[gmx-users] Benchmark performance question...

[Florian Haberl]

hi

On Thursday 08 December 2005 13:07, Stéphane Teletchéa wrote:
> Since my mail did not get much answer, i presume it was not clear enough...
> 
> I'm benchmarking my system and i had these results (for villin as in 
> gmxbench-3.0.tar.gz) :
>                        NbCPU     Time (real)        Perf (ns/day)
> single precision :     1          109.00           8.463
> single precision :     2          102.00           8.471
> double precision :     1          150.00           5.915
> double precision :     2          166.00           5.205
> 
got some more from some new machines ... still a bit incomplete
all in single p.

d.dppc


				time		gflops	ns/day	
gorgo	gcc	4	24:40		2.859	0.584
gorgo	gcc	2	37:39	1.873	0.382
gorgo	gcc	1			
gorgo	icc	4	24:35		2.869	0.586
gorgo	icc	2	39:21	1.793	0.366
gorgo	icc	1	64:17	1.097	0.224
snode	gcc	2			
snode	gcc	1			
snode	icc	2	41:00	1.721	0.351
snode	icc	1	67:04	1.052	0.215
opteron	icc	8	16:58	4.161	0.849
opteron	icc	4	24:34	2.871	0.586





d.lzm pme

				time		gflops	ns/day
gorgo	gcc	4	4:17		4.934		6.724
gorgo	gcc	2	8:05		2.514	3.563
gorgo	gcc	1	12:41	1.570	2.270
gorgo	icc	4	4:18		4.915	6.698
gorgo	icc	2	7:19		2.778	3.936
gorgo	icc	1	12:55	1.542	2.229
snode	gcc	2	7:42		2.639	3.740
snode	gcc	1			
snode	icc	2	8:34		2.108	4.431
snode	icc	1	12:51	1.549	2.239
opteron	icc	8	2:50		4.839	10.165
opteron	icc	4	4:34		4.628	6.307


d.lzm cutt off

				time	gflops	ns/day
gorgo	gcc	4	2:59	4.594		9.654
gorgo	gcc	2	4:46	2.874	6.042
gorgo	gcc	1	8:26	1.622	3.410
gorgo	icc	4	2:55	4.699	9.874
gorgo	icc	2	4:41	2.926	6.149
gorgo	icc	1	8:26	1.623	3.412
snode	gcc	2			
snode	gcc	1			
snode	icc	2	
snode	icc	1	8:37	1.588	3.338
opteron	icc	8	
opteron	icc	4	3:01	4.543		9.547


d.villin

				time	gflops	ns/day
gorgo	gcc	4	1:02	2.367	13.935
gorgo	gcc	2	0:41	3.491	20.571
gorgo	gcc	1	1:14	1.981	11.671
gorgo	icc	4	0:37	3.967	23.351
gorgo	icc	2	0:43	3.411	20.093
gorgo	icc	1	1:14	1.975	11.639
snode	gcc	2	0:45	3.258	19.200
snode	gcc	1	1:17	1.887	11.117
snode	icc	2	0:48	3.055	18.000
snode	icc	1	1:18	1.868	11.009
opteron	icc	8	not running
opteron	icc	4	0:28	5.242	30.857


gorgo = quad AMD Opteron(tm) Processor 848 with 2,2ghz 16gb ram
snode = dual Intel(R) Xeon(TM) CPU with 3.20GHz 2gb ram
opteron = dual Dual Core AMD Opteron(tm) Processor 270 with 2gb ram

Linux sfront03 2.6.14.2 #1 SMP Fri Nov 18 15:51:25 CET 2005 x86_64 GNU/Linux

gcc is 3.4.3 for 64 bit and normal mpich
icc is intel 8.1 with mpich 1.2.6 also for 64 bit
with normal compiling flags ...



> Is it possible to have only a small increase (single precision) on 1 vs 
> 2 processors (single precision) or do i need to multiply by nb of 
> processors (two) the reported performance for 2-cpus ?
> 
> Is there a file where this total performance is stored ?
> 
> Stéphane
> 
> System :
> Pentium IV 2*2.6 Ghz SMP
> 1GB of memory
> Gromacs 3.3 compiled with single/double precision with/without MPI 
> (mpich 1.2.5.2-5mdk)
> MandakeLinux LE2005, gcc 3.4.3-7mdk, kernel 2.6.11-6mdksmp
> 


greetings,
florian

-- 
-------------------------------------------------------------------------------
 Florian Haberl                        Universitaet Erlangen/ Nuernberg
                                       Computer-Chemie-Centrum   
                                       Naegelsbachstr. 25
                                       D-91052 Erlangen
 Mailto: florian.haberl AT chemie.uni-erlangen.de
-------------------------------------------------------------------------------