STREAMS on new AlphaServer 8400 5/300

From: John, dtn 381-0378 11-Apr-1995 1521 (henning@i4get.ENET.dec.com)
Date: Tue Apr 11 1995 - 14:19:48 CDT


Hello Professor McCalpin,

As promised a week or so ago, here are the Streams benchmark results for
the new Digital AlphaServer 8400 5/300 system announced today.

As with previous submissions, will you please place the full set of
results (1, 2, 4, 6, and 8 cpus) in both the base table.print and in
parallel.speedup?

Since "Digital AlphaServer 8400 5/300" is rather longer than your
current column width, I've suggested an abbreviation below. The new
naming convention for Alphas includes a model series indicator (8400 is
the refrigerator-sized SMP server); a chip generation number (5); and
the Mhz (300). Over time models may ship with differing Mhz ratings or
chip generations, so the trailing n/nnn is important.

If there are any questions that we can answer about the runs, please
feel free to contact me.

Best regards,
    /John Henning
     CSD Performance Group
     Digital Equipment Corporation
     henning@zko.dec.com
     603 881-0378

                            Bytes Bandwidth (MB/s)
Machine NCPUs /word Copy Scale Sum Triad
--------------- ----- ----- -------- -------- -------- --------
Digital 8400 5/300 1 8 186.3 174.4 198.3 195.2
Digital 8400 5/300 2 8 338.0 319.9 335.7 364.3
Digital 8400 5/300 4 8 611.5 574.0 611.9 656.7
Digital 8400 5/300 6 8 797.6 739.7 791.6 866.2
Digital 8400 5/300 8 8 899.8 834.2 892.8 978.8

                                      Copy Triad
                            Bytes Parallel Parallel
Machine NCPUs /word Speed-up Speed-up
--------------- ----- ----- -------- --------
Digital 8400 5/300 8 8 4.83 5.02
Digital 8400 5/300 6 8 4.28 4.44
Digital 8400 5/300 4 8 3.28 3.36
Digital 8400 5/300 2 8 1.81 1.87
Digital 8400 5/300 1 8 1.00 1.00

1 CPU
--------------------------------------
 Double precision appears to have 16 digits of accuracy
 Assuming 8 bytes per DOUBLEPRECISION word
--------------------------------------
 CPU time ; time = 26.7089843750000 hundredths of a second
 Wall Time ; time = 26.7423987388611 hundredths of a second
 Increase the size of the arrays if this is <100
  and your clock precision is =<1/100 second
 ---------------------------------------------------
Function Max MB/s RMS time Min time Max time
Assignment: 186.2891 0.0895 0.0859 0.0966
Scaling : 174.3984 0.0945 0.0917 0.1132
Summing : 198.3078 0.1268 0.1210 0.1279
SAXPYing : 195.1599 0.1253 0.1230 0.1366

2 CPUs - Add 2 tables together
--------------------------------------
 Double precision appears to have 16 digits of accuracy
 Assuming 8 bytes per DOUBLEPRECISION word
--------------------------------------
 CPU time ; time = 33.6669921875000 hundredths of a second
 Wall Time ; time = 33.6719989776611 hundredths of a second
 Increase the size of the arrays if this is <100
  and your clock precision is =<1/100 second
 ---------------------------------------------------
Function Max MB/s RMS time Min time Max time
Assignment: 167.2801 0.0972 0.0956 0.1054
Scaling : 159.1596 0.1015 0.1005 0.1021
Summing : 168.4258 0.1431 0.1425 0.1435
SAXPYing : 183.5086 0.1342 0.1308 0.1357
--------------------------------------
 Double precision appears to have 16 digits of accuracy
 Assuming 8 bytes per DOUBLEPRECISION word
--------------------------------------
 CPU time ; time = 33.1787109375000 hundredths of a second
 Wall Time ; time = 33.1840038299561 hundredths of a second
 Increase the size of the arrays if this is <100
  and your clock precision is =<1/100 second
 ---------------------------------------------------
Function Max MB/s RMS time Min time Max time
Assignment: 170.7650 0.0953 0.0937 0.1044
Scaling : 160.7202 0.0997 0.0996 0.1001
Summing : 167.2801 0.1441 0.1435 0.1450
SAXPYing : 180.8099 0.1339 0.1327 0.1347

