Home Go Up

(2) NVIDIA C2050's in a HP Z800 Workstation
(8) NVIDIA M2050's in a server

Machine Description

GPU Processors NVIDIA Fermi Tesla 2050
448 Cores, 1.15 Ghz clock
4.0 Run time library
Host Workstation HP Z800 Workstation
(1) Intel Xeon 5530 2.4 Ghz i7 quadcore processor
(48) Gbytes of memory
(32) Gbytes used by FMS
(6) SAS 15K disks
Host Server (2) Intel Xeon i7 6-core processor
(48) Gbytes of memory
(42) Gbytes used by FMS
(3) SATA 7.2K disks
Workstation Operating System Suse Linux 2.6.32.12 x86_64
Server Operating System GNU Linux 2.6.18-164.e15 x86_64

Problem Description

Code Used Example 11
Matrix Sparsity Full
Matrix Symmetry Nonsymmetric
Data Type 16-byte Complex
Number of Equations 20,000 to 250,000
Number of Vectors 1
Reduced Operation Algorithms None (IALGOR=0)

Results for Full Complex Nonsymmetric Matrices

:
Number of
Equations		 Storage Date
Started		 Num.
GPU'		 Times (Day::Hr:Min:Sec) Gigaflops
Loc. Gbytes Wall I/O Wait Overall
20,000 Mem. 8 12/27/10 0 9:30 0:00 37
9 04/07/11 1 1:21 0:00 264
10 04/06/11 2 0:46 0:00 449
10 04/14/11 8 0:21 0:00 1,000
Disk 8 12/22/10 0 9:34 0:03 37
9 04/07/11 1 1:27 0:06 245
8 04/06/11 2 0:52 0:06 413
8 04/06/11 8 0:23 - 942
40,000 Mem. 27 12/27/10 0 1:14:34 0:00 38
31 04/07/11 1 9:39 0:00 295
31 04/06/11 2 5:11 0:00 550
37 04/15/11 8 1:55 0:00 1,486
Disk 27 12/22/10 0 1:14:42 0:06 38
31 04/07/11 1 9:58 0:16 285
33 04/06/11 2 5:35 0:23 510
33 04/06/11 8 1:56 - 1,474
60,000 Disk 58 12/22/10 0 4:10:46 0:09 38
66 04/07/11 1 31:52 0:28 301
66 04/06/11 2 17:36 0:28 546
71 04/15/11 8 5:45 - 1,671
80,000 Disk 101 12/22/10 0 9:53:09 0:11 38
109 04/07/11 1 1:13:55 0:35 308
117 04/06/11 2 40:28 1:39 562
117 04/15/11 8 11:59 - 1,899
100,000 Disk 155 12/23/10 0 19:17:15 0:14 38
167 04/07/11 1 2:22:50 1:04 311
172 04/07/11 2 1:16:55 1:57 578
207 04/15/11 8 21:39 - 2,054
150,000 Disk 354 12/24/10 0 2::17:01:06 0:21 38
362 04/07/11 1 7:59:31 2:36 313
381 04/06/11 2 4:14:41 4:33 589
386 04/15/11 8 1:08:39 - 2,185
200,000 Disk 628 04/08/11 1 18:53:12 3:43 314
644 04/06/11 2 9:52:00 5:53 601
728 04/15/11 8 2:41:06 - 2,207
250,000 Disk 997 04/05/11 2 19:21:04 9:15 598
1,067 04/16/11 8 5:07:37 - 2,258

NOTES:

  1. 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 or transfers to complete.
  2. I/O Wait Time is the total time spent waiting for I/O to complete that is not overlapped by asynchronous I/O.
  3. Overall Gigaflops is the total number of floating point operations performed, divided by the Wall Time.
  4. Typical transfer rate across the PCIe X16 bus to the device is 7 Gbytes/sec. total.
  5. Typical transfer rate to/from the SAS disks in the workstation was 500 Mbytes/sec.
  6. The SATA disks in the server were only used to run a large problem. The overall performance of these disks was 225 Mbytes/sec. For that reason I/O wait time is not reported for the server.


HomeGo Up
Copyright © Multipath Corporation