- Joined
- Jul 20, 2002
I've been running some plink test jobs through SGE via qsub and comparing the results.
The plink test job on an old AMD Opteron @ 2.2 Ghz (2 GiB of RAM @ 400 Mhz) took 3 hr.s 8 minutes to run, while on a Intel Xeon X5675 @ 3.06 Ghz (96 GiB of RAM @ 1067 Mhz) it took 2 hrs. Plink is a single-process task so it only ran on one core in each case. What shocks me is that the Xeon X5675 was only 34% faster on this single-process task.
Instead of using the pre-compiled and tarred binary, would compiling the plink binary on the Intel Xeon X5675 w/all optimizations enabled have made any difference in the performance?
The plink test job on an old AMD Opteron @ 2.2 Ghz (2 GiB of RAM @ 400 Mhz) took 3 hr.s 8 minutes to run, while on a Intel Xeon X5675 @ 3.06 Ghz (96 GiB of RAM @ 1067 Mhz) it took 2 hrs. Plink is a single-process task so it only ran on one core in each case. What shocks me is that the Xeon X5675 was only 34% faster on this single-process task.
Instead of using the pre-compiled and tarred binary, would compiling the plink binary on the Intel Xeon X5675 w/all optimizations enabled have made any difference in the performance?