N. Tajima's FORTRAN benchmark tests (ver.2)
(2007/12/4 by Naoki TAJIMA: http://serv.apphy.u-fukui.ac.jp/~tajima/)
Summary of the results of the benchmark tests
In the following tabel, machines are sorted in the descending order of the
performance index called "fMark".
I define the "fMark" as a weighted geometrical average of the speed,
(reference time)/(CPU time),
namely,
(fMark) = exp ( A / B ),
A = Summation_{over 15 programs} weight * log{(CPU time)/(reference time)},
B = Summation_{over 15 programs) weight.
The advantages to employ the geometirical average are exlpained here.
Reference times are given in a line headed by [ reference time ] in units of
seconds. Weights used in the averaging procedure are given in a line headed
by [ weight ]. The numbers in the table are the ratio
(reference time)/(CPU time), where (CPU time) is the time to run each program.
When the results of many measurements are available (I often tried different
combinations of the compiler options), I have chosen the shortest CPU time.
From year 2007, I use the elapsed time rather than the user CPU time
in cases in which the former seems more appropriate than the latter.
: fMark: matvec: matvecz: leqs4h: leqs1k:jacobi11:jacobi1h: runge: intgl4:mathfnc1:mathfnc2: intosc: intpi3:permute1:permute2:permute3:
[ reference time ] : 296.00: 279.00: 256.00: 279.00: 261.00: 124.00: 204.00: 300.00: 123.00: 226.00: 236.00: 122.00: 23.00: 54.00: 228.00:
[ weight ] : 8: 4: 3: 3: 3: 3: 8: 4: 2: 2: 2: 2: 2: 2: 2:
Core2Duo 2.66GHz : 23.356: 30.737: 48.522: 43.986: 21.678: 19.107: 23.048: 14.216: 13.399: 11.988: 12.556: 41.043: 32.105: 12.778: 28.571: 53.396:
Core2Duo2.6G(2jobs): 17.932: 15.514: 28.097: 32.487: 16.707: 19.121: 23.178: 14.256: 13.447: 12.035: 12.563: 41.115: 32.275: 12.921: 27.000: 7.768:
Core2Duo 2.4GHz : 21.876: 27.561: 43.662: 40.125: 19.255: 16.456: 20.564: 13.609: 12.034: 10.827: 11.692: 36.991: 31.606: 15.333: 36.000: 52.900:
Core2Duo2.4G(2jobs): 16.454: 13.697: 25.738: 28.635: 15.736: 16.477: 20.530: 13.600: 12.092: 10.724: 11.680: 36.476: 31.771: 14.024: 23.478: 7.690:
Core2Duo 2.0GHz : 18.509: 25.170: 29.968: 33.118: 15.380: 14.216: 17.639: 11.327: 10.373: 13.667: 8.567: 36.817: 20.367: 13.939: 32.727: 41.455:
Core2Duo2GHz(2jobs): 14.274: 13.615: 24.282: 23.063: 12.786: 13.551: 16.337: 10.697: 9.753: 12.720: 8.040: 34.706: 19.334: 13.295: 20.377: 9.329:
NEC SX5 : 14.769: 46.614: 62.000: 100.392: 169.091: 1.780: 8.493: 3.380: 55.659: 130.851: 129.143: 2.193: 1.486: 1.041: 1.020: 3.070:
Pentium4 3.0EGHz : 13.184: 19.448: 25.714: 16.431: 10.288: 12.757: 13.177: 8.839: 12.131: 9.186: 6.250: 32.065: 13.466: 9.914: 19.565: 5.122:
