[Wien] Strange benchmark results

=?gb2312?B?wO698NpT?= lijy07 at lzu.cn
Thu Nov 19 15:39:44 CET 2009


Dear WIEN users:
  I get some strange benchmark results with a new computer (Intel Core i5 750 2.67GHz, 8 GB DDR3 1333 RAM) in our lab. I installed WIEN2k_09.1 with ifort 11.1 and mkl 10.0, set NMATMAX=20000 and  OMP_NUM_THREADS=1, ran serial benchmark and the result is "80.522u 0.724s 1:21.26 99.9%", faster than the results of i7-920 is 99 seconds(NMAT=3481,1 thread) on the benchmark page. When I set NMATMAX=3481 on my computer and recompile lapw1, the result is "32.006u 0.513s 0:32.65 99.5%" with OMP_NUM_THREADS=1, and "34.578u 1.179s 0:20.16 177.2%" with OMP_NUM_THREADS=2, and "44.262u 1.281s 0:15.50 293.8%" with OMP_NUM_THREADS=4. I can't understand how could i5-750 run so fast. My compile options are:

FOPT:-FR -mp1 -w -prec_div -pc80 -pad -align -DINTEL_VML -traceback
FPOPT:$(FOPT)
LDFLAGS:$(FOPT) -L/opt/intel/mkl/10.0.5.025/lib/em64t -pthread
DPARALLEL:'-DParallel'
R_LIBS:-lmkl_lapack -lmkl -liomp5
RP_LIBS:-lmkl -lmkl_lapack -lmkl_intel_lp64 -lmkl_scalapack_lp64 -lmkl_blacs_lp64 -lmkl_sequential -lguide -L /opt/local/fftw/lib/ -lfftw_mpi -lfftw

  In additional, I get a serial benchmark result on another computer(Intel Core2 Q9400 2.67GHz, 8 GB DDR2 800 RAM) is 125 seconds with OMP_NUM_THREADS=1, NMATMAX=13000, ifort 9.1 and mkl 10.0
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20091119/f57a1e22/attachment.htm>


More information about the Wien mailing list