NEC のサーバーを買ったので、ベンチマークしてみました。

ついでに VPS とか Mac のベンチマークもしてみました。

参考 : Linuxでベンチマーク – superπ/姫野ベンチ編 | データセンターの専用サーバ データセンターダイレクト ブログ
 

確認したマシンたち

CPU とメモリは cpuinfo とmeminfo で調べました。

CentOS の場合

# cat /proc/cpuinfo ← CPUの情報を表示
# cat /proc/meminfo ← メモリの情報を表示

Mac の場合

$ sysctl -a | grep cpu ← CPUの情報を表示
$ sysctl -a | grep mem ← メモリの情報を表示

以下、確認したマシンのスペック。

 

NEC Express5800/S70 タイプRB

CPU : Intel(R) Pentium(R) CPU G6950 @ 2.80GHz
メモリ : 6GB(2GB + 2GB + 1GB + 1GB)
参考 : 仕様:S70 タイプRB | NEC

 

Dell PowerEdge 430

CPU : Intel(R) Pentium(R) D CPU 2.80GHz
メモリ : 1GB(512MB + 512MB)
参考 : PowerEdge SC430 製品詳細 | Dell 日本

 

MacBook (Late 2007)

CPU : Intel(R) Core(TM)2 Duo CPU T7500 @ 2.20GHz
メモリ : 4GB(2GB + 2GB)
参考 : MacBook (Late 2007) – Technical Specifications

 

さくらのVPS

CPU : Intel(R) Core(TM)2 Duo CPU T7700 @ 2.40GHz
メモリ : 512MB
参考 : さくらのVPS

 

ServersMan@VPS(Entryプラン)

CPU : Intel(R) Xeon(R) CPU L5520 @ 2.27GHz
メモリ : 768MB
参考 : ServersMan@VPS

 

superπでベンチマークする方法

下記からsupperπをダウンロードして、
解凍して、実行するだけ。

ftp://pi.super-computing.org/ の一覧

# mkdir supper_pi
# cd supper_pi
# wget ftp://pi.super-computing.org/Linux_jp/super_pi-jp.tar.gz
# tar zxvf super_pi-jp.tar.gz
# ./supper_pi
 Version 2.0 of the super_pi for Linux OS
 Fortran source program was translated into C program with version 19981204 of
 f2c, then generated C source program was optimized manually.
 pgcc 3.2-3 with compile option of "-fast -tp px -Mbuiltin -Minline=size:1000 -Mnoframe -Mnobounds -Mcache_align -Mdalign -Mnoreentrant" was used for the
 compilation.
 ------ Started super_pi run : 2011年 2月 1日 火曜日 20:56:51 JST
 Parameter(%i) to super_pi is missing. Parameter value ? 20 ← 20 を入力(20 = 104 万桁)

 

姫野ベンチでベンチマークする方法

下記から C, static allocate version の M をダウンロードして、
解凍して、リネームして、make して、実行する。
(なんでリネームするんだろ……。)

ダウンロード (ソース & 実行形式(Win/Mac 用))|HPC High Performance Computing|ACCC. RIKEN

# mkdir himeno
# cd himeno
# wget http://accc.riken.jp/assets/files/himenob_loadmodule/cc_himenoBMTxp_m.lzh
# lha e cc_himenoBMTxp_m.lzh
# mv himenobmtxps.c himenoBMTxps.c
# make
# ./bmt

lha(lzhを解凍するコマンド)がない場合は、
yum か rpm でインストールする。

yum の場合。

# yum search lha
# yum -y install lha
# whereis lha
lha: /usr/bin/lha

rpm の場合。
下記から合ったものをダウンロードしてインストール。

Index of /lha

# wget http://packages.sw.be/lha/lha-1.14i-19.2.2.el5.rf.i386.rpm
# rpm -ivh lha-1.14i-19.2.2.el5.rf.i386.rpm
# whereis lha
lha: /usr/bin/lha

 

