diff --git a/comfy/model_management.py b/comfy/model_management.py
index bbfc3c7a1..c5b2b4ea9 100644
--- a/comfy/model_management.py
+++ b/comfy/model_management.py
@@ -1355,7 +1355,7 @@ def supports_fp8_compute(device=None):
     if SUPPORT_FP8_OPS:
         return True
 
-    if not is_nvidia():
+    if not (is_nvidia() or is_amd()):
         return False
 
     props = torch.cuda.get_device_properties(device)