Pentium4 3EG(2jobs): 6.872: 12.067: 15.449: 8.782: 5.711: 8.215: 8.010: 3.080: 3.548: 6.721: 4.391: 29.208: 8.866: 4.684: 9.541: 2.279:
CeleronD 3.06GHz : 12.516: 19.078: 27.058: 7.259: 4.956: 13.000: 13.804: 10.249: 12.269: 10.859: 4.883: 33.570: 15.135: 9.952: 19.424: 6.276:
CeleronD 2.66GHz : 10.222: 17.240: 23.763: 8.899: 2.129: 11.362: 11.589: 8.048: 10.739: 8.150: 5.532: 28.304: 11.861: 8.802: 12.091: 3.576:
CeleronD 2.53GHz : 9.679: 15.421: 16.998: 8.870: 2.169: 10.833: 10.969: 7.642: 10.199: 7.776: 5.327: 29.526: 11.477: 8.358: 16.285: 3.654:
Alpha 833MHz : 9.275: 9.472: 22.090: 7.199: 2.958: 9.285: 9.650: 8.793: 12.632: 10.540: 10.706: 10.886: 9.139: 7.077: 7.077: 8.669:
CoreSolo 1.06GHz : 8.145: 12.171: 16.118: 13.502: 9.366: 6.809: 8.115: 5.918: 4.235: 4.345: 4.758: 15.177: 8.925: 4.483: 10.465: 6.147:
Pentium4M 1.66GHz : 7.164: 8.204: 13.757: 8.886: 6.937: 8.344: 8.683: 6.082: 7.335: 6.301: 2.406: 9.161: 6.124: 6.005: 9.660: 2.578:
Celeron 2.4GHz : 6.486: 8.711: 14.968: 0.915: 0.571: 11.683: 10.773: 7.953: 10.313: 9.318: 4.514: 12.812: 8.640: 8.394: 9.375: 1.856:
Hitac S3800/480 : 6.104: 33.110: 51.859: 32.778: 51.381: 0.594: 2.903: 0.762: 26.339: 36.283: 40.071: 1.394: 0.438: 0.398: 0.915: 1.139:
SGI Origin3000 : 5.945: 6.407: 5.886: 16.516: 14.235: 5.699: 7.251: 6.053: 4.464: 7.593: 8.828: 5.187: 1.071: 2.674: 4.286: 3.529:
NEC SX4/2C : 5.836: 20.499: 19.900: 40.764: 78.151: 1.004: 4.980: 1.356: 11.596: 23.077: 25.111: 1.035: 0.473: 0.508: 0.670: 2.300:
Athlon1360MHz,FrBSD: 5.739: 2.769: 4.306: 4.706: 1.724: 9.491: 11.810: 7.969: 12.195: 5.442: 4.989: 19.187: 8.133: 5.750: 10.000: 2.137:
Pentium4 1.6GHz : 5.231: 5.586: 10.353: 6.119: 1.376: 7.700: 8.018: 5.368: 6.832: 4.406: 3.137: 5.122: 5.875: 5.588: 4.752: 1.742:
AMD Duron 1GHz : 3.897: 2.066: 3.809: 3.862: 2.145: 6.527: 7.603: 2.844: 7.726: 4.227: 3.300: 10.669: 5.535: 3.898: 7.040: 3.023:
Fujitsu VPP500 : 3.660: 22.239: 17.372: 13.545: 18.700: 0.535: 2.464: 0.864: 10.316: 9.535: 15.629: 0.495: 0.214: 0.349: 0.480: 1.729:
Fujitsu VPP700E : 3.542: 27.689: 15.389: 18.727: 26.983: 0.479: 2.283: 0.493: 11.270: 6.898: 22.782: 0.496: 0.288: 0.309: 0.378: 1.508:
Celeron 450MHz : 1.787: 1.970: 0.734: 2.210: 2.047: 1.748: 2.449: 0.986: 2.333: 2.228: 1.620: 5.884: 2.657: 1.627: 2.180: 2.844:
Pentium-II 450MHz : 1.692: 1.784: 0.712: 1.394: 1.147: 2.115: 2.267: 1.360: 2.266: 2.175: 1.793: 5.680: 2.298: 1.636: 1.233: 2.623:
Hitac MP5800/320 : 1.635: 1.479: 2.090: 2.788: 0.892: 1.757: 1.611: 1.469: 2.037: 2.295: 2.461: 2.296: 0.812: 0.706: 1.493: 2.600:
