john@ativax.UUCP (John Behrs) (03/08/88)
Although the IBM-AT disk controller runs with a sector interleave of 1, it misses consecutive sectors. Sector interleave of 2 or 3 is optimal, even on 16MHZ clones. Does anyone know the reason for this and if it is possible to make an IBM-AT run optimally with no sector interleave? -- John Behrs @ AEON Technologies boulder!fesk!ativax!john 2888 Bluff Street Suite 227 Boulder CO 80301 (303) 499-7304
sl@van-bc.UUCP (pri=-10 Stuart Lynne) (03/09/88)
In article <103@ativax.UUCP> john@ativax.UUCP (John Behrs) writes: >Although the IBM-AT disk controller runs with a sector interleave of 1, >it misses consecutive sectors. Sector interleave of 2 or 3 is optimal, >even on 16MHZ clones. Does anyone know the reason for this and if it >is possible to make an IBM-AT run optimally with no sector interleave? The standard WD controller only buffers one sector from the disk. When it successfully reads the sector in (ie, CRC is correct) you (well actually the disk driver) are notified that you can download it. The controller is unable to start reading the next sector until you have finished empting the sector buffer. With a 1-1 interleave you only have the inter sector gap time, probably under 50 microseconds. (The disk rotates at 3600 RPM, or 60 cps, giving a rotation time of 17 milliseconds, there are typically 17 sectors per track, giving an overall time from the start of one sector to the next of 1 millisecond. Assume less than 5% overhead for inter-sector gap.) 50 microseconds is not enough time to download much :-) Even a 386 at 20Mhz will only get part way through. Western Digital has solved this problem however. Their new controller the WD-1006 has a buffer capable of holding the entire track, so that it can continue to read more sectors in while you start taking them out. Unfortunately this controller is currently only available sans floppy disk controller. No word yet on when it will be available with same. It is also available now with RLL support. -- {ihnp4!alberta!ubc-vision,uunet}!van-bc!Stuart.Lynne Vancouver,BC,604-937-7532
adonis@tahoe.unr.edu (Derrick Hamner) (03/11/88)
In article <103@ativax.UUCP> john@ativax.UUCP (John Behrs) writes: >Although the IBM-AT disk controller runs with a sector interleave of 1, >it misses consecutive sectors. Sector interleave of 2 or 3 is optimal, >even on 16MHZ clones. Does anyone know the reason for this and if it >is possible to make an IBM-AT run optimally with no sector interleave? >-- >John Behrs @ AEON Technologies boulder!fesk!ativax!john >2888 Bluff Street Suite 227 >Boulder CO 80301 >(303) 499-7304 The interleave is dependant upon how fast the controller can pass data, and how fast the computer can receive data. The optimal interleave will probably be different for almost every different type of computer and controller combination. On my AST Premuim/286 (10 Mhz. - 0 wait states) I find that an interleave of 2 is best. As far as I know, the only way to possibly reduce the interleave is to get a faster controller (or a faster computer :-). -- A computer program does what you tell it to do, | Derrick Hamner not what you want it to do. | adonis@tahoe.unr.edu
palowoda@megatest.UUCP (Bob Palowoda) (03/11/88)
in article <1686@van-bc.UUCP>, sl@van-bc.UUCP (pri=-10 Stuart Lynne) says: > > In article <103@ativax.UUCP> john@ativax.UUCP (John Behrs) writes: >>Although the IBM-AT disk controller runs with a sector interleave of 1, >>it misses consecutive sectors. Sector interleave of 2 or 3 is optimal, >>even on 16MHZ clones. Does anyone know the reason for this and if it >>is possible to make an IBM-AT run optimally with no sector interleave? There is an OMTI RLL AT HD controller (model ?) that clams to do this. In fact I was down at Sunnyvale Memories and seen it running on a 10Mhz AT clone. 485Kb a sec. I just talk to them recently and they said there was a bug in it, and that the new model should be in about a months time. The benchmark used was Coretest. ---Bob