[3701] in Release_7.7_team

home help back first fref pref prev next nref lref last post

Our issues with Matrox G450DVI and Linux

daemon@ATHENA.MIT.EDU (Bill Cattey)
Thu Jan 16 17:31:28 2003

From: Bill Cattey <wdc@MIT.EDU>
To: Pam Huntley <phuntley@us.ibm.com>
Cc: Amy Alsheskie <amyalshe@us.ibm.com>, Steven Sakata <ssakata@us.ibm.com>,
        tjim@us.ibm.com, release-team@mit.edu
Content-Type: text/plain
Content-Transfer-Encoding: 7bit
Date: 16 Jan 2003 17:31:20 -0500
Message-Id: <1042756280.10070.62.camel@tokata.mit.edu>
Mime-Version: 1.0

Pam:

I hope I'm right in addressing this note to you, primarily.

Amy and Stephen are interested in this issue because they made a site
visit when it looked like we were getting dead Matrox G450 cards on our
big Intellistation order.

Since the first batch of Intellistations arrived we have been concerned
that there might be something wrong in how XFree86 auto-configures the
monitor sync rates under Red Hat Linux.

We at MIT understand that it's common policy to point users at the
proprietary Matrox device driver as a way of getting better performance
or functionality.

We did extensive testing with various combinations of monitor, drivers,
resolutions, and software versions.  Our detailed findings are published
at:

	http://web.mit.edu/release/www/ibm-video.html

We want IBM to understand a few simple findings of ours:

Under the current XFree86 v4.2, there is a problem with
auto-configuration at the 1280X1024 resolution.  This problem was missed
because testers apparently assumed all would be well at lower
resolutions than the proven working 1600x1200 resolution.

The proprietary Matrox driver solves the one problem of
auto-configuration at the 1280x1024 resolution for a cathode ray tube
monitor.  But it creates new, far worse problems when the monitor is the
IBM T860 flat panel monitor.  For example, whereas the free driver works
if you hand-tool the monitor spec, the Matrox proprietary driver will
hang the machine and require a power cycle when you try to use the IBM
LCD display (or even a 3rd party LCD display for that matter).

The IBM T860 LCD display apparently does not comply with the VESA spec
to report its supported sync rates.  We find the third party Planar
brand monitors do so.  We find the Planar displays superior to the IBM
product at much lower cost.

With LCD's the digital DVI interface simply does not work under Linux.
Perhaps this is a problem that Matrox can address with their G450 card.

The new XFree86 v4.3, which we tested in beta, remedies all the problems
we had at the 1280x1024 resolution, creates no new problems, and does
not require acquisition and integration of a proprietary driver off the
net.

----

Our findings lead us to suggest the following IBM policies moving
forward:

1. Quit advocating the Matrox proprietary driver.  It causes more
problems than it solves.

2. Work with Matrox to find out why DVI does not work.

3. Recognize and perhaps remedy the current state of the T860 LCD of:
	Not properly identifying its valid scan rates in analog mode.
	Not being as functional and cost effective as units from Planar.

At this time MIT is well positioned to find its own way without any
intervention by IBM in these areas.  (All we have to do is not buy IBM
LCD's and to wait for XFree86 v4.3 to arrive.)

I am sending you this note because, as a business partner here, I see
value to IBM if the above 3 policies are implemented.  IBM will have
more satisfied customers as a consequence of internalizing what MIT has
learned.

I hope this information proves useful to you,

-wdc



home help back first fref pref prev next nref lref last post