Hitac SR2201 : 1.552: 2.674: 4.464: 3.816: 5.158: 0.754: 1.054: 1.590: 1.470: 1.873: 1.538: 0.768: 0.289: 0.304: 0.525: 0.662:
UltraSPARC-2 296MHz: 1.464: 1.118: 1.279: 1.204: 0.651: 2.366: 1.734: 2.454: 2.505: 1.839: 1.451: 1.151: 0.768: 0.907: 1.429: 1.334:
HP Exemplar PA-RISC: 1.368: 0.821: 1.077: 7.254: 1.892: 1.707: 2.708: 1.742: 1.050: 1.321: 1.234: 1.679: 0.349: 0.623: 1.363: 1.152:
HP9000/889 180MHz : 1.172: 1.459: 2.073: 0.476: 0.248: 1.872: 2.705: 1.702: 1.042: 1.321: 1.587: 1.531: 0.349: 0.622: 1.339: 0.695:
Pentium-II 300MHz : 1.124: 1.259: 1.020: 1.090: 0.897: 1.098: 1.395: 0.621: 1.492: 1.427: 1.388: 1.495: 1.514: 1.477: 1.015: 1.718:
UltraSPARC-2 200MHz: 1.085: 1.092: 1.093: 0.871: 0.472: 1.595: 1.235: 1.659: 1.691: 1.252: 0.973: 0.777: 0.518: 0.612: 1.019: 0.941:
Pentium Pro 200MHz : 1.001: 0.998: 1.001: 0.998: 1.001: 1.001: 1.001: 1.000: 1.001: 0.999: 1.001: 0.999: 1.004: 1.018: 1.004: 1.001:
UltraSPARC-1 167MHz: 0.901: 0.937: 0.916: 0.651: 0.373: 1.331: 1.029: 1.382: 1.411: 1.035: 0.818: 0.649: 0.432: 0.511: 0.843: 0.797:
IBM RS6000/590 : 0.858: 2.726: 3.214: 0.735: 0.182: 0.784: 0.836: 0.697: 0.579: 0.827: 0.309: 0.952: 0.207: 0.472: 1.016: 0.754:
HP9000/735 125MHz : 0.709: 0.463: 0.788: 0.258: 0.345: 1.080: 0.942: 1.495: 0.706: 1.057: 1.140: 1.727: 0.276: 0.344: 0.754: 0.643:
Pentium 120MHz : 0.240: 0.232: 0.358: 0.128: 0.102: 0.280: 0.178: 0.273: 0.323: 0.591: 0.673: 0.439: 0.216: 0.078: 0.114: 0.267:
Pentium ODP 83MHz : 0.135: 0.100: 0.104: 0.068: 0.044: 0.165: 0.079: 0.168: 0.227: 0.408: 0.285: 0.323: 0.119: 0.157: 0.134: 0.169:
Motrola 68040 66MHz: 0.058: 0.054: 0.066: 0.093: 0.074: 0.054: 0.043: 0.041: 0.051: 0.055: 0.073: 0.100: 0.035: 0.035: 0.066: 0.189:
The following are the explanations of the machines:
---------- Vector and/or parallel super & main-frame computers ----------------
Hitac S3800/480
test date : 1997/7/6 (year/month/date)
Debut year of the machine : Service began on 1993/2/1 at C.C. of Univ.Tokyo
Vector Super Computer at Computer Center of the University of Tokyo
8 GFlops * 4 CPU, Memory:2GB, RAM disk 16GB (4GB/s), 128 IO channels
water cooling, OS:VOS3/AS
256KB buffer, 99 vector instructions, pipeline pitch=2ns
#pipelines = 8(add,mult)+1(div)+1(mask)+4(load)+4(load,store)
NB) This test used only 1 CPU.
Results compiled with com ,,hap,parm(dcom,comary),lp(ex(ea),ld(any))
SGI Origin3000
test date : 2001/9/7
Debut year of the machine : ?
parallel machine (48 CPU, 48GB shared memory) SPECfp95=54 per CPU at Yukawa Inst of Theor Phys, Kyoto Univ.
NB) This test used only 1 CPU.
