 | |||||||||||||||
 |
 |
 |
|
 |
|
 |
|
 |
|
 |
|
 |
|
 |
 |
AMBER Archive (2001)
Subject: Re: Dual Origin vs dual Alfa
From: Nathalie Godbout (godbout_at_sgihud.hudson.sgi.com)
Date: Wed Feb 14 2001 - 10:18:23 CST
- Next message: Chuck Schneider: "Re: dynamic memory allocation"
- Previous message: Jake Isaacs: "how to handle a metal ion"
- In reply to: Anton Guliaev: "Dual Origin vs dual Alfa"
- Next in thread: Tru Huynh: "Athlon 1GHz (was: Dual Origin vs dual Alfa)"
- Reply: Tru Huynh: "Athlon 1GHz (was: Dual Origin vs dual Alfa)"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Anton Guliaev wrote:
>
> Dear Amber Users -- I am currently using dual 225MhzR10K Origin 200 for
> my MD simulations. I am looking to upgrade this machine with the dual
> 360MhzR12K ($9K). Does anybody use this configuration and know how much
> would I gain in CPU time (30-40%?)? Also, to my knowledge, dual CPU
> Compaq alfa in general outperform the Origin 200 -- is this really true
> for Amber calculations? I would appreciate any feedback..
> Also, how good is dual 400MhzR12K Octane, compare with dual 225MhzR10K
> Origin...
>
> THanks a lot,
>
> Anton
>
> P.S. The benchmarks on Amber page do not cover new 400Mhz or 360Mhz R12K
> CPUs
>
> --
> ************************************
> * Anton B. Guliaev, Ph.D *
> _ * Life Sciences Division * _
> / )* LBNL, Berkeley CA 94720 *( \
> / / * abguliaev_at_lbl.gov * \ \
> _( (_ * T: 510-486-5773 F:(510)-486-6488 * ) )_
> (((\ \)**/ )**************************( \**(/ /)))
> (\\\\ \_/ / \ \_/ ////)
> \ / \ /
> \ _/ \_ /
> / / \ \
> /___/ \___\
Hi Anton,
I've attached a text file with the latest Amber6 benchmarks on most
sgi platforms. You will notice in the table that I have done the
calculations using two versions of the source code; the original source
code and a modified version of the source code. The modified version is
a version of Amber6 that has been made to run faster by making one of
the most computing intensive part of the code (namely subroutine
short_ene) faster. Because this subroutine is in a parallel part of the
code, the scalability is reduced a little but time to solution is
better.
I don't know what type of run you do but the more time you spend in
this subroutine, the more you will benefit from the improvement. This
version will be available in the Amber7 distribution so you can compare
your 225MhzR10k results with the results using the modified version of
the source code. I don't know if the modified version of the code will
run faster than the original source code on other computer
architectures, I only have access to sgi machines ;-)
Let me know if you have questions,
Nathalie
-- ------------------------------------------------------------------------ Nathalie Godbout, Ph.D. SGI godbout_at_hudson.sgi.com Chemistry and Biology Applications Group (781)839-2147 One Cabot Road, Suite 250 (978)562-7450 fax Hudson, MA vnet:483-2147 01749, USA ------------------------------------------------------------------------ Amber6/sander benchmarks across sgi platforms.Nathalie Godbout, sgi 2/14/01
*on sgi1400: - 4xPIII 500MHz/2048KB L2 cache - Linux - PGI compilers - MPICH (1.2) and sgi MPI - original Amber6 source code
*on sgi1200: - 16xPIII (8 nodes) 700MHz/256KB L2 cache - Linux - PGI compilers - MPICH (1.2) - ethernet (2 jobs/node) and myrinet (1 job/node) - original Amber6 source code
*on Origin200: - 4xR12k IP27 360MHz/4MB L2 cache - Irix 6.5 - MIPSpro 7.3.1 compilers - sgi MPI - original and modified Amber6 source code
*on sgi2800: - 32xR12k IP27 400MHz/8MB L2 cache - Irix 6.5 - MIPSpro 7.3.1 compilers - sgi MPI - original and modified Amber6 source code
