ngeow@bu-cs.BU.EDU (Yee Ngeow) (12/17/89)
Hi! Just a question about hard disk transfer rates. I recently purchased a SCSI drive, the ST-02 (no flames please) and ST-157N. Running with the zero wait jumper enabled, with a 1:1 interleave I am getting 500K of transfer rate, but running at 2:1 I am getting 350K or so. I thought 1:1 is twice as fast as 2:1, since the disk must rotate twice to read all the sectors. Anyone know what is going on there? Also, how does one get the theoretical maximum transfer rate for any one encoding scheme (RLL/MFM/ESDI)? Doesn't the transfer rate be affected by the number of read/write heads you have on your drive? I though the sequence on sequential read/write on a hard disk is: (sector 0) head 0 1 2 3 .. (sector 1) head 0 1 2 3 .. (sector 2) head 0 1 .. It seems this way when I format a disk. So the more heads you have, the higher your transfer rate, since switching from head to head doesn't take any time?! Or is it the other way: (head 0) sector 0 1 2 3 4 ... 17 (head 1) sector 0 1 2 3 .... which provides a constant rate regardless of number of heads? Also, I though Winchester drives spin at a rate of 3000 revolution per second, as opposed to 300 rps on floppy drives. Assuming heads don't matter, the transfer rate formula on a 1:1 intereave is: 3000 * SECTOR/TRACK * SECTOR SIZE which will give me a HUGE number. Lets assume RLL, 26 sectors/track and 512 bytes/sector: 3000 * 26 * 512 = 39,000 Kbytes which is over 39 Megabytes/second, and is about 500 times greater then 7.5Mbits/second!! What is going on here??! I will appreciate if anyone can answer my question. Sorry if they seem naive, but I just can not figure it out.. Thanks in advance, Kwong
u-gclapp%ug.utah.edu@cs.utah.edu (Glenn Clapp) (12/17/89)
In article <44807@bu-cs.BU.EDU> ngeow@cs.bu.edu (Yee Ngeow) writes: >Hi! > > Just a question about hard disk transfer rates. I recently purchased a >SCSI drive, the ST-02 (no flames please) and ST-157N. Running with the >zero wait jumper enabled, with a 1:1 interleave I am getting 500K of >transfer rate, but running at 2:1 I am getting 350K or so. I thought 1:1 is >twice as fast as 2:1, since the disk must rotate twice to read all the sectors. >Anyone know what is going on there? > Well, if your drive has its optimum transfer rate at 2:1 (consecutive sectors are not really consecutive, but have another sector between them), then going to a 1:1 interleave will cause the drive to make a full EXTRA revolution+ to read the next sector, because the drive is too slow to be ready for it the first time it comes by. This will DRASTICALLY reduce your transfer rate. Going to a higher interleave such as 3:1 will only slightly reduce transfer rate as the drive will be waiting for data. Great utilities exist for checking and changing interleave such as Gibson Research's SpinRite II, but I don't think it works with SCSI drives. In any case, transfer over the SCSI may be very high, but it still has to get the data into the PC via the out-dated AT bus at 8MHz (some are faster, but this get's dicey (I know, just try an ATI VGA Wonder in a 12MHz bus and see what happens). The AT bus has a maximum *Theoretical* transfer rate of 1Mb/s, but in practice the maximum is somewhat less. Glenn
rfarris@serene.UUCP (Rick Farris) (12/17/89)