Results
Fujitsu VPP700E
test date : 2003/9/10
Debut year of the machine : 1997/2/4
Vector Parallel super computer with 160CPU at Riken (Research
Institute for chemical and physical studies, Wako, Saitama, Japan)
NB) This test used only 1 CPU. 2.4GFlops/CPU
Results
Fujitsu VPP500
test date : 1998/7/3
Debut year of the machine : 1995
Vector Parallel super computer with 28CPU at Riken (Research
Institute for chemical and physical studies, Wako, Saitama, Japan)
NB) This test used only 1 CPU. 1.6GFlops/CPU
Results
NEC SX5
test date : 2003/9/17
Debut year of the machine : ?
Vector Parallel super computer at Yukawa Institute of Kyoto Univeristy
9.052GFlops*4CPU, shared memory 32GB
NB) This test used only 1 CPU.
Results
NEC SX4/2C
test date : 1997/7/3
Debut year of the machine : 1996
Vector Parallel super computer at Yukawa Institute of Kyoto Univeristy
2GFlops*2CPU, Memory 2GB, vector register 144KB*2
NB) This test used only 1 CPU.
Results compiled with "sxf90 -C hopt"
Hitac SR2201
test date : 1998/7/31
Debut year of the machine : Service began on 1996/3/4 at C.C.of Univ.Tokyo
Massively parallel computer at Comp.Center,Univ.Tokyo
(300 MFlops, 256MB)*1024 CPU, 0.3 micro meter CMOS
128 real*8 registers, 1MB SRAM level-2 cache /CPU
inter-CPU communication = 300MB/s
NB) This test used only 1 CPU
Results with optiomizations o(3),o(4),o(s),o(ss)
HP Exemplar PA-RISC
test date : 1998/7/6
Debut year of the machine : 1997 ?
Parallel computer at the Yukawa Institute of Kyoto Univeristy
PA-RISC8000*32, Shared memory 8GB
NB) This test used only 1 CPU.
Compiled with f77 +Oall, Results
Hitac MP5800/320
test date : 1998/7/6
debut year of the machine : Service began on 1996/4/1 at C.C.of Univ.TOkyo
Main-Frame machine of the Computer Center of the University of Tokyo
3 processors, Memory=10GB (incl. RAM disk)
Results compiled with "f77" with default optimizations
---------------------------- Work Stations ------------------------------------
Alpha 21264(EV68) 833MHz
test date : 2001/5/18
alpha1.yukawa.kyoto-u.ac.jp at the Yukawa Institute of Kyoto Univeristy
833MHz Alpha 21264(EV68), Compaq Tru64 UNIX V5.1 (Rev. 732)
compiled with [ f77 -O -tune ev6 ]
Results
IBM RS6000/590
test date : 1998/7/2
Debut year of the machine : ?
ps[12].yukawa.kyoto-u.ac.jp at the Yukawa Institute of Kyoto Univeristy
IBM Risc Station 6000/590, 1 CPU with unknown clock speed, Memory=280MB
Results compiled with "xlf -O3 -qarch=pwr2":No.1,No.2
Results compiled with "xlf -O3":No.1
UltraSPARC-2 296MHz
test date : 1998/7/6
Debut year of the machine : ?
sun4.yukawa.kyoto-u.ac.jp of the Yukawa Institute of Kyoto Univeristy
(SUN UltraSparc-2 296MHz)*4 CPU, Memory=1024MB
Results with optimizations -fast -O4,-fast -O4 -cg92
UltraSPARC-2 200MHz
test date : 1998/7/2
Debut year of the machine : ?
sun[123].yukawa.kyoto-u.ac.jp of the Yukawa Institute of Kyoto Univeristy
(SUN UltraSparc-2 200MHz) x 2 CPU, Memory=256-320MB
Results with optimizations "-fast -O4" : No.1,No.2,No.3
Results with optimizations "-fast -O4 -cg92" : No.1,No.2,No.3
UltraSPARC-1 167 MHz
test date : 1998/7/2
Debut year of the machine :
sun[56].yukawa.kyoto-u.ac.jp of the Yukawa Institute of Kyoto Univeristy
(SUN UltraSparc-1 167MHz)*2 CPU, Memory=128-192 MB
Results compiled with "-fast -O4" : No.1,No.2
HP9000/k260 180 MHz
test date : 1998/7/8
Debut year of the machine : ?
