Memory performance - Stream
Website: University of Virginia
Stream is a synthetic memory bandwidth benchmark developed by John McCalpin at the University of Virginia.
It comprises four individual tests, each of which represents a typical long vector operation, a typical workload in server and HPC applications.
Unfortunately, the Windows version of Stream is hardcoded to support 1, 2, 4, 8, 16 or 32 threads, not the 64 threads native in the Xeon X7560 test system or 24 threads in the Xeon X5650 or Opteron 6174 systems.
In addition, the Xeon X7560 system crashed every time we tried to run the 32-thread version of the Stream memory benchmark on it, so we were forced to use the 16-thread version instead. However, this is the same version we ran on the other two servers.
-
2 x AMD Opteron 6174
-
2 x Intel Xeon X5650
-
4 x Intel Xeon X7560
0
5000
10000
15000
20000
25000
30000
35000
MB/sec (higher is better)
-
2 x AMD Opteron 6174
-
2 x Intel Xeon X5650
-
4 x Intel Xeon X7560
0
5000
10000
15000
20000
25000
30000
35000
MB/sec (higher is better)
-
2 x AMD Opteron 6174
-
2 x Intel Xeon X5650
-
4 x Intel Xeon X7560
0
5000
10000
15000
20000
25000
30000
35000
MB/sec (higher is better)
-
2 x AMD Opteron 6174
-
2 x Intel Xeon X5650
-
4 x Intel Xeon X7560
0
5000
10000
15000
20000
25000
30000
35000
MB/sec (higher is better)
For an explanation of what these numbers mean, head to the
Conclusion.
Want to comment? Please log in.