As I am exploring material simulations one really hard truth that I have realized is that my gaming laptop is not at all going to cut it for my research work. Currently I am trying to find the absorption spectra of a quantum dot which has 45 atoms in total and I am using the PBE functional (so not that much accuracy). The geometrical optimization converged after 10-12 hours by running the executable for absorption spectra (turbo_lanczos.x) did not go that well. The computation ran for 23 hours after which my laptop gave out.

So I am looking for clusters which do not burn a big hole in my pocket. AWS and GCS came first to my mind. Although I found a tutorial for running on AWS I am not sure how to know how many cores, nodes, RAM or instances I need. I was also not able to find any such documentation for Google Cloud.

Could you guys help me with some guidance?

  • 1
    $\begingroup$ Does this answer your question? What are some cloud services for computing? $\endgroup$ Jul 1, 2021 at 13:38
  • 2
    $\begingroup$ Does this help you? mattermodeling.stackexchange.com/q/1122/88 $\endgroup$
    – Thomas
    Jul 1, 2021 at 13:45
  • 2
    $\begingroup$ +1 but Is your question "how many cores/nodes, and how much RAM do I need?" or is it "is there a tutorial for using Google Cloud?" $\endgroup$ Jul 1, 2021 at 15:26
  • 1
    $\begingroup$ If youre low on RAM and computational resources, I believe using linearly scaling codes like SIESTA would be a good alternative. Moreover since your working on a quantum dot have you tried codes utilizing a local basis like gaussian: researchgate.net/post/… or ORCA (its opensource) $\endgroup$ Jul 2, 2021 at 9:09
  • 1
    $\begingroup$ @ParmeetSinghEP066 That depends on the elements your considering. It would be expensive if you're trying to simulate quantum dots of heavy elements. I'm not a fluent user of Gaussian so I'm not entirely sure! :) $\endgroup$ Jul 3, 2021 at 22:27