karamich@uiucuxc.CSO.UIUC.EDU (11/03/86)
I have a ST225 with a Western Digital combo hard/floppy controller.
I've had no problems so far (about a year now). I did a low-level
format using debug, with 1 way interleaving. The drive formatted
perfectly. NO BAD SECTORS. Freaked me out! But I wouldn't use the
original PC power supply to handle the hard drive. Who knows
what might fry .........
tom karamichos
U. of Illinois at Urbana-Champaign
Computing Services Office -- Microcomputer Consulting
karamich@uxc.CSO.UIUC.EDU
ihnp4!pyrchi!uiucuxc!karamich
-------------------------------------------------------------
Women -- you can't live without them, you can't shoot them!
-------------------------------------------------------------timothym@tekigm2.UUCP (Timothy D Margeson) (11/07/86)
In article <21900185@uiucuxc> karamich@uiucuxc.CSO.UIUC.EDU writes: > I have a ST225 with a Western Digital combo hard/floppy controller. > I've had no problems so far (about a year now). I did a low-level > format using debug, with 1 way interleaving. The drive formatted > perfectly. NO BAD SECTORS. Freaked me out! But I wouldn't use the > original PC power supply to handle the hard drive. Who knows > what might fry ......... > tom karamichos I can't believe people do the things they do so blindly. Does this person thinks he is gaining something by setting the interleave factor to 1? He more than likely destroyed his system performance by doing this. At the least he made a 40% reduction in disk access performance. For your own good, when you use an interleave other than that recommended by the system manufacturer, do a few benchmarks to see if there is an increase in system throughput. The best test I have found is to change the interleave, format the disk, hi-level DOS format the disk, install a BIG program (200 to 400k), then run the program, and time how long it takes to see my prompt. I then redo the process using a different interleave. The range I go through is 2 to 8 for IBM PC's. 1 through 8 for 8MHz AT's. The best interleave for my machine, a Compaq Deskpro (8086 type), is 6. Now my machine isn't quite as fast as an AT, and doesn't have a 16bit disk data path, but it is still a fast machine (3 times the typical PC), but an inter- leave of 3 (recommended by Western Digital) slows my system down by 30%. I have tried an ATT-PC-6300 with the exact same results, 6 is the best inter- leave factor for the machine (again we're talking about an 8 MHz 8086 based product with an 8 bit bus to disk controller path). Optimum interleave factors will vary between different makes of disk control devices, and between different uProcessor system bus configurations (e.g. an 8 MHz 8086 vs, a 4.77 MHz 8088, an 8 bit controller vs. a 16 bit controller), but not between hard disks on the same controller. THE MAIN THING IS TRY DIFFERENT INTERLEAVES BEFORE PUTTING YOUR SYSTEM BACK TOGETHER! THAT WAY YOU WILL KNOW WHAT INTERLEAVE IS BEST FOR YOUR SYSTEM ! A few more notes, ALL ST506 or ST412 disk drives use an 8 bit data path, an most typically spin at about 3600 RPM. Data goes on and comes off the disk at about 5 megabyts per second, WITH A GOOD CONTROLLER AND FAST MEMORY BUS! An IBM PC, with a standard controller (WD or Xebec), has a data transfer rate to the disk drive of .5 to 2 megabytes per second. This being limited by the memory bus bandwidth, and the disk controller's inefficiencies. AN IBM PC-AT, with it's standard controller, moves along at somewhere around 2 to 4 megabytes per second (on a good day). With a good and fast cacheing controller (capable of >15 sector read and write cache), you can push data rates near the 5 megabyte per second disk limit, but this will be true only for the one cyclinder you are writing to (in some cases this may be as many as 16 tracks on an 8 platter disk). After this single track write, disk access times come into play to limit system performance. In my experience, I have seen an IBM PC (8088 at 4.77 MHz) outperform my Compaq Deskpro (8086 at 7.16 MHz) in disk I/O. The reason for this is due to my Western Digital Controller being SLOW. I have optimized the interleave for best system perfomance, but there is data rate problems within the controller itself. The IBM PC that is faster than my Deskpro (for DISK I/O) has an Adaptec 2010A controller (not the RLL model). Both controllers use the ST412 interface into Seatgate ST425 disk drives. My Deskpro performs best at I=6, the PC works best at I=2. The difference being the controllers internals. The actual times recorded for completeing a 226k file copy are 6.9 seconds for the PC, and 7.5 seconds for my Deskpro. Similar results are obtained for loading and executing programs. An IBM PC-AT on the other hand completes the same copy in 6.2 seconds. Keep in mind that the above figures are with VERIFY ON. Verify off is a state that I NEVER run my computer in. These figures indicate that the PC comes very close to AT performance when the hard disk is being accessed. All because of the disk controller, now if only the PC had a fast bus to keep up with the controller. Finally, it would be nice if controller manufacturers specified their data rates given some standard (fast) bus architecture, say an 8 megahertz 8086 based machine for 8 bit controllers, or an 8 megahertz 80286 for 16 bitters. Enough raving.... off to work, I've wasted more than one lunch hour. -- Tim Margeson (206)253-5240 PO Box 3500 d/s C1-937 @@ 'Who said that?' Vancouver, WA. 98668 {allegra..inhp4..decvax..ucbvax}!tektronix!tekigm2!timothym