4 CPUs - Add 4 tables together
--------------------------------------
 Double precision appears to have 16 digits of accuracy
 Assuming 8 bytes per DOUBLEPRECISION word
--------------------------------------
 CPU time ; time = 38.6718750000000 hundredths of a second
 Wall Time ; time = 38.8447999954224 hundredths of a second
 Increase the size of the arrays if this is <100
  and your clock precision is =<1/100 second
 ---------------------------------------------------
Function Max MB/s RMS time Min time Max time
Assignment: 153.2098 0.1067 0.1044 0.1171
Scaling : 142.5517 0.1130 0.1122 0.1142
Summing : 154.6552 0.1561 0.1552 0.1571
SAXPYing : 164.3835 0.1483 0.1460 0.1493
--------------------------------------
 Double precision appears to have 16 digits of accuracy
 Assuming 8 bytes per DOUBLEPRECISION word
--------------------------------------
 CPU time ; time = 39.4531250000000 hundredths of a second
 Wall Time ; time = 39.3327951431274 hundredths of a second
 Increase the size of the arrays if this is <100
  and your clock precision is =<1/100 second
 ---------------------------------------------------
Function Max MB/s RMS time Min time Max time
Assignment: 153.2098 0.1064 0.1044 0.1181
Scaling : 143.8021 0.1123 0.1113 0.1138
Summing : 152.7340 0.1582 0.1571 0.1591
SAXPYing : 162.8487 0.1477 0.1474 0.1484
--------------------------------------
 Double precision appears to have 16 digits of accuracy
 Assuming 8 bytes per DOUBLEPRECISION word
--------------------------------------
 CPU time ; time = 40.2343750000000 hundredths of a second
 Wall Time ; time = 40.0159955024719 hundredths of a second
 Increase the size of the arrays if this is <100
  and your clock precision is =<1/100 second
 ---------------------------------------------------
Function Max MB/s RMS time Min time Max time
Assignment: 154.6553 0.1061 0.1035 0.1191
Scaling : 145.0748 0.1108 0.1103 0.1118
Summing : 150.8599 0.1600 0.1591 0.1610
SAXPYing : 162.8488 0.1477 0.1474 0.1489
--------------------------------------
 Double precision appears to have 16 digits of accuracy
 Assuming 8 bytes per DOUBLEPRECISION word
--------------------------------------
 CPU time ; time = 41.4062500000000 hundredths of a second
 Wall Time ; time = 41.5776014328003 hundredths of a second
 Increase the size of the arrays if this is <100
  and your clock precision is =<1/100 second
 ---------------------------------------------------
Function Max MB/s RMS time Min time Max time
Assignment: 150.3985 0.1085 0.1064 0.1220
Scaling : 142.5517 0.1136 0.1122 0.1152
Summing : 153.6885 0.1596 0.1562 0.1636
SAXPYing : 166.6112 0.1491 0.1440 0.1503