hp1.yukawa.kyoto-u.ac.jp of the Yukawa Institute of Kyoto Univeristy
Hewlett Packard 9000/K260, 4 CPU working at 180 MHz, Memory=512MB
Results compiled with f77 +Oall
HP9000/889 180 MHz
test date : 1998/7/8
Debut year of the machine :
hp2.yukawa.kyoto-u.ac.jp of the Yukawa Institute of Kyoto Univeristy
Hewlett Packard 9000/889, 2 CPU working at 180 MHz, Memory=256MB
Results compiled with f77 +Oall
HP9000/735 125MHz
test date : 1998/7/2
Debut year of the machine : ?
hp01.yukawa.kyoto-u.ac.jp of the Yukawa Institute of Kyoto Univeristy
Hewlett Packard 9000/735, 1 CPU working at 125 MHz, Memory=144MB
Results compiled with f77 +Oall
HP9000/735 99MHz
test date : 1998/7/3
Debut year of the machine : ?
hp0[23].yukawa.kyoto-u.ac.jp of the Yukawa Institute of Kyoto Univeristy
Hewlett Packard 9000/735, 1 CPU working at 99 MHz, Memory=128MB
Results compiled with f77 +Oall
Motrola 68040 66MHz
test date : 1998/7/2
Debut year of the machine : <1993
nt1.c.u-tokyo.ac.jp
Next Station Turbo, CPU=Motrola 68040 (CPU clock = 66MHz, bus clock=33MHz)
OS = NEXTSTEP
compiler = absoft fortran77, options = "f77 -N53 -f -s -O"
Results
-------------- PC with Intel or AMD's CPU operationg with Linux or Free BSD ----------------------
Core2Duo(dual core) E6750 2.66GHz, 4MB L2 Cache, 1333MHz FSB, M/B Intel DG33BU
Note: Each moment, only one job was running using one core. Elapsed time was used to calculate fMark.
test date : 2007/11/21
machine : Fedora7 (Linux 2.6.23.1-21.fc7), gcc,gfortran 4.1.2
compiled with [ gfortran -malign-double -funroll-loops -fexpensive-optimizations -O3 ]
Results
Core2Duo(dual core) E6750 2.66GHz, 4MB L2 Cache, 1333MHz FSB, M/B Intel DG33BU, 2jobs
Note: Each moment, two same jobs were running using two cores. Elapsed time was used to calculate fMark.
test date : 2007/11/21
machine : Fedora7 (Linux 2.6.23.1-21.fc7), gcc,gfortran 4.1.2
compiled with [ gfortran -malign-double -funroll-loops -fexpensive-optimizations -O3 ]
Results of job No.1, Results of job No.2
Core2Duo(dual core) E6600 2.4GHz, 4MB L2 Cache, 1066MHz FSB, M/B ASUSTek P5LD2-VM R2(i945G)
Note: Each moment, only one job was running using one core. Elapsed time was used to calculate fMark.
test date : 2006/12/28, 2007/8/22(revised)
machine : SUSE Linux 10.1 (Linux 2.6.16.13-4-smp), gcc,gfortran 4.1.0
compiled with [ gfortran -malign-double -funroll-loops -fexpensive-optimizations -O3 ]
Results
Core2Duo(dual core) E6600 2.4GHz, 4MB L2 Cache, 1066MHz FSB, M/B ASUSTek P5LD2-VM R2(i945G), 2jobs
Note: Each moment, two same jobs were running using two cores. Elapsed time was used to calculate fMark.
test date : 2006/12/28, 2007/8/22(revised)
machine : SUSE Linux 10.1 (Linux 2.6.16.13-4-smp), gcc,gfortran 4.1.0
compiled with [ gfortran -malign-double -funroll-loops -fexpensive-optimizations -O3 ]
Results of job No.1, Results of job No.2
Core2Duo(dual core) T7300 2.0GHz, 4MB L2 Cache, 800MHz FSB
Note: Each moment, only one job was running using one core. Elapsed time was used to calculate fMark.
test date : 2007/8/21
machine : DELL Inspiron Notebook 1420, cygwin on Windows Vista
compiled with [ g77 -O3 ]
Results
Core2Duo(dual core) T7300 2.0GHz, 4MB L2 Cache, 800MHz FSB, 2jobs
Note: Each moment, two same jobs were running using two cores. Elapsed time was used to calculate fMark.
