Miscellanea

Home IH site Publications Software Blog e-mail ReadmeMiscellaneaCurriculum Valid HTML 4.01! Valid CSS! CC License
Music | FLOSS | Logic | Gaussian

I hope the info in this page will be helpful to the reader...

Tips on calculation efficiency | Various Benchmarks | Troubleshooting

Benchmarks

Computers

Eneas Pentium 4 1.7 GHz desktop computer, running Debian Sarge
Used as reference (1GHz = 1.7 times the elapsed time of Eneas).
Arina Itanium 2 cluster node, running Red Hat.
Ulises AMD Athlon 2800+ desktop PC, running Debian Sarge.
Bart Pentium 4 3.2GHz desktop computer, running Debian Etch.
Maggie Intel Core 2 Duo T7200 2.0GHz laptop, running Ubuntu 6.06 LTS AMD64.
Mopsus Pentium D 2.8GHz desktop computer, running Fedora Core 4 x86_64.
Opteron AMD Opteron 250 cluster node, running Red Hat Advanced Server 4.

Benchmark 01

bench_01 C5v Ferrocene RHF/6-311G(d,p) Optimization
21 atoms
286 basis functions
481 primitives
96 electrons

10,831 sec @ 1GHz
95% CPU
xmgrace

Comments

This takes longer with G03, maybe because the electronic state is not well described as a RHF singlet, and G03 takes some shortcuts in the optimization which fail to save time under these circumstances.

Benchmark 02

bench_02 CH3B(CCl3)3- MP2/6-311G(d,p) Single Point
17 atoms
342 basis functions
625 primitives
186 electrons

11,540 sec @ 1GHz
79% CPU
xmgrace

Comments

This benchmark stresses the disk I/O. It creates a 11GB rwf file (if your filesystem does not support files of more than 2GB, you'll have to split the rwf into 6 fragments).

Benchmark 03

bench_03 C3v CH3B(CF2Cl)3- B3LYP/6-311G(d,p) Optimization
17 atoms
294 basis functions
523 primitives
138 electrons

9,731 sec @ 1GHz
97% CPU
xmgrace

Comments

None

Benchmark 04

bench_04 C3v CH3B(CF2Cl)3- B3LYP/6-311G(d,p) Frequency
17 atoms
294 basis functions
523 primitives
138 electrons

19,754 sec @ 1GHz
95% CPU
xmgrace

Comments

None

Benchmark 05

bench_04 CH3B(C6F5)3- RHF/6-31G Optimization
38 atoms
321 basis functions
782 primitives
258 electrons

30,414 sec @ 1GHz
98% CPU
xmgrace

Comments

None

Benchmark 06-n

06-n
bench_06
Proline n-mer B3LYP/6-31G Optimization
14n+3 atoms
77n+13 basis functions
182n+30 primitives
52n+10 electrons
Used at Arina to test the effect of %mem in the optimization time. From the tetramer on, only 10 opt cycles have been carried out, not untill full convergence. The conclusions of such test are outlined in this section of the Gaussian Tips
06b-6
bench_06
Proline hexamer B3LYP SP with 4 different basis sets
87 atoms
691, 863, 1254 and 2469 basis functions
322 electrons
Used at Arina to test the effect of %mem in the parallel calculation. It looks like more CPUs (with shared memory) require more memory to do the same job, as can be seen in this section of the Gaussian Tips

Benchmark 07

bench_04 CS CnCl2n+2 B3LYP/6-311+G* Frequency
3n+2 atoms
82n+60 basis functions
142n+106 primitives
40n+34 electrons

Comments

Used at tx41 to test the effect of %mem in the frequency calculation time. The conclusions of such test are outlined in this section of the Gaussian Tips page.

Benchmark 08

bench_03 Peniciline B3LYP/6-31G Optimization
38 atoms
234 basis functions
572 primitives
168 electrons

~35,000 sec @ 1GHz
99% CPU
xmgrace

Comments

None