upstream-archive/ComfyUI
1
0
Fork 0
mirror of https://github.com/Comfy-Org/ComfyUI.git synced 2026-09-21 13:38:08 -05:00
Code Issues Packages Projects Releases Wiki Activity

main: bump the AMD Windows VA quota to 4TB (#16199)

Browse Source 
AMD+windows sets an unusually low maximum GPU virtual memory quota
compared to other plaforms. For reference, the hardware limits on
XT9060 are 128TB. On Nvidia RTX3060 it is 1TB. On nvidia RTX5090
it is 128TB.

Bump this into the same league, but keep it to a modest 4TB pending
further usage case need + testing going even higher to the hardware
limits.
This commit is contained in:
rattus
2026-09-09 08:51:34 -07:00
committed by GitHub
parent 54e03f5367
commit 6f3895ed8d
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
main.py
+1 -1
View File
@@ -111,7 +111,7 @@ if __name__ == "__main__":
os.environ['CUBLAS_WORKSPACE_CONFIG'] = ":4096:8"
if "rocm" in cuda_malloc.get_torch_version_noimport():
os.environ['OCL_SET_SVM_SIZE'] = '262144' # set at the request of AMD
os.environ['OCL_SET_SVM_SIZE'] = '4194304' # 4TB. Much larger than the ROCM 64GB/256GB defaults for Aimdos liberal VA use
def handle_comfyui_manager_unavailable():