X performance on mini-itx boards

[Lee Revell]

On Thu, 2004-09-02 at 04:30, Francois Tigeot wrote:
> On Thu, Sep 02, 2004 at 03:38:57AM -0400, Lee Revell wrote:
> > On Thu, 2004-09-02 at 03:26, Francois Tigeot wrote:
> > > On Tue, Aug 31, 2004 at 09:43:16PM +0100, Hamie wrote:
> > > > What sort of performance on the VIA mini-itx boards can be expected with 
> > > > XOrg? Is there hardware acceleration available for these boards? WHat 
> > > > sort of fps can you get with glxgears at default size?
> > > 
> > > I have found fanless VIA ME-6000 perfectly adequate for 2D work.
> > 
> > The current 2D driver has some recently identified problems, where
> > excessive 2D engine activity (dragging around the mplayer window) would
> > completely stall the PCI bus, with the symptoms that interrupt handlers
> > can be delayed for arbitrary lengths of time, leading to audio (and
> > probably other) problems.
> 
> [...]
> 
> I have yet to see any of these.
> The small number of machines already deployed are only used for word
> processing, web browsing and typical office work, which may help.

Exactly, I am sure VIA thinks they can get away with it for this
reason.  It's not the kind of problem that would show up with normal
desktop use, but becomes glaringly obvious once you try to do low
latency audio.  The behavior is also almost certainly intentional, see
this article (scroll down to 4.5 Problem: Antisocial Video Cards):

http://research.microsoft.com/~mbj/papers/tr-98-29.html

The problem is trivial to demonstrate.  Once you start a PCM stream, the
sound card starts to DMA, and the capture interrupts should come at a
fixed rate, corresponding to the period size, and it can be trivially
demonstrated that overloading the 2D engine by dragging windows will
cause long gaps in the stream of interrupts.

Doesn't really matter anymore, this is supposed to be fixed in the
latest Unichrome driver (tagged release-25rc).  Please see
unichrome-devel for more info.

Lee