Subject: Re: ATAPI UltraDMA questions...
To: Jukka Marin <>
From: Wolfgang Rupprecht <>
List: netbsd-users
Date: 02/28/2001 11:34:36
Jukka Marin writes:
> On Wed, Feb 28, 2001 at 10:45:13AM -0800, Wolfgang Rupprecht wrote:
> > I think the raw performance of IDE is quite good (especially if you
> > only put one disk on a cable). The IBM DTLA-307075 (an IDE drive)
> > reads at 36.4 MBytes/sec.
> I get over 37 MB/s on my IBM's.

This is confirmation enough. ;-) My numbers were from a live system,
so I was expecting a slightly conservative number.

What I find most impressive is the honestly on IBM's web site.  Quite
often when vendors give a number they are lying through their teeth.
The drive's spec sheet at claims
37MBytes/sec sustained data rate and this appears to be 100% true for
the outer tracks.

> > The one caveat is that all current drives are zoned, meaning that they
> > are faster on the outer (low-numbered) tracks than the inner ones.
> > The above IBM slows down by a factor of ~2 on the inner tracks.
> Isn't this true for SCSI disks as well?

Sure is.  I wasn't trying to imply otherwise.  For a while most
manufacturers were selling the exact same drive with two different
controller boards.  In effect the question they asked at the checkout
was "would you like SCSI with that?".

(Moving folks to SCSI was simply a way for the manufacturer to get a
more reasonable mark-up in the highly cut-throat disk business.)

> I have two IDE disks on the same cable and they read/write over 22 MB/s
> simultaneously.  I get over 20 MB/s through RAID1 from these two disks.

Interesting.  It sure seems like something is maxing out around

> I have never had an SCSI device that could beat this ;-)  (Yeah, they
> do exist, but they are expensive and need an UW controller..)

Neither have I.  Its only been recently that the manufacturers have
only been releasing the good drives in SCSI and waiting a while before
releasing them in IDE.

Wolfgang Rupprecht    <>
Coming soon: GPS mapping tools for Open Systems.