test date : 2007/8/21
machine : DELL Inspiron Notebook 1420, cygwin on Windows Vista
compiled with [ g77 -O3 ]
Results of job No.1, Results of job No.2
CoreSolo U1300 1.06GHz, 2MB L2chache, 533MHz FSB
test date : 2007/8/22
Note : Elapsed time was used to calculate fMark.
machine : Panasonic Let's note CF-W5, SuSE linux 10.1
compiled with [ gfortran -malign-double -O3 -funroll-loops -fexpensive-optimizations ]
Results, Results with cygwin on Windows XP(fMark=8.011)
Pentium4 3.0EGHz, 1MB L2 Cache, 800MHz FSB, PC3200 dual channel DDR SDRAM
Note: "E" of 3.0E stands for 90nm process. Pentium4 without "E" is made by 130nm process.
test date : 2005/9/15, 2007/8/22(revised)
machine : EPSON LX-7500 TurboLinux10 (linux 2.6.0), g77-3.3.1
compiled with [ g77 -malign-double -funroll-loops -fexpensive-optimizations -O3 ]
Results
Pentium4 3.0EGHz, 1MB L2 Cache, 800MHz FSB, PC3200 dual channel DDR SDRAM, 2jobs
Note: Each moment, two same jobs were running using a single core. Elapsed time was used to calculate fMark.
This CPU features the HyperThreading, which makes a single CPU core look two cores logically.
test date : 2007/8/22
machine : EPSON LX-7500 TurboLinux10 (linux 2.6.0), g77-3.3.1
compiled with [ g77 -malign-double -funroll-loops -fexpensive-optimizations -O3 ]
Results of job No.1, Results of job No.2
Celeron D345, 3.06GHz, 256kB L2 Cache, 533MHz FSB
test date : 2007/2/10
machine : EPSON Endeavor MT7700, Windows XP, cygwin g77
compiled with [ g77 -malign-double -O3 -funroll-loops -fexpensive-optimizations ]
Results
CeleronD-330, 2.66GHz, 256kB L2 Cache, 533MHz FSB, PC2700 dual channel DDR SDRAM
Note: Unlike PentiumD, "D" of CeleronD does not stand for "Dual core" but for "Desktop"
test date : 2005/9/15
machine : EPSON LX-7500 TurboLinux10 (linux 2.6.0) g77-3.3.1
compiled with [ g77 -malign-double -funroll-loops -fexpensive-optimizations -O3 ]
Results
CeleronD-325, 2.53GHz, 256kB L2 Cache, 533MHz FSB (DDR-SDRAM 333MHz, PC2700 DIMM)
test date : 2005/6/24
machine : EPSON LX-7500 TurboLinux10 (linux 2.6.0) pre-installed
compiled with [ g77 -O3 ]
Results
Celeron 2.4GHz, FSB 400MHz(Memory Bus 100MHz), 128kB L2 Cache
test date : 2004/4/12
machine : EPSON DIRECT AT-950, Cygwin on Windows-XP
compiled with [ g77 -O3 ]
Results (fMark=6.486)
test date : 2007/7/6
machine : EPSON DIRECT AT-950, linux 2.6.0, g77-3.3.1
compiled with [ g77 -malign-double -O3 -funroll-loops -fexpensive-optimizations ]
Results (fMark = 6.232)
Mobile Pentium4-M 1.66GHz, 1MB L2 Cache
test date : 2003/4/5
machine : DELL inspiron 8200 (Note Book Type), Cygwin on Windows-XP
compiled with [ g77 -O3 ]
Results
Athlon1360MHz,FrBSD
test date : 2001/7/12
machine : Athlon 1360MHz (FSB 136MHz), FreeBSD, g77
Tested by prof. Y. R. Shimizu of Kyushu University
compile with [ g77 -w -funroll-loops -malign-double -fexpensive-optimizations -funroll-loops -O4 ]
Results
Pentium4 1.6GHz, 256kB L2 Cache
test date : 2005/8/29
machine : DELL dimension 4300S, Linux 2.6.0, g77-3.3.1
compiled with [ g77 -O3 ]
Results
AMD Duron 1GHz (1.6GHz CPU is downclocked for a thermal problem), FSB 200MHz
test date : 2002/10/17
machine : AMD Duron 1GHz, (FSB 200MHz), Linux (Vine 2.5)
compile with [ g77 -O3] <- faster than -funroll-loops -malign-double -fexpensive-optimizations -funroll-loops -O4
Results
Celeron 450MHz
test date : 2001/6/4
Debut year of the machine : 2000 ?
