Re: SUMMARY: SCSI-2 on SPARC-10 [ correction ]

From: Dave Olson (olson@anchor.esd.sgi.com)
Date: Sat Dec 19 1992 - 20:46:48 CST


In <BzEuxK.7KL@ra.nrl.navy.mil> wang@cedar.nrl.navy.mil (Dennis Wang) writes:
| Here is what I got back as answers to my query re SCSI-2
| (Interesting that two of the responses to a nationwide broadcast come from my own domain :)
|
| ----Begin Include ----
|
| From schuette@roadkill.nrl.navy.mil Tue Dec 15 16:40:58 1992
| Newsgroups: nrl.misc,comp.sys.sun.hardware
| Organization: Naval Research Lab, Physical Acoustics, Washington, DC
|
| you cannot mix fast and slow scsi. the drives will work, but only
| at slow scsi speeds. This is true for the Western Digital
| Version WD33C93B, revision C controllers.
| (which are used on newer SGI machines)

This is complete bull****. I just love people who post
authoritatively with no knowledge. The drives can be
mixed just fine, since the sync rate varies with the
connected target (or async for that matter).

Now, there *is* a grain of truth here, since the 93B has
different timings if you want to sustain 10 MB/sec than
the 93A did (not surprising, given that WD's chips run
at the same speed on both the SCSI and the host bus). Given
that, all the Indigo can sustain is 5.5 MB/sec, even though
10 MB/sec can be negotiated with the device. Future systems
have newer DMA engines, and will sustain the 10 MB/sec rate.

Of course, this is irrelevant to the Sun question, since they
use different hardware and software.

|
| From buck@ra.nrl.navy.mil Wed Dec 16 10:37:32 1992
| Return-Path: <cedar.nrl.navy.mil>
| Organization: Naval Research Lab, Washington, DC
| In article <BzBIoq.2zC@ra.nrl.navy.mil> you write:
| >Does anyone know of any problems using the SCSI-2 on a SPARC 10 ? I am getting some
| >conflicting stories.
| >
| >1. "No problem mixing fast SCSI-2 and slow SCSI" - Sun rep
|
| He is probably right, no problem unless you want the speed from your fast
| SCSI-2 devices. If you mix them you will get the speed of the slowest device
| on the bus.

This is false. That isn't they way SCSI works (unless you have
a very poorly written SCSI driver or braindead SCSI chip, of
course).

| >2. "Big problem mixing fast SCSI and slow SCSI, you should put all your
| >slow SCSI devices on the SPARC 10 embedded controller and all the fast SCSI on
| >a separate differential SCSI controller board (you will have to buy diff SCSI
| >drives also - about $200 more per drive). Sun really messed up with the active
| >termination on the embedded controller. " - 3rd party drive supplier
|
| This is the correct solution. We are ordering two HP 735s, each with one
| slow SCSI and one fast wide SCSI. There are several advantages to this
| solution, the big and obvious one is fast drives on the fast bus will go
| fast. The differential bus will have significantly less errors and you will
| not have to pay as much attention to your total bus length. By putting
| your CD-ROM drive and tape drive(s) on the slow bus and your hard disks on
| the fast bus you will speed up software installation and backups (although the
| gain may be small depending on other factors). Another thing to be aware of
| is if you ever have a flakey device on one of the busses, by moving devices
| from one bus to another you will be able to isolate the problem

You may or may not have fewer bus problems, but it certainly will
not affect the speed with which either device operates, unless
of course you are using the full bus bandwidth, and the slower
devices are using too much of the bandwidth (because they remain
on the bus longer to transfer a given amount of data).

| >3. "No problem mixing fast SCSI and slow SCSI" - another 3rd party drive supplier
|
| We recently bought 3 fast SCSI drives and have them all on slow SCSI busses.
| They are all working well. Another thing to be aware of is that the raw
| data rate of reading from or writing to a drive is limited to the bits that
| pass under a single head. Most advertising will not mention this number.

Most *concientious* suppliers will mention this. There are, of
course, many retailers/distributors who don't care, or just want
to sell whatever they have.

--
Let no one tell me that silence gives consent,  |   Dave Olson
because whoever is silent dissents.             |   Silicon Graphics, Inc.
    Maria Isabel Barreno                        |   olson@sgi.com



This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:06:54 CDT