tj@utcs.uucp (tj) (12/06/85)
We had a guy come talk to us about this. In particular he said that the Western Digital controller's default interleave factor of 3 is too low and thus you are missing sectors and waiting for full revolutions of the disk. You can almost double your performance (half the access time) by setting it to 5 (maybe 4 but thats close) Appreciate that this is what this guy said and that he adds hard disks to systems for a living almost and he has spent more time formatting drives at the low level than I have spent formatting floppies so I think he might have something here. The real important thing that came out of his discussion is that performance suffers a little when interleave is too high but suffers greatly when it is too low. Speed up your hard drives for free guys, just spend some time playing. (On western digital controllers you can get at the low level formatting routines in ROM from debug using G=c800:5 I think (rumour only... worked on a friends)) t.jones (now if only I had the money to play with these things myself...)
andy@sdcarl.UUCP (Andrew Voelkel) (12/18/85)
Question:
Is the most desirable interleave factor affected by cpu clock speed. I am sure
those of us with 8 mhz machines would love to know.
--
Andrew Voelkel
{ucbvax,ihnp4,akgua,hplabs,sdcsvax}!sdcarl!andytimothym@tekigm2.UUCP (Timothy D Margeson) (12/21/85)
In article <273@sdcarl.UUCP> andy@sdcarl.UUCP (Andrew Voelkel) writes: > >Question: > >Is the most desirable interleave factor affected by cpu clock speed. I am sure >those of us with 8 mhz machines would love to know. Yes, I have a Compaq Deskpro running at 7.14 Mhz. I have tried four different interleaves (yesterday on my machine, today on an ATT-6300). I have WD WX2 controller, with the default interleave of 3. I have been living with it for almost a year. Byte had an article saying 3 is slow, 7 is better, so I tried it. BYTE is right. To do a 112k file copy and load of a 272k program, the times are listed below: Operation Source Destination File Size Time in Seconds Interleave Copy Goo/B Foo 112,276 10.89 3 Copy Goo/B+Goo/B Foo 224,552 19.48 3 Load Framework II Memory ~272,000 29.27 3 Copy Goo/B Foo 112,276 8.26 5 Copy Goo/B+Goo/B Bar 224,276 15.52 5 Load Framework II Memory ~272,000 25.91 5 Copy Goo/B Foo 112,276 4.66 6 Copy Goo/B+Goo/B Bar 224,276 8.28 6 Load Framework II Memory ~272,000 19.40 6 Copy Goo/B Foo 112,276 5.11 7 Copy Goo/B+Goo/B Bar 224,276 10.91 7 Load Framework II Memory ~272,000 21.45 7 Note that the /B switch causes PC/MS-DOS to copy the entire file as binary. I have also tried these benchmarks on the ATT-6300 as noted above with similar results. The ATT has an 8Mhz clock. The BYTE article suggested 6 as an interleave for IBM PC-AT's and 7 for normal IBM PC-XT's. Although the AT has a full 16 bit wide harddisk buss and the ATT and Compaq's are only 8 bit busses, an interleave of 6 still seems best. This is contradictory to WD and XEBEC ads. (Again, the Compaq Deskpro has the WD 1002-WX2 controller and a Seagate ST412 drive, the ATT has a XEBEC IBM work- a-like, with a Green Mountain drive). The defaults for both these systems was an interleave of 3, not 7, the value used by IBM. For your information......Hope it helps you all! -- Tim Margeson (206)253-5240 tektronix!tekigm2!timothym @@ 'Who said that?' PO Box 3500 d/s C1-465 Vancouver, WA. 98665