Path: utzoo!mnetor!uunet!husc6!mailrus!ames!think!ephraim
From: ephraim@think.COM (ephraim vishniac)
Newsgroups: comp.sys.mac
Subject: Re: Jasmine Direct Drive 50 and a Plus
Message-ID: <18879@think.UUCP>
Date: 5 Apr 88 20:56:01 GMT
References: <23013@bbn.COM>
Sender: usenet@think.UUCP
Reply-To: ephraim@vidar.think.com.UUCP (ephraim vishniac)
Organization: Thinking Machines Corporation, Cambridge, MA
Lines: 50

In article <23013@bbn.COM> levin@BBN.COM (Joel B Levin) writes:
>I just got a Jasmine Direct Drive 50 for the office Plus, and I wanted
>to pass along a small gotcha that you might keep in mind if you are
>looking at hard disks.
...
>I was surprised to find the performance numbers substantially
>lower than I had been led to believe they would be:

>  Read   117
>  Write  126
>  Seek     0 (no seeking due to the large number of sectors per
>              cylinder, I believe).

>I called Jasmine Tech Support and discovered that the drive is
>formated with 1:1 interleave; that it comes from the factory that way;
>and that the interleave cannot be changed by software, according to
>the person who answered my call (and consulted with someone while I
>was on the phone to verify that this was normal performance).

This report is rather misleading.  There is a good reason that the
drive's interleaving is fixed at 1:1, and it's unrelated to the
drive's performance on a Plus.

First, the interleaving: The Quantum drives have a large internal
buffer/cache which allows them to cache a track in one pass.  So, 1:1
is the best interleaving for the drive without regard to the host's
transfer rate.

Second, the performance: The Quantum drives have rather complicated
controller software, partially due to their caching scheme.  So, the
externally visible timing of the controller is more variable than
some.  This causes it to trip over various bugs in the SCSI manager in
the Mac Plus, and also makes blind transfers impossible on the Plus'
very limited SCSI hardware.  (The SE and II have slightly better
hardware.)  Consequently, the Quantum drives must use fully
synchronized transfers at a cost of about three instructions per byte
instead of one instruction per byte for blind transfers.

How do I know this stuff?  I wrote Jasmine's SCSI driver.

Disclaimer: I once had a business relationship with Jasmine
(obviously), but now they own the software outright and it's their
problem.


Ephraim Vishniac					  ephraim@think.com
Thinking Machines Corporation / 245 First Street / Cambridge, MA 02142-1214

     On two occasions I have been asked, "Pray, Mr. Babbage, if you put
     into the machine wrong figures, will the right answers come out?"