supperπの結果

NEC Express5800/S70 タイプRB

 Version 2.0 of the super_pi for Linux OS
 Fortran source program was translated into C program with version 19981204 of
 f2c, then generated C source program was optimized manually.
 pgcc 3.2-3 with compile option of "-fast -tp px -Mbuiltin -Minline=size:1000 -Mnoframe -Mnobounds -Mcache_align -Mdalign -Mnoreentrant" was used for the
 compilation.
 ------ Started super_pi run : 2011年 2月 1日 火曜日 20:56:51 JST
 Parameter(%i) to super_pi is missing. Parameter value ? 20
 Start of PI calculation up to 1048576 decimal digits
 End of initialization. Time=       0.204 Sec.
 I= 1 L=       0        Time=       0.588 Sec.
 I= 2 L=       0        Time=       0.674 Sec.
 I= 3 L=       1        Time=       0.674 Sec.
 I= 4 L=       2        Time=       0.676 Sec.
 I= 5 L=       5        Time=       0.675 Sec.
 I= 6 L=      10        Time=       0.674 Sec.
 I= 7 L=      21        Time=       0.675 Sec.
 I= 8 L=      43        Time=       0.675 Sec.
 I= 9 L=      87        Time=       0.675 Sec.
 I=10 L=     174        Time=       0.675 Sec.
 I=11 L=     349        Time=       0.675 Sec.
 I=12 L=     698        Time=       0.674 Sec.
 I=13 L=    1396        Time=       0.675 Sec.
 I=14 L=    2794        Time=       0.671 Sec.
 I=15 L=    5588        Time=       0.670 Sec.
 I=16 L=   11176        Time=       0.666 Sec.
 I=17 L=   22353        Time=       0.655 Sec.
 I=18 L=   44707        Time=       0.634 Sec.
 I=19 L=   89415        Time=       0.590 Sec.
 End of main loop
 End of calculation.    Time=      13.291 Sec.
 End of data output.    Time=       0.067 Sec.
 Total calculation(I/O) time=      13.358(       0.558) Sec.
 ------ Ended super_pi run : 2011年 2月 1日 火曜日 20:57:06 JST

Dell PowerEdge 430

 Version 2.0 of the super_pi for Linux OS
 Fortran source program was translated into C program with version 19981204 of
 f2c, then generated C source program was optimized manually.
 pgcc 3.2-3 with compile option of "-fast -tp px -Mbuiltin -Minline=size:1000 -Mnoframe -Mnobounds -Mcache_align -Mdalign -Mnoreentrant" was used for the
 compilation.
 ------ Started super_pi run : 2011年 2月 1日 火曜日 20:17:58 JST
 Parameter(%i) to super_pi is missing. Parameter value ? 20
 Start of PI calculation up to 1048576 decimal digits
 End of initialization. Time=       0.452 Sec.
 I= 1 L=       0        Time=       1.358 Sec.
 I= 2 L=       0        Time=       1.543 Sec.
 I= 3 L=       1        Time=       1.539 Sec.
 I= 4 L=       2        Time=       1.557 Sec.
 I= 5 L=       5        Time=       1.546 Sec.
 I= 6 L=      10        Time=       1.538 Sec.
 I= 7 L=      21        Time=       1.537 Sec.
 I= 8 L=      43        Time=       1.536 Sec.
 I= 9 L=      87        Time=       1.536 Sec.
 I=10 L=     174        Time=       1.554 Sec.
 I=11 L=     349        Time=       1.541 Sec.
 I=12 L=     698        Time=       1.538 Sec.
 I=13 L=    1396        Time=       1.545 Sec.
 I=14 L=    2794        Time=       1.539 Sec.
 I=15 L=    5588        Time=       1.533 Sec.
 I=16 L=   11176        Time=       1.522 Sec.
 I=17 L=   22353        Time=       1.497 Sec.
 I=18 L=   44707        Time=       1.437 Sec.
 I=19 L=   89415        Time=       1.339 Sec.
 End of main loop
 End of calculation.    Time=      30.382 Sec.
 End of data output.    Time=       0.197 Sec.
 Total calculation(I/O) time=      30.579(       1.017) Sec.
 ------ Ended super_pi run : 2011年 2月 1日 火曜日 20:18:30 JST

MacBook (Late 2007)

 Version 2.0 of the super_pi for Mac OS/X
 Fortran source program was translated into C program with version 19981204 of
 f2c, then generated C source program was optimized manually.
 gcc-3.2.2 with compile option of "-O3 -ffast-math -finline-limit=1000" was used
 for the compilation.
 ------ Started super_pi run : 2011年 2月 1日 火曜日 19時35分52秒 JST
 Parameter(%i) to super_pi is missing. Parameter value ? 20
 Start of PI calculation up to 1048576 decimal digits
 End of initialization. Time=       0.496 Sec.
 I= 1 L=       0        Time=       1.252 Sec.
 I= 2 L=       0        Time=       1.406 Sec.
 I= 3 L=       1        Time=       1.399 Sec.
 I= 4 L=       2        Time=       1.395 Sec.
 I= 5 L=       5        Time=       1.396 Sec.
 I= 6 L=      10        Time=       1.389 Sec.
 I= 7 L=      21        Time=       1.391 Sec.
 I= 8 L=      43        Time=       1.387 Sec.
 I= 9 L=      87        Time=       1.388 Sec.
 I=10 L=     174        Time=       1.394 Sec.
 I=11 L=     349        Time=       1.383 Sec.
 I=12 L=     698        Time=       1.378 Sec.
 I=13 L=    1396        Time=       1.391 Sec.
 I=14 L=    2794        Time=       1.377 Sec.
 I=15 L=    5588        Time=       1.467 Sec.
 I=16 L=   11176        Time=       1.394 Sec.
 I=17 L=   22353        Time=       1.336 Sec.
 I=18 L=   44707        Time=       1.295 Sec.
 I=19 L=   89415        Time=       1.206 Sec.
 End of main loop
 End of calculation.    Time=      27.604 Sec.
 End of data output.    Time=       0.255 Sec.
 Total calculation(I/O) time=      27.858(       2.727) Sec.
 ------ Ended super_pi run : 2011年 2月 1日 火曜日 19時36分26秒 JST