6 CPUs - add 6 tables together
--------------------------------------
 Double precision appears to have 16 digits of accuracy
 Assuming 8 bytes per DOUBLEPRECISION word
--------------------------------------
 CPU time ; time = 44.9218750000000 hundredths of a second
 Wall Time ; time = 44.8559999465942 hundredths of a second
 Increase the size of the arrays if this is <100
  and your clock precision is =<1/100 second
 ---------------------------------------------------
Function Max MB/s RMS time Min time Max time
Assignment: 132.2052 0.1232 0.1210 0.1366
Scaling : 121.4329 0.1326 0.1318 0.1333
Summing : 135.8573 0.1781 0.1767 0.1796
SAXPYing : 142.9660 0.1693 0.1679 0.1708
--------------------------------------
 Double precision appears to have 16 digits of accuracy
 Assuming 8 bytes per DOUBLEPRECISION word
--------------------------------------
 CPU time ; time = 44.5312500000000 hundredths of a second
 Wall Time ; time = 44.6607947349548 hundredths of a second
 Increase the size of the arrays if this is <100
  and your clock precision is =<1/100 second
 ---------------------------------------------------
Function Max MB/s RMS time Min time Max time
Assignment: 132.2052 0.1231 0.1210 0.1376
Scaling : 121.4329 0.1327 0.1318 0.1333
Summing : 133.6423 0.1804 0.1796 0.1811
SAXPYing : 146.3701 0.1673 0.1640 0.1679
--------------------------------------
 Double precision appears to have 16 digits of accuracy
 Assuming 8 bytes per DOUBLEPRECISION word
--------------------------------------
 CPU time ; time = 45.3125000000000 hundredths of a second
 Wall Time ; time = 45.3439950942993 hundredths of a second
 Increase the size of the arrays if this is <100
  and your clock precision is =<1/100 second
 ---------------------------------------------------
Function Max MB/s RMS time Min time Max time
Assignment: 133.2800 0.1231 0.1200 0.1396
Scaling : 123.2590 0.1304 0.1298 0.1314
Summing : 131.4983 0.1836 0.1825 0.1850
SAXPYing : 144.6480 0.1666 0.1659 0.1669
--------------------------------------
 Double precision appears to have 16 digits of accuracy
 Assuming 8 bytes per DOUBLEPRECISION word
--------------------------------------
 CPU time ; time = 46.8750000000000 hundredths of a second
 Wall Time ; time = 46.7104077339172 hundredths of a second
 Increase the size of the arrays if this is <100
  and your clock precision is =<1/100 second
 ---------------------------------------------------
Function Max MB/s RMS time Min time Max time
Assignment: 134.3724 0.1221 0.1191 0.1396
Scaling : 125.1408 0.1289 0.1279 0.1298
Summing : 130.1067 0.1860 0.1845 0.1884
SAXPYing : 144.6481 0.1660 0.1659 0.1665
--------------------------------------
 Double precision appears to have 16 digits of accuracy
 Assuming 8 bytes per DOUBLEPRECISION word
--------------------------------------
 CPU time ; time = 46.4843750000000 hundredths of a second
 Wall Time ; time = 46.3199973106384 hundredths of a second
 Increase the size of the arrays if this is <100
  and your clock precision is =<1/100 second
 ---------------------------------------------------
Function Max MB/s RMS time Min time Max time
Assignment: 134.3725 0.1218 0.1191 0.1396
Scaling : 125.1409 0.1280 0.1279 0.1288
Summing : 128.7443 0.1874 0.1864 0.1884
SAXPYing : 144.6481 0.1668 0.1659 0.1675
--------------------------------------
 Double precision appears to have 16 digits of accuracy
 Assuming 8 bytes per DOUBLEPRECISION word
--------------------------------------
 CPU time ; time = 48.0468750000000 hundredths of a second
 Wall Time ; time = 48.0767965316772 hundredths of a second
 Increase the size of the arrays if this is <100
  and your clock precision is =<1/100 second
 ---------------------------------------------------
Function Max MB/s RMS time Min time Max time
Assignment: 131.1474 0.1250 0.1220 0.1415
Scaling : 123.2590 0.1314 0.1298 0.1337
Summing : 131.7870 0.1856 0.1821 0.1893
SAXPYing : 142.9661 0.1691 0.1679 0.1718

8 CPUs - add 8 tables together
--------------------------------------
 Double precision appears to have 16 digits of accuracy
 Assuming 8 bytes per DOUBLEPRECISION word
--------------------------------------
 CPU time ; time = 51.9531250000000 hundredths of a second
 Wall Time ; time = 52.2159934043884 hundredths of a second
 Increase the size of the arrays if this is <100
  and your clock precision is =<1/100 second
 ---------------------------------------------------
Function Max MB/s RMS time Min time Max time
Assignment: 113.0582 0.1443 0.1415 0.1591
Scaling : 102.4590 0.1575 0.1562 0.1591
Summing : 114.9073 0.2103 0.2089 0.2118
SAXPYing : 121.1338 0.2001 0.1981 0.2026
--------------------------------------
 Double precision appears to have 16 digits of accuracy
 Assuming 8 bytes per DOUBLEPRECISION word
--------------------------------------
 CPU time ; time = 55.4687500000000 hundredths of a second
 Wall Time ; time = 55.4368019104004 hundredths of a second
 Increase the size of the arrays if this is <100
  and your clock precision is =<1/100 second
 ---------------------------------------------------
Function Max MB/s RMS time Min time Max time
Assignment: 113.0582 0.1448 0.1415 0.1610
Scaling : 101.8226 0.1584 0.1571 0.1601
Summing : 114.9073 0.2101 0.2089 0.2108
SAXPYing : 121.7335 0.2011 0.1972 0.2030
--------------------------------------
 Double precision appears to have 16 digits of accuracy
 Assuming 8 bytes per DOUBLEPRECISION word
--------------------------------------
 CPU time ; time = 58.9843750000000 hundredths of a second
 Wall Time ; time = 59.0079903602600 hundredths of a second
 Increase the size of the arrays if this is <100
  and your clock precision is =<1/100 second
 ---------------------------------------------------
Function Max MB/s RMS time Min time Max time
Assignment: 113.0583 0.1444 0.1415 0.1601
Scaling : 103.1035 0.1565 0.1552 0.1577
Summing : 113.8434 0.2114 0.2108 0.2128
SAXPYing : 121.7335 0.1984 0.1972 0.2001
--------------------------------------
 Double precision appears to have 16 digits of accuracy
 Assuming 8 bytes per DOUBLEPRECISION word
--------------------------------------
 CPU time ; time = 54.2968750000000 hundredths of a second
 Wall Time ; time = 54.2255997657776 hundredths of a second
 Increase the size of the arrays if this is <100
  and your clock precision is =<1/100 second
 ---------------------------------------------------
Function Max MB/s RMS time Min time Max time
Assignment: 112.2839 0.1456 0.1425 0.1659
Scaling : 104.4168 0.1538 0.1532 0.1552
Summing : 110.4728 0.2178 0.2172 0.2186
SAXPYing : 122.9508 0.1963 0.1952 0.1981
--------------------------------------
 Double precision appears to have 16 digits of accuracy
 Assuming 8 bytes per DOUBLEPRECISION word
--------------------------------------
 CPU time ; time = 54.6875000000000 hundredths of a second
 Wall Time ; time = 54.7135949134827 hundredths of a second
 Increase the size of the arrays if this is <100
  and your clock precision is =<1/100 second
 ---------------------------------------------------
Function Max MB/s RMS time Min time Max time
Assignment: 113.0583 0.1450 0.1415 0.1649
Scaling : 105.0862 0.1525 0.1523 0.1532
Summing : 109.7775 0.2206 0.2186 0.2216
SAXPYing : 122.9508 0.1966 0.1952 0.1977
--------------------------------------
 Double precision appears to have 16 digits of accuracy
 Assuming 8 bytes per DOUBLEPRECISION word
--------------------------------------
 CPU time ; time = 55.4687500000000 hundredths of a second
 Wall Time ; time = 55.3968071937561 hundredths of a second
 Increase the size of the arrays if this is <100
  and your clock precision is =<1/100 second
 ---------------------------------------------------
Function Max MB/s RMS time Min time Max time
Assignment: 111.5201 0.1466 0.1435 0.1708
Scaling : 105.0862 0.1526 0.1523 0.1538
Summing : 110.7665 0.2192 0.2167 0.2216
SAXPYing : 124.1927 0.1958 0.1932 0.1968
--------------------------------------
 Double precision appears to have 16 digits of accuracy
 Assuming 8 bytes per DOUBLEPRECISION word
--------------------------------------
 CPU time ; time = 56.2500000000000 hundredths of a second
 Wall Time ; time = 56.3728094100952 hundredths of a second
 Increase the size of the arrays if this is <100
  and your clock precision is =<1/100 second
 ---------------------------------------------------
Function Max MB/s RMS time Min time Max time
Assignment: 112.2839 0.1462 0.1425 0.1718
Scaling : 107.1467 0.1503 0.1493 0.1523
Summing : 109.2896 0.2220 0.2196 0.2245
SAXPYing : 122.9508 0.1963 0.1952 0.1972
--------------------------------------
 Double precision appears to have 16 digits of accuracy
 Assuming 8 bytes per DOUBLEPRECISION word
--------------------------------------
 CPU time ; time = 55.8593750000000 hundredths of a second
 Wall Time ; time = 56.1776041984558 hundredths of a second
 Increase the size of the arrays if this is <100
  and your clock precision is =<1/100 second
 ---------------------------------------------------
Function Max MB/s RMS time Min time Max time
Assignment: 111.5201 0.1466 0.1435 0.1679
Scaling : 105.0862 0.1540 0.1523 0.1552
Summing : 108.8061 0.2230 0.2206 0.2274
SAXPYing : 121.1338 0.1994 0.1981 0.2011



This archive was generated by hypermail 2b29 : Tue Apr 18 2000 - 05:23:04 CDT