In article <44807@bu-cs.BU.EDU> ngeow@cs.bu.edu (Yee Ngeow) writes: > Winchester drives spin at a rate of 3000 revolution per second So close, Kwong, yet so far away :-). 5 1/4 inch winchesters spin at 3600 rotations per *minute*, not per second. That's 60 rps. That would make your calculation: 60 * 26 * 512 = 798,720 bytes/sec But consider what would happen if each head on a disk were feeding it's own buffer, so that one revolution could read an entire *cylinder*! A 12 head drive could approach 10 MB/s (80 Mbits/sec YOW!). It goes without saying that the controller would need to optimize cylinder writes.... Rick Farris RF Engineering POB M Del Mar, CA 92014 voice (619) 259-6793 rfarris@serene.uu.net ...!uunet!serene!rfarris serene.UUCP 259-7757
ralf@b.gp.cs.cmu.edu (Ralf Brown) (12/20/89)
In article <1989Dec16.192948.27568@hellgate.utah.edu> u-gclapp%ug.utah.edu@cs.utah.edu (Glenn Clapp) writes: }In any case, transfer over the SCSI may be very high, but it still has }to get the data into the PC via the out-dated AT bus at 8MHz (some }are faster, but this get's dicey (I know, just try an ATI VGA Wonder }in a 12MHz bus and see what happens). The AT bus has a maximum }*Theoretical* transfer rate of 1Mb/s, but in practice the maximum is }somewhat less. Better not tell the HD controllers with hardware disk caches that blast 3-4M/s over the AT bus.... The AT bus is nowhere near the bottleneck with current hard disk technology. MFM drives do up to 500K/s, RLL up to 750K/s, and ESDI do up to 1M/s for 10MHz controllers (there are also some 15 and 20MHz controllers available, but I wonder whether current hard disks could actually make use of the extra speed--you'd need about 68 sectors per physical track to max out a 20 MHz controller). Compare that to a bus bandwidth of 5.33M/s for an 8 MHz/1ws bus doing 16-bit transfers. SCSI can theoretically do 4M/s, but that assumes that both the hard disk and the host adapter can keep up (I think the new SCSI II does 10M/s). -- {backbone}!cs.cmu.edu!ralf ARPA: RALF@CS.CMU.EDU FIDO: Ralf Brown 1:129/46 BITnet: RALF%CS.CMU.EDU@CMUCCVMA AT&Tnet: (412)268-3053 (school) FAX: ask DISCLAIMER? | _How_to_Prove_It_ by Dana Angluin 9. proof by funding: What's that?| How could three different government agencies be wrong?
ralf@b.gp.cs.cmu.edu@canremote.uucp (ralf@b.gp.cs.cmu.edu) (12/21/89)
From: ralf@b.gp.cs.cmu.edu (Ralf Brown)
Orga: Carnegie-Mellon University, CS/RI
In article <1989Dec16.192948.27568@hellgate.utah.edu>
u-gclapp%ug.utah.edu@cs.utah.edu (Glenn Clapp) writes: }In any case,
transfer over the SCSI may be very high, but it still has }to get the
data into the PC via the out-dated AT bus at 8MHz (some }are faster,
but this get's dicey (I know, just try an ATI VGA Wonder }in a 12MHz
bus and see what happens). The AT bus has a maximum }*Theoretical*
transfer rate of 1Mb/s, but in practice the maximum is }somewhat less.
Better not tell the HD controllers with hardware disk caches that
blast 3-4M/s over the AT bus.... The AT bus is nowhere near the
bottleneck with current hard disk technology. MFM drives do up to
500K/s, RLL up to 750K/s, and ESDI do up to 1M/s for 10MHz
controllers (there are also some 15 and 20MHz controllers available,
but I wonder whether current hard disks could actually make use of
the extra speed--you'd need about 68 sectors per physical track to
max out a 20 MHz controller). Compare that to a bus bandwidth of
5.33M/s for an 8 MHz/1ws bus doing 16-bit transfers. SCSI can
theoretically do 4M/s, but that assumes that both the hard disk and
the host adapter can keep up (I think the new SCSI II does 10M/s).
--
{backbone}!cs.cmu.edu!ralf ARPA: RALF@CS.CMU.EDU FIDO: Ralf Brown
1:129/46 BITnet: RALF%CS.CMU.EDU@CMUCCVMA AT&Tnet: (412)268-3053
(school) FAX: ask DISCLAIMER? | _How_to_Prove_It_ by Dana Angluin
9. proof by funding: What's that?| How could three different
government agencies be wrong?
---
* Via MaSNet/HST96/HST144/V32 - UN IBM PC
* Via Usenet Newsgroup comp.sys.ibm.pc