さくらのVPS

 Version 2.0 of the super_pi for Linux OS
 Fortran source program was translated into C program with version 19981204 of
 f2c, then generated C source program was optimized manually.
 pgcc 3.2-3 with compile option of "-fast -tp px -Mbuiltin -Minline=size:1000 -Mnoframe -Mnobounds -Mcache_align -Mdalign -Mnoreentrant" was used for the
 compilation.
 ------ Started super_pi run : 2011年 2月 1日 火曜日 20:18:26 JST
 Parameter(%i) to super_pi is missing. Parameter value ? 20
 Start of PI calculation up to 1048576 decimal digits
 End of initialization. Time=       0.271 Sec.
 I= 1 L=       0        Time=       0.740 Sec.
 I= 2 L=       0        Time=       0.798 Sec.
 I= 3 L=       1        Time=       0.772 Sec.
 I= 4 L=       2        Time=       0.878 Sec.
 I= 5 L=       5        Time=       0.798 Sec.
 I= 6 L=      10        Time=       0.742 Sec.
 I= 7 L=      21        Time=       0.692 Sec.
 I= 8 L=      43        Time=       0.695 Sec.
 I= 9 L=      87        Time=       0.679 Sec.
 I=10 L=     174        Time=       0.682 Sec.
 I=11 L=     349        Time=       0.700 Sec.
 I=12 L=     698        Time=       0.729 Sec.
 I=13 L=    1396        Time=       0.722 Sec.
 I=14 L=    2794        Time=       0.707 Sec.
 I=15 L=    5588        Time=       0.709 Sec.
 I=16 L=   11176        Time=       0.698 Sec.
 I=17 L=   22353        Time=       0.668 Sec.
 I=18 L=   44707        Time=       0.659 Sec.
 I=19 L=   89415        Time=       0.617 Sec.
 End of main loop
 End of calculation.    Time=      14.499 Sec.
 End of data output.    Time=       0.078 Sec.
 Total calculation(I/O) time=      14.577(       0.613) Sec.
 ------ Ended super_pi run : 2011年 2月 1日 火曜日 20:18:42 JST

ServersMan@VPS(Entryプラン)

 Version 2.0 of the super_pi for Linux OS
 Fortran source program was translated into C program with version 19981204 of
 f2c, then generated C source program was optimized manually.
 pgcc 3.2-3 with compile option of "-fast -tp px -Mbuiltin -Minline=size:1000 -Mnoframe -Mnobounds -Mcache_align -Mdalign -Mnoreentrant" was used for the
 compilation.
 ------ Started super_pi run : 2011年 2月 1日 火曜日 20:18:56 JST
 Parameter(%i) to super_pi is missing. Parameter value ? 20
 Start of PI calculation up to 1048576 decimal digits
 End of initialization. Time=       0.364 Sec.
 I= 1 L=       0        Time=       1.059 Sec.
 I= 2 L=       0        Time=       1.225 Sec.
 I= 3 L=       1        Time=       1.271 Sec.
 I= 4 L=       2        Time=       1.327 Sec.
 I= 5 L=       5        Time=       1.287 Sec.
 I= 6 L=      10        Time=       1.267 Sec.
 I= 7 L=      21        Time=       1.248 Sec.
 I= 8 L=      43        Time=       1.261 Sec.
 I= 9 L=      87        Time=       1.298 Sec.
 I=10 L=     174        Time=       1.320 Sec.
 I=11 L=     349        Time=       1.259 Sec.
 I=12 L=     698        Time=       1.251 Sec.
 I=13 L=    1396        Time=       1.265 Sec.
 I=14 L=    2794        Time=       1.231 Sec.
 I=15 L=    5588        Time=       1.222 Sec.
 I=16 L=   11176        Time=       1.241 Sec.
 I=17 L=   22353        Time=       1.288 Sec.
 I=18 L=   44707        Time=       1.158 Sec.
 I=19 L=   89415        Time=       1.091 Sec.
 End of main loop
 End of calculation.    Time=      24.936 Sec.
 End of data output.    Time=       0.149 Sec.
 Total calculation(I/O) time=      25.085(       1.269) Sec.
 ------ Ended super_pi run : 2011年 2月 1日 火曜日 20:19:28 JST

