You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi Oleg, and thanks for the great portal.
My 2080 is not working at expected speeds.
I have 9-10it/s, but Comfy only returned a picture every 5 seconds.
I tied to add --hihram in comfy arguments and lowered it around 1 sec.
[rgthree] Using rgthree's optimized recursive execution.
Requested to load SDXL
Loading 1 new model
100%|████████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 6.94it/s]
Prompt executed in 1.16 seconds
got prompt
Using SDXL lightning and seems Comfy is using most of the time reloading the model each iteration
Tried SDXL Turbo with LCM Lora but it throws a
!!! Exception during processing !!!
Rather new to Comfy coming from a1111, but great to be in noodle mode again.
Cheers, Boya
The text was updated successfully, but these errors were encountered:
Did a fresh Comfy install, no custom nodes. Updated gpu drivers.
SD1.5 models run better: 10it/s but 0.6 sec to return to TD.
Comfy still reloads SDXL Turbo / Lightning every it. at around 3 seconds
I see you are getting almost 20 fps.
Could I ask with what card and what models, and could you share those exact .jsons for me to test?
I understand I need to switch my 2080 at some point, but would like to se this running smoother first.
Hi Oleg, and thanks for the great portal.
My 2080 is not working at expected speeds.
I have 9-10it/s, but Comfy only returned a picture every 5 seconds.
I tied to add --hihram in comfy arguments and lowered it around 1 sec.
[rgthree] Using rgthree's optimized recursive execution.
Requested to load SDXL
Loading 1 new model
100%|████████████████████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 6.94it/s]
Prompt executed in 1.16 seconds
got prompt
Using SDXL lightning and seems Comfy is using most of the time reloading the model each iteration
Tried SDXL Turbo with LCM Lora but it throws a
!!! Exception during processing !!!
Rather new to Comfy coming from a1111, but great to be in noodle mode again.
Cheers, Boya
The text was updated successfully, but these errors were encountered: