QuantumATK Forum
QuantumATK => General Questions and Answers => Topic started by: joon9885 on November 21, 2018, 03:48
-
Dear experts,
Hello. From some time ago, I started to run a PLDOS calculation of Tellurium nanowire FET, which is built based on the article 'Silicon nanowire field-effect-transistor' at Quantumwise webpage (https://docs.quantumwise.com/tutorials/silicon_nanowire/silicon_nanowire.html). When I repeated the silicon nanowire example with our lab's computer, it ran well without error. However, with my newly created one it continuously put error message 'Out-Of-Memory' and very recently the program just stop without any error message. So, I tried to lower the energy or do not add the VanderWaals effect or the basis to lower the memory but it did not quite work. I used MPI with 4 cores so that it could share the memory and I did not try the serial calculation yet. The log file and script file is posted. Would you mind finding what is the problem here?
Thanks!
-
It is a pretty big system, and not something I would expect to be possible to run on a laptop. It is not really comparable to the quite small and simple system in the tutorial. I suggest you try to run it on a cluster with significantly more processes and memory. Note that for a system with this much vacuum, the grid is probably dominating the memory consumption, so if you could have less vacuum, that would also make it more feasible.
-
Hello.
Firstly thank you for the reply.
Actually, I did calculations on our lab's server and previously when I assigned 22 cores it also got out-of-memory. Anyway, I will modify some structure and try it again.
Thanks!
-
Unfortunately, for these big systems, the grid can take quite a bit of memory. One thing you could also try is to change the Poisson solver to the [Parallel] Conjugate Gradient. It should lower the memory consumption quite a bit when running on many cores. See also the short discussion of the different Poisson Solvers here: https://docs.quantumwise.com/tutorials/inas_p-i-n_junction/inas_p-i-n_junction.html#choice-of-the-poisson-solver