 

Total calculation time が早い方がいいんだよね、多分。

ってことで、

1位 : NEC Express5800/S70 タイプRB → 13.358秒
2位 : さくらのVPS → 14.577秒
3位 : ServersMan@VPS(Entryプラン) → 25.085秒
4位 : MacBook (Late 2007) → 27.858秒
5位 : Dell PowerEdge 430 → 30.579秒

という結果になりました。
MacBook が意外と遅いね。

 

姫野ベンチの結果

NEC Express5800/S70 タイプRB

mimax = 129 mjmax = 129 mkmax = 257
imax = 128 jmax = 128 kmax =256
 Start rehearsal measurement process.
 Measure the performance in 3 times.

 MFLOPS: 1133.231415 time(s): 0.362958 1.693779e-03

 Now, start the actual measurement process.
 The loop will be excuted in 495 times
 This will take about one minute.
 Wait for a while

 Loop executed for 495 times
 Gosa : 9.987461e-04 
 MFLOPS measured : 1160.141740	cpu : 58.498923
 Score based on Pentium III 600MHz : 14.148070

Dell PowerEdge 430

mimax = 129 mjmax = 129 mkmax = 257
imax = 128 jmax = 128 kmax =256
 Start rehearsal measurement process.
 Measure the performance in 3 times.

 MFLOPS: 887.306594 time(s): 0.463555 1.693779e-03

 Now, start the actual measurement process.
 The loop will be excuted in 388 times
 This will take about one minute.
 Wait for a while

 Loop executed for 388 times
 Gosa : 1.071915e-03 
 MFLOPS measured : 955.270751	cpu : 55.687660
 Score based on Pentium III 600MHz : 11.649643

MacBook (Late 2007)

mimax=  257  mjmax=  129  mkmax=  129
  imax=  256  jmax=  128  kmax=  128
  Start rehearsal measurement process.
  Measure the performance in 3 times.
   MFLOPS:  250.762   time(s):  1.64026  1.733593E-03
 Now, start the actual measurement process.
 The loop will be excuted in  109  times.
 This will take about one minute.
 Wait for a while.
  Loop executed for   109  times
  Gosa :  1.369256E-03
  MFLOPS:  251.219   time(s):  59.4877
  Score based on Pentium III 600MHz :  3.03259

さくらのVPS

mimax = 129 mjmax = 129 mkmax = 257
imax = 128 jmax = 128 kmax =256
 Start rehearsal measurement process.
 Measure the performance in 3 times.

 MFLOPS: 1096.811840 time(s): 0.375010 1.733593e-03

 Now, start the actual measurement process.
 The loop will be excuted in 479 times
 This will take about one minute.
 Wait for a while

 Loop executed for 479 times
 Gosa : 9.995485e-04 
 MFLOPS measured : 1350.076831	cpu : 48.644165
 Score based on Pentium III 600MHz : 16.464352

ServersMan@VPS(Entryプラン)

mimax = 129 mjmax = 129 mkmax = 257
imax = 128 jmax = 128 kmax =256
 Start rehearsal measurement process.
 Measure the performance in 3 times.

 MFLOPS: 584.931168 time(s): 0.703186 1.693779e-03

 Now, start the actual measurement process.
 The loop will be excuted in 255 times
 This will take about one minute.
 Wait for a while

 Loop executed for 255 times
 Gosa : 1.184707e-03 
 MFLOPS measured : 430.082239	cpu : 81.290987
 Score based on Pentium III 600MHz : 5.244905

 

見方が分かんないけど、
数値が大きい方がいいんだよね?

ってことで、

1位 : さくらのVPS
2位 : NEC Express5800/S70 タイプRB
3位 : Dell PowerEdge 430
4位 : ServersMan@VPS(Entryプラン)
5位 : MacBook (Late 2007)

という結果になりました。

 

っということは、
最近買った「NEC Express5800/S70 タイプRB」と「さくらのVPS」が、
ベンチマーク的に良い勝負ってことですね。

すげーな、さくらのVPS。