Hi Ahmed,
Conda packages are precompiled binaries. There is no sane way of working around a glibc incompatibility between binaries. Given that you are already compiling 'pmemd.cuda', it would be easiest to follow the manual instructions and compile all of Amber.
The conda package is more of a convenient way to get, say, cpptraj working on your laptop. When installing amber on a cluster, is better to do a proper install.
Having said that, you mentioned you're using gcc15. That's a pretty recent gcc version, and it requires CUDA 13, which Amber does not support, and may not even be available on your cluster.
Check your available CUDA toolkit versions, pick the latest 12v you can find, check its compatibility table<
https://docs.nvidia.com/cuda/archive/12.5.0/cuda-installation-guide-linux/index.html>, and try to use the indicated gcc version, or older.
Let me know how it goes.
Best,
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Jul 06 2026 - 06:30:03 PDT
