jcurtis@simasd.navy.mil (Jay Curtis) (09/09/90)
Thanks to all who responded to my request for SCSICNTL.EXE. I have
received a copy of it from several sources. I have one furthur problem.
It won't run on my system! The aha-1542b is set up with all the default
parameters except the ROM location which I set to CC00 vice DC00. For
some reason the system wouldn't find the Adaptec bios at DC00.
Any way, what happens is the copywrite screen displays, then a window
pops up with the message "Unable to get inquiry data from host adapter".
The program then terminates. The docs don't explain what to do in case
the program won't run. Any one ever seen this problem before?
Here are the results of the Byte Benchmark file system tests.
Start Benchmark Run (BYTE Version 2.6)
Fri Sep 7 22:54:48 PDT 1990
1 interactive users.
sysname=XENIX
nodename=simasd
release=2.3.3
version=SysV
machine=i80386
=============================================================================
Filesystem Throughput Test:
-----------------------------------------------------------------------------
Test Time: 1 secs
Arithmetric Geometric Variance
Mean Mean (6 tests)
Read (Kbytes/sec): 434 417 13406.27
Write (Kbytes/sec): 613 613 2.97
Copy (Kbytes/sec): 177 170 3437.47
-----------------------------------------------------------------------------
Test Time: 10 secs
Arithmetric Geometric Variance
Mean Mean (6 tests)
Read (Kbytes/sec): 91 90 152.67
Write (Kbytes/sec): 71 69 266.67
Copy (Kbytes/sec): 49 49 69.87
-----------------------------------------------------------------------------
Test Time: 20 secs
Arithmetric Geometric Variance
Mean Mean (6 tests)
Read (Kbytes/sec): 53 53 4.17
Write (Kbytes/sec): 72 71 170.00
Copy (Kbytes/sec): 33 33 6.97
1 interactive users.
End Benchmark Run (Fri Sep 7 22:54:48 PDT 1990) ....
This is on a 33MHz 386 with an AHA-1542b and an Imprimiss Wren V
hardrive! I ran the same test on an MFM drive attached to a wd1003.
Drive was formated at three to one interleave. The numbers were approx.
3 times as great.
Any ideas? I have changed the Xenix configure parameters to match the
ones suggested by the tuning portion of the SCO manuals. I haven't tried
the test with the stock, out of the box parameters in the Kernel. I
suspect that it wouldn't make much difference in this case.
--Jay
--
Jay Curtis
jcurtis@simasd.UUCP
{nosc;ncr-sd;ucsd;}!simasd!jcurtis
Of course my opinions are my own... Who would let me speak for them?!caf@omen.UUCP (Chuck Forsberg WA7KGX) (09/10/90)
You may have to slow down the CPU in order to get scsicntl.exe to work properly. With the track read buffer turned on, I saw Xenix file reads increase from 36 kHz to up to say 200 kHz, extremely dependent on how fragmented the file was. On SCO Unix I now get around 400 to just over 600 kHz when doing "time cp file /dev/null", depending on how fragmented the file is and how many daemons go bump in the night. Backups to tape are very much faster except for directories with many tiny files. My file browser program loads files in a flash. But you won't see much difference in most commands. After all, what compiler can do 400,000 lines per minute on a 386?