My Community

I am doing some benchmark tests for the newly built of gpu version of vasp 6.2 using a system with 750 atoms. The vasp out was complete and "reached required accuracy - stopping structural energy minimization. " However, the executive files were still running in the server until the time run out as I ssh into the running node. When running small systems below 500, it runs beautifully and has no problem.
Any suggestions would be greatly appreciated to help solve the problem.

The OpenACC GPU port of VASP was compiled using makefile.include.linux_pgi_acc from the vasp wiki website. The compile process works very well.

Please advise. Thanks very much.

Hello,

please upload the corresponding input files and file a bug report as written in the forum posting guidelines.