<html><head><meta http-equiv="content-type" content="text/html; charset=utf-8"></head><body style="overflow-wrap: break-word; -webkit-nbsp-mode: space; line-break: after-white-space;"><div>Hi Folks,</div><div><br></div><div>I installed a new i9-13900k and I was just curious to see how the P cores and E cores compared. I did an SCF of a phyllosilicate using each and I find that it converges nicely in all cases but that the number of iterations to reach convergence is different. I’m just curious why.</div><div><br></div><div>Run on the 16 E cores only:</div><div><br></div><div><font color="#000000"><span style="caret-color: rgb(0, 0, 0);">mpirun —cpu-set 8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23 —bind-to core -n 16 pw.x < Cronstedtite.scf | tee Cronstedtite.scf.16Ecore.out</span></font></div><div><span style="caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0);"><br></span></div><div>Run on the 8 P cores only:</div><div><br></div><div><span style="caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0);">mpirun —cpu-set 0,1,2,3,4,5,6,7 —bind-to core -n 8 pw.x < Cronstedtite.scf | tee Cronstedtite.scf.8Pcore.out</span></div><div><span style="caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0);"><br></span></div><div>Run on all 24 cores:</div><div><br></div><div>mpirun —bind-to core -n 24 pw.x < Cronstedtite.scf | tee Cronstedtite.scf.8P16Ecore.out</div><div><br></div><div>I find 1 hour for the E cores, 53 mins for the P cores, and 29 mins for all cores — which is practically linear scaling.</div><div><br></div><div>However, 22 iterations to converge on E cores, 26 to converge on P cores, and 15 to converge on all cores. It looks like the more cores the fewer the number of convergence steps? </div><div><br></div><div>(Input and output files and a spreadsheet attached if anyone is interested.)</div><div><br></div><div>Thanks,</div><div><br></div><div>Zack Gainsforth</div><div>Space Sciences Laboratory</div><div>University of California, Berkeley</div><div><br></div><div></div></body></html>