home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
OS/2 Shareware BBS: 8 Other
/
08-Other.zip
/
diski117.zip
/
diskio.doc
< prev
next >
Wrap
Text File
|
1998-07-05
|
7KB
|
158 lines
DISKIO - Fixed Disk Benchmark, Version 1.15
(C) 1994-1998 Kai Uwe Rommel
Dhrystone 2.1 C benchmark routines (C) 1988 Reinhold P. Weicker
This is a simple fixed disk and CD-ROM drive benchmark program for
OS/2 and Windows NT. It measures (approximately) the following
numbers:
Hard disks:
- drive cache/bus transfer rate (introduced with version 1.6)
- data transfer rate (on inner and outer cylinders)
- CPU usage of full speed disk I/O (introduced with version 1.9)
- average latency time (discontinued with version 1.15)
- average data access time
- multi threaded I/O throughput (introduced with version 1.15)
- concurrent data transfer on all disks (introduced with version 1.15)
CD-ROM drives (introduced with version 1.10):
- data transfer rate (at beginning and nearly end of the inserted CD)
- CPU usage of full speed data reads (in the middle of the inserted CD)
- average data access time
Concurrent hard disk and CD-ROM drive benchmark (introduced with version 1.10):
- hard disk data transfer rate
- CD-ROM drive data transfer rate
- CPU usage for concurrent full speed I/O on both devices
Note: for all CD-ROM related tests, a CD medium should be inserted that
contains at least 600 MB of data as a single track, to get comparable
results (do not use a CD-R or CD-RW media). Since some drives cache
data, delayed spin-up can cause wrong results. Therefore, you
should remove and re-insert the test CD immediately before running
this benchmark program.
The program also measures CPU speed without disk I/O using the Dhrystone
CPU benchmark which is needed as calibration for determining the CPU load
caused by disk I/O. This programs dhrystone results may vary compared to
those of other programs. However, since the dhrystone results are used
for relative comparisons within the program, the absolute values do no
have any influence on the CPU usage results of the I/O benchmarks.
The CPU load caused by the disk I/O of the benchmark routines is
"measured" by first calculating the Dhrystone benchmark result without
any disk I/O and then the number of Dhrystone's the CPU could run in
the background during full speed read operations on the devices. The
lower the percentage number, the less CPU time is (ab)used by the disk
interface. This depends on the quality and technology of the disk
interface (i.e. if it uses PIO or DMA, IDE or SCSI etc.), but also on
the speed of the host CPU and the speed of the devices.
Of course all values are influenced by other activity on the system,
so stop all other software if you want reliable results. The results
will also differ slightly between repeated runs.
The benchmark program needs about 115 seconds to complete on a typical
system with one hard disk and one CD-ROM drive. On system with more
disk drives it needs about 5 seconds to calibrate, 70 seconds per hard
disk, 30 seconds per CD-ROM drive and 20 seconds for the concurrency
benchmarks.
The individual benchmarks are run on all hard disks and CD-ROM drives
in the system, if started without arguments. The concurrency benchmarks
are run using all hard disks (for the all disks concurrent read benchmark)
and the first hard disk and first CD-ROM drive in the system (for the
hard disk and CD-ROM concurrent read benchmark).
To run it only for a specific drives, use command line arguments.
Enter "diskio -?" to get a syntax description for the command line
arguments.
