at the moment (should also mainly affect memory allocation)
  • Going beyond 768 was not able for the M1 with 32GB shared memory