Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!uwm.edu!zaphod.mps.ohio-state.edu!swrinde!ucsd!nosc!crash!pnet01!jca From: jca@pnet01.cts.com (John C. Archambeau) Newsgroups: comp.sys.ibm.pc Subject: Re: Disk interleave. Message-ID: <3000@crash.cts.com> Date: 5 Jun 90 03:06:20 GMT Sender: root@crash.cts.com Organization: People-Net [pnet01], El Cajon CA Lines: 72 jawitz@ursa.UUCP (Eric Jawitz) writes: > writes - > >>Don't forget to ask for 1:1 interleave for the controller card !. >>I had some experience to purchase computer(s) from CompuAdd, they >>usually give us 3:1 or 2:1 if we didn't ask. > >I'm curious to know what a 1:1 interleave disk controller card does for >performance. I have two 40 meg. Seagate (ST 251-1) drives in my 386/25 >system on which I recently ran SpinRite which supposedly optimizes >the disk interleaving. SpinRite reported that the interleaving was set >at 3:1 (yielding about a 176K/sec. throughput) and that the optimum >interleave would be 2:1 (resulting in a 50% greater throughput). So I >let SpinRite do its stuff and change the interleaving. The Norton SI >disk index shows no change in performance between 3:1 and 2:1 disk >interleaving (both yield an index of 2.6). I assume this is because >SI does not measure throughput to derive the index. On the other hand, >the Checkit disk benchmark shows a throughput of about 250K/sec. with 2:1 >compared with 176K/sec. with 1:1. It seems, therefore, that SpinRite did >make a difference. > >The talk about a 1:1 interleave disk controller puzzles me because if >the interleave can be changed with software, what's the need for a >special controller card? How is a 1:1 controller any different from >my (apparently) 3:1 controller? Should I consider getting one? > >On a related note, my system has a selectable bus speed (8/12 megahertz I >believe). I realize that some cards may not work at the higher speed. However >assuming that everything checks out OK, are there any drawbacks to increasing >the speed? Would the system run hotter? Would I be shortening the life >of my boards? Am I likely to see an increase in disk throughput? Is this >in turn likely to affect the optimum interleaving for my system? > >If you have insights into any or all of the above questions, I'd be most >grateful to hear from you. Your 1:1 interleave controllers have the ability to buffer an entire track on the card. I interleave the drive at what the controller is rated at. My current controller is a 2:1 WA3-16 (WD1003-WA2 clone) and the optimal interleave via inspection was what the manufacturer rated the controller at which is 2:1. If you jump to a WD1006V controller and interleave the drive at 1:1, I'd be surprised if you didn't get a performance increase. Yes, it is true you can change the interleave via software, the issue of optimal interleave though is if your hardware configuration can handle 1:1 or not and yield optimal performance. So, thus, the interleave of the controller is the optimal interleave the controller is rated at by the manufacturer of the controller. As for bus speed, it may or may not. I doubt it would since it's your controller that's reading the drive, it's only shooting the data across the bus. A word of warning, if you deviate from the de facto ISA bus speed standard of 8 MHz, you can damage cards that are pushed at a higher speed. I killed an internal modem that couldn't handle the increase in bus speed when I pulled it out of my XT and put it into my 286 a couple of years back. I neglected to configure the bus speed at 8 MHz. My drive controller worked just fine since it's rated to run up to 16 MHz. So a word to the wise, unless you buy high speed cards, stay at the de facto standard of 8 MHz on the ISA bus. // JCA /* **--------------------------------------------------------------------------* ** Flames : /dev/null | Small memory model only for ** ARPANET : crash!pnet01!jca@nosc.mil | Unix? Get the (*bleep*) out ** INTERNET: jca@pnet01.cts.com | of here! ** UUCP : {nosc ucsd hplabs!hd-sdd}!crash!pnet01!jca **--------------------------------------------------------------------------* */