ESDI vs SCSI

[Mike McNally]

In article <98835 at sun.Eng.Sun.COM> williamt at sun.UUCP (William A. Turnbow) writes:
>1)  I said bus transfer speed.  Not DMA.  I don't know if the DMA speed
>is different.
>
>2) Can you explain how bot the ESDI and SCSI drives you mention (1.2M/s and
>.6M/s) are faster than the maximum DMA rate you claim?  

I thought about that after I posted the note.  The SCSI drive, of course,
uses the interface I described which has its own DMA controller.  The
ESDI driver (which also does ST506) copies from the buffer on the 
controller.  I stand corrected.  (Of course, DMA is preferrable in a 
multi-tasking environment.)

>5) As for SCSI controllers supporting variable number of sectors per track,
>that is only in the interface presented to the user (computer).  Since
>the SCSI controllers have intelligence in them, they can remap the physical
>sectors in whatever fashion they choose.  This does not change the
>actual physical number of sectors/track, the spin rate, nor the final
>maximum transfer rate.  

This is not correct.  The CDC Wren IV definitely varies the number of
sectors per track across the disk.  (I intuit that outer tracks have
more sectors, but I haven't had any coffee yet so I could be wrong.)
This is the big advantage of the drive over an ESDI drive: the Wren has
an average of 45 sectors per track.  Given a controller that can keep
up (ours does), the Wren will of necessity show a higher transfer rate
(not to mention the fact that the ESDI interface won't go faster than
10 megabits/sec, while asynchronous SCSI is 16 megabits/sec (the WD
33C93A SCSI chip actually does up to 2.5 megabytes/sec, which is 24
megabits/sec.)).

>7) Be careful about how you measure disk xfer speed.  I have seen 'coretest'
>be totally fooled by one controller that kept an on board cache into believing
>it could xfer over 5Meg/sec on a 386.  Some tests also seem to be less
>than reliable on ESDI drives because it is apparently common or standard
>to include a track buffer.

We're all well aware here of how bogus a lot of these figures can be.  I
think drive manufacturers get their transfer rates by issuing one command
to read a contiguous megabyte and timing it.  Although we do perform that
test, we also test reading/writing to a regular file (in a UNIX-ish file
system), reading/writing to a contiguous file (direct transfers from 
user memory to/from the disk),  and reading/writing straight to the block
or raw disk (the block view of the disk goes through the disk cache in the
OS).  Modern SCSI disks have very low command overhead, so this is not
really an advantage of ESDI (with its own relatively complex command
setup).

Here's the beef: we can read from the Wren IV via SCSI at 1.4 megabytes per
second, and write at just over a megabyte per second.  ESDI drives don't
go that fast.  Of course, the Wren is rather expensive, but I can get 
40 or 80 megabyte Quantum SCSI drives that are about as fast as fast as
any ESDI disk on the block, and the Quantums are cheap (and small and
very quiet).  I can also plug in a SCSI TEAC streaming cartiridge drive and
back up 150 megabytes on one cassette.  Or, I can plug in my Qualstar
reel-to-reel and read 1600 bpi GNU tapes.

I should of course add that these arguments may not apply to all SCSI 
interfaces.  Our SCSI controller uses a WD 31C93A controller chip at
10 Mhz (16 Mhz for faster systems, like 25Mhz 386 machines) and a 
bursting DMA controller.  Your performance may vary.  I really don't
know much about other PC SCSI controllers.



-- 
Mike McNally                                    Lynx Real-Time Systems
uucp: {voder,athsys}!lynx!m5                    phone: 408 370 2233

            Where equal mind and contest equal, go.