*on sgi3800: - 128xR12k IP35 400MHz/8MB L2 cache - Irix 6.5.9 - MIPSpro 7.3.1 compilers - sgi MPI - original and modified Amber6 source code PROWAT: ------- sgi1400 sgi1400 sgi1200 sgi1200 origin200 origin200 (500mhz) (500mhz) (700mhz) (700mhz) (360mhz) (360mhz) mpich sgi-mpi ethernet myrinet orig code modif code ------------------------------------------------------------------------ nproc time spup time spup time spup time spup time spup time spup 1 278.72 276.80 187.42 186.14 150.21 129.43 2 148 1.88 145 1.91 103 1.82 99 1.88 79 1.90 68 1.90 4 87 3.20 84 3.30 64 2.93 55 3.38 45 3.34 39 3.32 8 44 4.26 33 5.64 16 34 5.51 21 8.86 -------------------------------------------------------------------------------- sgi2800 sgi2800 sgi3800 sgi3800 (400mhz) (400mhz) (400mhz) (400mhz) code: original modified original modified -------------------------------------------------------------------------------- nproc time spup %P time spup %P time spup %P time spup %P 1 136.20 115.94 133.82 114.96 2 71 1.92(95.83) 61 1.90(94.74) 69 1.94(96.91) 60 1.92(95.83) 4 39 3.49(95.13) 33 3.52(95.45) 38 3.52(95.45) 32 3.59(96.19) 8 22 6.19(95.82) 19 6.10(95.55) 20 6.69(97.20) 18 6.39(96.40) 16 13 10.48(96.49) 12 9.67(95.64) 14 9.56(95.51) 11 10.45(96.46) 32 11 12.38(94.89) 10 11.60(94.33) 8 16.73(97.06) 8 14.37(96.04) 64 6 22.30(97.03) 6 19.16(96.29) --------------------------------------------------------------------------------
DHFR: ----- sgi1400 sgi1400 sgi1200 sgi1200 origin200 origin200 (500mhz) (500mhz) (700mhz) (700mhz) (360mhz) (360mhz) mpich sgi-mpi ethernet myrinet orig code modif code ------------------------------------------------------------------------ nproc time spup time spup time spup time spup time spup time spup 1 441.21 437.73 331.64 330.49 197.67 170.28 2 254 1.74 241 1.82 203 1.63 183 1.81 110 1.80 96 1.77 4 157 2.81 146 3.00 146 2.27 101 3.27 61 3.24 55 3.10 8 106 3.13 60 5.51 16 79 4.20 41 8.06 ------------------------------------------------------------------------------- sgi2800 sgi2800 sgi3800 sgi3800 (400mhz) (400mhz) (400mhz) (400mhz) code: original modified original modified ------------------------------------------------------------------------------- nproc time spup %P time spup %P time spup %P time spup %P 1 166.47 142.39 163.20 138.97 2 94 1.77(87.01) 82 1.73(84.39) 88 1.85(91.89) 75 1.85(91.89) 4 51 3.26(92.43) 45 3.16(91.14) 47 3.47(94.91) 41 3.39(94.00) 8 30 5.55(93.69) 27 5.26(92.56) 26 6.28(96.09) 23 6.04(95.36) 16 20 8.32(93.85) 18 7.89(93.15) 17 9.60(95.56) 16 8.68(94.38) 32 18 9.25(92.07) 18 7.89(90.14) 13 12.55(95.00) 11 12.63(95.05) 64 11 14.84(94.74) 11 12.63(93.54) -------------------------------------------------------------------------------
4096WAT: -------- sgi1400 sgi1400 sgi1200 sgi1200 origin200 origin200 (500mhz) (500mhz) (700mhz) (700mhz) (360mhz) (360mhz) mpich sgi-mpi ethernet myrinet orig code modif code ----------------------------------------------------------------------- nproc time spup time spup time spup time spup time spup time spup 1 223.14 223.04 166.22 167.95 96.70 80.32 2 126 1.77 121 1.84 106 1.57 94 1.79 56 1.73 47 1.71 4 79 2.82 71 3.14 80 2.08 53 3.17 31 3.12 28 2.87 8 56 2.97 33 5.09 16 43 3.86 22 7.63 ------------------------------------------------------------------------------- sgi2800 sgi2800 sgi3800 sgi3800 (400mhz) (400mhz) (400mhz) (400mhz) code: original modified original modified ------------------------------------------------------------------------------- nproc time spup %P time spup %P time spup %P time spup %P 1 85.44 70.53 84.18 69.17 2 45 1.90(94.74) 37 1.91(95.29) 45 1.87(93.05) 37 1.87(93.05) 4 24 3.56(95.88) 21 3.36(93.65) 24 3.51(95.35) 20 3.46(94.80) 8 14 6.10(95.55) 12 5.88(94.85) 14 6.01(95.27) 12 5.76(94.44) 16 10 8.54(94.18) 9 7.84(93.06) 10 8.42(94.00) 8 8.65(94.34) 32 9 9.49(92.35) 8 8.82(91.52) 7 12.02(94.64) 6 11.53(94.27) 64 7 12.02(93.14) 6 11.53(92.78) -------------------------------------------------------------------------------
- Next message: Chuck Schneider: "Re: dynamic memory allocation"
- Previous message: Jake Isaacs: "how to handle a metal ion"
- In reply to: Anton Guliaev: "Dual Origin vs dual Alfa"
- Next in thread: Tru Huynh: "Athlon 1GHz (was: Dual Origin vs dual Alfa)"
- Reply: Tru Huynh: "Athlon 1GHz (was: Dual Origin vs dual Alfa)"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
|
VU Home |
VUMC Home |
People Finder |
University Calendar
webmaster- modified on January 30, 2009 |