Hello everyone, i had a 3080ti 12gb and added a 3080 20gb, so it has a bit less speed but more memory than my main card. I could finally get some speed with the usual suspects (i am testing gemma 4 31b/26b-a4b and qwen 3.6 27b/35b-a3b), BUT to some good speed I have to fit the whole dataset (weight,
