Serverworks chip sets
Steffen Persvold
sp at scali.no
Mon Dec 11 10:10:41 PST 2000
Here are a more meaningful comparision. The comparison is between a Tyan
S2510 based on the LE chipset and a Supermicro 370DER based on the HE-SL
chipset.
Both machines are equipped with 2x PIII-800EB(133) CPUs and 512MByte
PC-133 SDRAM
Using one CPU:
370DER:
----------------------------------------------
Double precision appears to have 16 digits of accuracy
Assuming 8 bytes per DOUBLE PRECISION word
----------------------------------------------
Array size = 2000000
Offset = 0
The total memory requirement is 45 MB
You are running each test 20 times
--
The *best* time for each test is used
*EXCLUDING* the first and last iterations
----------------------------------------------------
Your clock granularity appears to be less than one microsecond
Your clock granularity/precision appears to be 1 microseconds
----------------------------------------------------
Function Rate (MB/s) Avg time Min time Max time
Copy: 374.2689 0.0855 0.0855 0.0856
Scale: 401.9594 0.0796 0.0796 0.0797
Add: 475.5536 0.1010 0.1009 0.1010
Triad: 475.1911 0.1011 0.1010 0.1012
----------------------------------------------------
Solution Validates!
----------------------------------------------------
S2510:
----------------------------------------------
Double precision appears to have 16 digits of accuracy
Assuming 8 bytes per DOUBLE PRECISION word
----------------------------------------------
Array size = 2000000
Offset = 0
The total memory requirement is 45 MB
You are running each test 20 times
--
The *best* time for each test is used
*EXCLUDING* the first and last iterations
----------------------------------------------------
Your clock granularity appears to be less than one microsecond
Your clock granularity/precision appears to be 1 microseconds
----------------------------------------------------
Function Rate (MB/s) Avg time Min time Max time
Copy: 376.9848 0.0849 0.0849 0.0851
Scale: 375.0191 0.0854 0.0853 0.0854
Add: 483.7248 0.0993 0.0992 0.0994
Triad: 484.5352 0.0991 0.0991 0.0991
----------------------------------------------------
Solution Validates!
----------------------------------------------------
Using both CPUs (OMP makes it a bit easier and adds the numbers for me):
370DER:
----------------------------------------------
Double precision appears to have 16 digits of accuracy
Assuming 8 bytes per DOUBLE PRECISION word
----------------------------------------------
Array size = 2000000
Offset = 0
The total memory requirement is 45 MB
You are running each test 20 times
--
The *best* time for each test is used
*EXCLUDING* the first and last iterations
----------------------------------------------------
Your clock granularity appears to be less than one microsecond
Your clock granularity/precision appears to be 1 microseconds
----------------------------------------------------
Function Rate (MB/s) Avg time Min time Max time
Copy: 468.8234 0.0686 0.0683 0.0693
Scale: 479.3794 0.0669 0.0668 0.0671
Add: 541.0826 0.0888 0.0887 0.0888
Triad: 542.6429 0.0885 0.0885 0.0886
----------------------------------------------------
Solution Validates!
----------------------------------------------------
S2510:
----------------------------------------------
Double precision appears to have 16 digits of accuracy
Assuming 8 bytes per DOUBLE PRECISION word
----------------------------------------------
Array size = 2000000
Offset = 0
The total memory requirement is 45 MB
You are running each test 20 times
--
The *best* time for each test is used
*EXCLUDING* the first and last iterations
----------------------------------------------------
Your clock granularity appears to be less than one microsecond
Your clock granularity/precision appears to be 1 microseconds
----------------------------------------------------
Function Rate (MB/s) Avg time Min time Max time
Copy: 353.0918 0.0916 0.0906 0.0926
Scale: 355.3543 0.0918 0.0901 0.0931
Add: 397.2821 0.1213 0.1208 0.1222
Triad: 397.7756 0.1218 0.1207 0.1225
----------------------------------------------------
Solution Validates!
----------------------------------------------------
As you can see the two-way interleaving helps a bit on a Dual-CPU run
(~35% more bandwidth)
Best regards,
--
Steffen Persvold Systems Engineer
Email : mailto:sp at scali.no Scali AS (http://www.scali.com)
Tlf : (+47) 22 62 89 50 Olaf Helsets vei 6
Fax : (+47) 22 62 89 51 N-0621 Oslo, Norway
More information about the Beowulf
mailing list