[MAINIMAGE]
 
  Welcome
  Community
  Documentation
  Download
  Machines
 
Projects 
For Developers 
Dev Plans 
Bug Tracking 
Feature Requests 
Mailing Lists 
Viz Tools 
  Showcase
  Links
 



Quick Subscription:
News Mailing List


more Mailing Lists
  Search CactusCode:


BenchADM (40-cubed)

The BenchADM benchmark application is the computational kernal representative of many applications in numerical relativity. It is characterised by being compute intensive, with around 600 Floating point operations per grid point and with six grid functions synchronised each timestep. The benchmark is run here with a local workload of 40-cubed on each processor).

Machine ID Architecture Fortran Compiler Single Processor
(secs) [1]		 MFlops [2][3]
(%Peak)		 Scaling
(16 procs) [4]
AMD  
 
Amarok Athlon Box Portland
3.2-3 		17.8 236 94% (at 2)
 
Amarok Athlon Box Intel 5.0.1
(010730D0) 		16.7 251 92% (at 2)
 
Amarok Athlon Box Absoft
2.1 		34.31 122 98% (at 2)
 
Apple  
 
G4 Mac Absoft
7.0 		74 57 N/A
Compaq  
 
TCSini Alpha Native 8.6 488
(32%) 		 
Sierra Alpha Native 8.8 477 83% (at 12)
gif
holodec Alpha Digital F90
V5.2-705 		9.7 433 N/A
Cray  
 
Turing T3E Native 44.7 94
(8%) 		96%
GRAPH
Fujitsu  
 
Fujitsu VPP700/52 Native 56 75  
 
Hitachi  
 
Hitachi SR-8000 SR-8000 Native 11 381
(25%) 		48%
gif
IA32 Linux  
 
Platinum Pentium III Portland 24.6 [a] 181
(18%) 		????
gif
Platinum Pentium III Portland 3.2-3 17.7 237
(24%) 		 
Platinum Pentium III Absoft 2.3 23.9 176
(18%) 		 
Platinum Pentium III Intel 5.0.1
(010730D0) 		17.5 240
(24%) 		 
albert Pentium III
Fast-Ethernet		 Intel 5.0.1 (010730D0)
LAM 5.4.5		 14.1   46%
gif
albert Pentium III
Fast-Ethernet		 GNU gcc 2.96
LAM 5.4.5		 37.9   75%
gif
Zap Pentium III Portland 3.2-3 20.6 204
(24%) 		N/A
Zap Pentium III Absoft 2.1 55.3 78
(10%) 		N/A
Zap Pentium III Absoft 2.3 27.9 150
(18%) 		N/A
Zap Pentium III Intel 5.0.1
(010730D0) 		20.7 203
(24%) 		N/A
Xeon Pentium IV Xeon Intel 5.0.1
(010730D0) 		10.2 437
(26%) 		93% (at 2)
gif
IA64 Linux  
 
Titan IA64 Intel 6.0 Beta (-O1)
(20010905) (???) 		9.4 446 (14%) 82%
gif
Titan IA64 Intel 6.0 Beta (-O2)
(20010905) 		9.8 428 (13%) 82%
gif
Titan IA64 Intel 6.0 Beta (-O3)
(20010905) 		13.2 318 (10%) 82%
gif
Titan IA64
VMI/GigE		 Intel 6.0 (-O3)
(20020320) 		13.7 306.3 (9.5%) 80%
data
Titan IA64
VMI / Myrinet		 Intel 6.0 (-O3)
(20020320) 		13.8 304.1 (9.5%) 91%
data
IBM  
 
Blue Horizon SP2 Native 14.2 296
(20%) 		88% (at 8)
gif
Seaborg SP3 Native 13.8 304
(20%) 		85%
(at 8)
gif
Psi Power4 xlc (-O2)
 Interact.: 5.7
RZG: 4.8		 735 14%
RZG: 874 -		 75%
data
NEC  
 
hwwsx5 NEC SX-5Be/32M2 Native 6.1 322
15% 		95% (at 4)
gif
SGI  
 
AEI Origin 2000 Native
MIPSpro 7.3.1.2m 		38 110.5
(29%) 		85%
gif
Lux Octane Native
MIPSpro 7.2.1 		28.3 148
(25%) 		N/A
Harpo SGI ?? Native
MIPSPro 7.30 		30 139
(28%) 		71%(at 4)
gif
Green Origin 3000 Native
???? 		19 220
(28%) 		87%
gif
Sun  
 
Inca Sun E6500 Native 32 131
20% 		74% (at 16)
gif

Notes

  1. Measured using Cactus timer (Total time) with gettimeofday clock.
  2. MFlops: This is calculated using the number of floating point operations calculated on the Origin 2000 using perfex including standard optimisation but switching off multadds (compiling with -TARG:madd=OFF). The number of floating point operations for the different cases are listed on the pages describing the individual benchmarks (to come).
  3. MFlops here means 1,000,000 Flops.
  4. Scaling is defined by user time on one processor/user time on 16 processors

Run Notes

[a] Before additional optimisation flags used
      

Cactus Webmaster Last Modified: Thursday, 25-Sep-2003 10:07:02 CDT