Dear Amber experts,
in our group, we are using several different simulation techniques.
Since updating to Amber 22, we are experiencing memory (RAM) problems.
In a simulation of about 2 days walltime, Amber fills up our 32 GB of
RAM. The leak happens with aMD, REMD-TI, and cMD, although it is much
slower with cMD. All the simulations are running using pmemd.cuda,
except REMD, which is using MPI as well.
The same leak happens on different computers in our system. The tested
compiler / CUDA combinations are:
(Debian 11) gcc-10.2.1 + cuda-11.4
(Debian 10) gcc-8.3.0 + cuda-11.2
(Debian 9) gcc-6.3.0 + cuda-11.6
(Debian 9) gcc-6.3.0 + cuda-11.2
Is anyone else experiencing similar issues? Is there something else that
we could try to fix the problem?
Best regards,
Franz Waibl
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Sep 20 2022 - 06:00:02 PDT
