This workflow is specifically made for benchmarking AMD Ryzen AI Max+ 395 systems. It offers very fast, good quality and large image outputs even on sparse (<256GB/s) bandwidth. The workflow model can be used for SFW or NSFW images. However, it should be suitable and fast for anyone else as well.

Model:

Juggernaut XL - Ragnarok_by_RunDiffusion | Stable Diffusion XL Checkpoint | Civitai

https://huggingface.co/RunDiffusion/Juggernaut-X-v10/tree/main

Nodes:

Res4Lyf

Comfy-Easy-Use

1024 x 1024 Settings + Gen Times for AMD Ryzen AI Max+ 395

-------

Middle distance shots with ok detail - Time 18s

Steps 10 - CFG 8 - Sampler res_2 - Scheduler beta57 or beta

-------

Overall ok quality - Time: 18-25s

Steps 25-30 - CFG 8 - Sampler dpmpp_3m_sde_gpu - Scheduler exponential

-------

Vintage head shots, hazy old picture quality feel - Time: 9s

Steps 10 - CFG 4-6 - Sampler lcm - Scheduler bong_tangent

-------

Clear, good quality head shots - Time: 33s

Steps 20 - CFG 4-6 - Sampler ttm - Scheduler sgm_uniform

-------

Fast decent head shots - Time: 12s

Steps 12 - CFG 5 - Sampler rk - Scheduler simple

-------

Nice vintage landscapes - Time: 9s

Steps 10 - CFG 5 - Sampler er_sde - Scheduler ddim_uniform or karras