Home Go Up

SGI Altix UV1000

Machine Description

Model Number SGI Altix UV1000
Processor Type Intel i7 6-Core @ 2.66 GHz
Number of Processors 64
Number of Cores 384
Memory Installed 2 Tbytes
Memory Used by FMS 1 Tbytes
Operating System Linux 2.6.32.24-0.2.1.2230.2.PTF-default

Problem Description

Code Used Example 11
Matrix Sparsity Full
Matrix Symmetry Symmetric and Nonsymmetric
Data Type 16-byte Complex
Number of Equations 20,000 to 500,000
Number of Vectors 1
Reduced Operation Algorithms None (IALGOR=0)

Results for Complex Nonsymmetric Matrices

Date
Started		 Number of
Equations		 Storage Times (Day::Hr:Min:Sec) Megaflops Parallel
Speedup
Loc. Gbytes CPU Wall I/O Wait CPU Overall
03/21/11 20,000 Mem. 11.9 2:23:17 1:10 0:00 2,481 306,658 123.6
03/21/11 20,000 Disk 11.9 2:23:17 1:29 0:08 2,540 239,583 94.3
03/21/11 40,000 Mem. 37.0 8:36:49 2:47 0:00 5,504 1,021,101 185.5
03/21/11 40,000 Disk 37.0 8:38:13 3:25 0:13 5,489 830,560 151.3
03/21/11 60,000 Disk 74.1 1::03:32:03 7:54 0:18 5,811 1,214,834 209.0
03/21/11 80,000 Disk 116.3 2::10:12:47 13:32 0:17 6,515 1,681,083 258.0
03/19/11 100,000 Disk 185.2 4::10:10:09 23:51 0:26 6,977 1,863,944 267.1
03/21/11 150,000 Disk 411.4 13::07:33:00 1:09:35 0:58 7,823 2,155,681 275.5
03/21/11 200,000 Disk 726.8 31::00:42:59 2:31:46 1:43 7,957 2,342,937 294.4
03/19/11 250,000 Disk 1,131.2 58::19:33:27 4:36:53 2:47 8,199 2,508,073 305.9
03/19/11 500,000 Disk 4,385.8 NA(7) 1::09:27:39 48:01 NA(7) 2,767,216 NA(7)

NOTES:

  1. CPU Time is the total amount of User and System time used by all processors.
  2. Wall Time is the elapsed time measured on a dedicated machine. This includes the time spent processing and any time spent waiting for I/O to complete.
  3. I/O Wait Time is the total time spent waiting for I/O to complete that is not overlapped by asynchronous I/O.
  4. Megaflops per Processor is the theoretical number of floating point operations performed, divided by the CPU time.
  5. Overall Megaflops is the total number of floating point operations performed, divided by the Wall Time.
  6. Parallel Speedup is overall increase in performance due to parallel processing.
  7. NA(7). Due to the size of this problem, the CPU time counter wrapped. Therefore CPU time, CPU Mflops and parallel scaling are not available. The overall performance, which is based on wall time, is accurate.


HomeGo Up
Copyright © Multipath Corporation