A few hard disk sample results, numbers listed are the data transfer
rate on track 0 (k/sec), latency time (ms), access time (ms),
cache/bus speed (k/sec), CPU usage (last two numbers only if result
was available):
- Pentium PCI/SCSI systems:
Quantum Atlas + NCR 53c810 + PPro-200: 6932, 0.6, 14, 8060, 15%
Quantum Atlas + NCR 53c810 + P-100: 6914, 0.8, 14, 7515, 18%
Quantum Atlas + AHA-2940 + P-133: 6800, 1.3, 13, 6900, 20%
HP C3323 + NCR 53c810 + P-100: 4569, 1.2, 18, 7016, 21%
IBM DPES-31080 + NCR 53c810 + P-100: 4186, 1.1, 18, 7008, 15%
DEC DSP-3160 + NCR 53c810 + P-120: 3930, 2, 16, 6370, 15%
DEC DSP-3133L + NCR 53c810 + P-100: 4393, 1.5, 17, 6124, 22%
- Pentium PCI/EIDE (Intel Triton/Triton-II) systems:
Quantum LPS-540A + P-100: 2177, 1.3, 19, 2227, 92%
Quantum LPS-540A + P-100 (TRIOS2.ADD): 3532, 1.5, 20, 6083, 9%
Quantum Fireball 1280A + P-100: 6188, 0.7, 16, 6294, 89%
- 486 PCI/SCSI systems:
HP C3323 + NCR 53c810 + 486-100: 3945, 2, 18, 5045, 54%
IBM DPES-31080 + NCR 53c810 + 486-100: 4239, 1.3, 18, 6613, 46%
DEC DSP-3160 + NCR 53c810 + 486-66: 3930, 2, 16
DEC DSP-3053L + NCR 53c810 + 486-66: 4130, 3, 16
Quantum LPS-340S + NCR 53c810 + 486-66: 1900, 5, 23
Maxtor 7345S + NCR 53c810 + 486-66: 1881, 10, 17
- 486 PCI/IDE systems:
Quantum LPS-540A + IBM1S506.ADD + 486-100: 1796, 1.6, 20, 1820, 94%
- 486 EISA/SCSI systems:
DEC DSP-3210 + AHA-2740 + 486-50: 3650, 2, 17
DEC DSP-3160 + AHA-1742 + 486-66: 3050, 3, 18
- 486 ISA systems:
Quantum LPS-340S + AHA-1542CF + 486-50: 1508, 7, 26
Quantum LPS-540A + IDE + 486-66: 1492, 4, 21
And a few CD-ROM sample results, listing transfer rate (k/sec),
access time (ms) and CPU usage (%):
- Pentium PCI/SCSI systems:
Toshiba 3701 (6.6x) + NCR 53c810 + PPro-200: 990, 160, 14%
Plextor 8x + NCR 53c810 + P-100: 1222, 134, 29%
Plextor 12x + NCR 53c810 + P-100: 1722, 123, 37%
- Pentium PCI/EIDE (Intel Triton/Triton-II) systems:
Sony CDU-311 (8x) + P-100: 1209, 175, 49%
Concurrency benchmark sample results, listing hard disk and CD-ROM
transfer rate (k/sec) and CPU usage (%):
- Pentium PCI/SCSI systems:
Quantum Atlas + Toshiba 3701 (6.6x) + NCR 53c810 + PPro-200: 5121, 861, 21%
Quantum Atlas + Plextor 8x + NCR 53c810 + P-100: 4975, 1206, 38%
- Pentium PCI/EIDE (Intel Triton/Triton-II) systems:
Quantum Fireball 1280A + Sony CDU-311 (8x) + P-100: 3565, 1165, 99%
The above values are getting old and outdated. However, I have not
enough time to keep an up to date list of typical results ...
There is no warranty. Use this software on your own risk. Due to the
complexity and variety of today's hardware and software which may be
used to run this program, I am not responsible for any damage or loss of
data caused by use of this software. It was tested very well and is
expected to work correctly, but nobody can actually guarantee this for
any circumstances. And because this software is free, you get what you
pay for ...
This program can be used freely for private or educational purposes.
If you want to use it for commercial purposes or want to integrate it
into another package or find any bugs or have suggestions about
further enhancement, please contact the author.
Author: Kai Uwe Rommel
Muenchen, Germany
Fax: +49 89 324 4524
E-Mail: rommel@ars.de (preferred)
rommel@leo.org