ARP 1.1 (Disk 85) (Mar 1988) : DriveTest / autotest.doc

     Autotest is a program to test the low-level performance of disk drives,
the associated hardware interface, and the driver software.  This program
was originally written to compare hard disk drives on the IBM PCs (See
"Fixed-disk Benchmarks" by William Hunt in the November 1984 issue of PC
Tech Journal).  I converted it to run on the AMIGA to test the performance
of the hard disk, interface, and driver on my AMIGA.

     This program performs two basic tests:

1)   Raw transfer rate.  20 reads of 1 sector, 8 sectors, 16 sectors,
     and 24 sectors are performed.  The average time per read is reported
     for each transfer size.  This will give an indication of how fast
     data can be transferred from the disk.  I use it varying the interleave
     on the hard disk to get the smallest interleave.

2)   Seek speeds.  The "Random reads" test reads the disk varying the sector
     position from .1 to .9 times the full disk capacity, first with 1 sector
     reads, then with 8 sector reads.  The average time per read is again
     reported.  This test will give an indication of the seek times of the
     hard disk, as well as the effect of multiple sector reads combined with
     seeks.

     To run autotest, just enter the CLI command:
          autotest <drive name>
where <drive name> is the disk drive name, i.e. DF0: or HD2:.  The test is
non-destructive (does not do any writes to the disk).


     The files floppy.rpt and harddisk.rpt are the results on my floppy
drive and the current hard disk drive (a 42MB partition on a SeaGate ST4096,
96MB unformatted, 28ms average seek time).

     NOTE:  The times reported for the 2 "random read" tests will vary
between drives with different capacities, since the width of the seek is
a percentage of the total disk capacity.  With drives of equivalent capacity,
the results will generally show the difference between the slower step rate
drives and the faster drives.

     ALSO NOTE:  The transfer rate test measures the raw read capability of
the drive and does not take into consideration the type of I/O done by
AmigaDOS.  It may be that a disk with a very fast transfer rate may run slower
when doing normal AmigaDOS operations.  Multi-tasking will also vastly affect
the disk performance (although a faster seeking drive should perform better).

               Michael L. Hitch
               Office of Systems and Computing Services
               Montana State University