My cpu is not Fully used

[AymanLyesri]

Hi, am a new stable diffusion user, using this technologie has help me a lot understanding how image generation works.

One thing i dont understand is that when i use CPU ONLY it only uses 50% of my cpu capabilities, i need like 30 min to generate a 512x512 image with 75 steps and 7.5 guidance scale, and i think its a ridiculous amount of time for one image.

CPU : AMD ryzen 3400G

RAM : 16gb

GPU : Radeon RX vega 11 grafics

VRAM : 2gb

I hope somebody tells me if thats normal, is there a way to generate an image using 100% of my CPU,

Could i use the vram of my GPU to minimise the amount of time needed.

[michaeljanich]

I read that PyTorch supports OpenMP, that way it should be significantly faster with multi-core CPUs. Try to enable that.

[Night-Chan]

This seems like a bottleneck by RAM speeds. An external GPU wont face this bottleneck, vRAM is fast enough to deliver all the data thats needed, but i suspect your RAM isnt fast enough to supply all data.

[Paul-Williams]

I have the same issue (much later). However if I run two instances on different ports, then the CPU gets maxed.

PS. My RAM is running at 2400Mhz - CPU is i7-7700K