NetNews Usenet Archive 1992 #31/NN_1992_31.iso/spool/biz/sco/general/4909

Internet Message Format | 1992-12-31 | 3.6 KB

Path: sparky!uunet!shady!kevin
From: [e-mail] (Kevin Smith)
Newsgroups: biz.sco.general
Subject: Re: Configuring an Adaptec 1542B ...
Message-ID: <[e-mail]>
Date: 31 Dec 92 20:25:35 GMT
References: <[e-mail]>
Organization: ShadeTree Software, Inc.
Lines: 66

In article <[e-mail]> jrsalb1![e-mail] (Tim Northrup) writes:
:>Howdy & Happy New Year!
:>
:>I have a 486/50 system running Xenix 2.3.3.
:>
:>It has an Adaptec 1542b controller with 2 Maxtor SCSI drives, and a SCSI
:>Syquest 44 (all internal, ID's 0, 1 and 2, terminators on the Syquest drive
:>& adapter only).  Everything works okay at the moment.
:>
:>Currently, the 1542b is configured for the default of 5 mb/s transfer speed.
:>Has anyone had any problems using the higher settings?  A friend (who set
:>up this system) claims to have had some kind of disk corruption occur when
:>the higher speeds were used.  What else needs to be changed (if any) to
:>use the higher settings?
:>
:>Any info at all on configuration of this puppy would be most appreciated!
:>
:>I got zilch for documentation on the 1542b with the system; the document
:>154XB_IG.DOC from the Adaptec BBS is a little helpful; but I'm looking for
:>a little better direction.
:>
:>What prompted this was the recent performance discussion going on ... I ran
:>one of the tests (dd if=/dev/rhd0a of=/dev/null bs=8k count=2048) and it
:>took about a minute on my system as currently configured ... this seemed a
:>bit long to me.  Is this slow, or does it just *seem* slow?
:>
:>ANYWAY .... Thanks in advance for any pointers!
:>

Either here or in comp.unix.sysv386 I have heard of people having
problems at higher transfer rates.  I have a 386/33 AMI motherboard
and AMI bios with a 1542b controller.  I was able to run with no
problems at the highest transfer rate setting (don't remember what it
was) under UNIX but DOS would throw a fit.  Since I also did not
detect any significant performance improvement at higher speeds, I
left it at the default.

What I have noticed, however, is that disk drives without a caching
controller will perform horribly during boot and in raw disk transfers
(as in your test).  I don't have any hard numbers but during boot,
with a non caching seagate drive, it would take about 30 seconds to
get through all the dots whereas with a cached seagate (or my HP
drives) it would take only 2-3 seconds.  On my system (multi-user,
light load, no special priorities), your 'dd' test took 38 seconds on
one drive and 30 on the other (both HP drives, 660MB and 1.2GB).

Block io (i.e.  files) suffered but not to as great an extent.  My HP
drives have a one track cache as do a lot of the drives out there and
it makes all the difference in the world.  I would have to say, don't
leave home without one.  Note:  these drives did not have one of the
fancy buffered intelligent controllers, just a track buffer.

You might also try the sychronous transfer enable (J5-1).  This allows
the adaptor to negotiate sychronous mode with each drive.  If the
drives support sychronous mode this will significanly boost the data
transfer rate over the SCSI bus.  The adaptor will negotiate transfer
mode with each device independantly.  This jumper allows the adaptor
to use sychronous mode if a device will agree to it.

All in all, transfer rates (DMA, motherboard bus and SCSI bus) still tend
to stay way ahead of seek times and rotational latency.
--
        | Email - !shady!kevin uunet!shady!kevin [e-mail]
Kevin Smith | Voice - (+1) [phone]
        | Mail  - ShadeTree Software, Inc., 192 Capricorn Dr. #10,
        |         Somerville, NJ  08876, USA