bec.apphy.fukui-u, Dual Celeron450MHz,linux
compile with [ g77 -w -funroll-loops -malign-double -ffast-math -fexpensive-optimizations -O3 ]
Results
Pentium-II 450MHz
test date : 2001/6/9
Debut year of the machine : 1998
nok.apphy.fukui-u.ac.jp, PenII 450MHz, linux
compile with [ g77 fastest results ]
An example of the results
Pentium-II 300MHz
test date : 1998/7/2
Debut year of the machine : 1997
nt4.c.u-tokyo.ac.jp of Komaba nuclear theory group
Proside Pro-2300AL-S, Memory=192MB (64MB DIMM x 3)
single Intel Pentium II CPU at 300MHz, 512KB cache at 150MHz
Results compiled with g77 -w -funroll-loops -malign-double -ffast-math -fexpensive-optimizations -O3
Results compiled with g77 -O3
Results compiled with g77 w/o optimizations
Pentium Pro 200MHz
test date : 1998/7/2
Debut year of the machine : 1996
tjm.apphy.fukui-u.ac.jp(=nt3.c.u-tokyo.ac.jp)
single Intel Pentium Pro CPU at 200MHz, 256KB cache at 200MHz
tjm : EPSON VA620V, Memory=128MB (=32MB SIMM x 4)
phys22 : Proside...,Memory=64MB
compliler = g77, Absoft fortran77 for Intel
Results No.1,No.2 compiled with g77 -w -funroll-loops -malign-double -ffast-math -fexpensive-optimizations -O3
Results compiled with g77 -O3
Results compiled with g77 w/o optimizations
Results compiled with absoft f77 -s -O
Results compiled with absoft f77 -s w/o optimizations
Pentium 120MHz
test date : 1998/7/2
Debut year of the machine : 1995
nt2.c.u-tokyo.ac.jp
EPSON VT512R
Pentium 120MHz, bus clock=60MHz,
compiler = absoft f77 for intel, compiled with "f77 -s -O"
Results
Pentium ODP 83MHz
test date : 1998/7/2
Debut year of the machine : 1995
phys20.c.u-tokyo.ac.jp
Fujitsu FM-V 425D with Pentium OverDriveProcessor at 83MHz, bus clock=33MHz
compiled with "g77 -w -funroll-loops -malign-double -ffast-math -fexpensive-optimizations -O3"
Results
-------------------------------------------------------------------------------
I roughly counted the number of floating-point operations for three programs:
(program) : #add,sub #mult #div total matvec : 1*n 1*n 0*n 2*n where n=6.3E+9 runge : 32*n 44*n 4*n 80*n where n=2.0E+8 intgl4 : 6*n 3*n 1*n 10*n where n=1.6E+9I estimated the FLOPS values for a few systems as
-------------------------------------------------------------------------------
(machine) (CPU clock) program : FLOPS value :Officially claimed speed
-------------------------------------------------------------------------------
Hitac S3800/480 : 500MHz : matvec : 1.4 GFLOPS : 8 GFLOPS
: : intgl4 : 1.4 GFLOPS :
: : runge : 60 MFLOPS :
-------------------------------------------------------------------------------
UltraSPARC-2 : 296MHz : matvec : 48 MFLOPS : SEPCint95 = 12.1
: : intgl4 : 134 MFLOPS : SPECfp95 = 18.3
: : runge : 192 MFLOPS :
-------------------------------------------------------------------------------
Pentium II : 300MHz : matvec : 54 MFLOPS : SEPCint95 = 11.6
: : intgl4 : 80 MFLOPS : SPECfp95 = 7.2
: : runge : 49 MFLOPS :
-------------------------------------------------------------------------------