[Sidefx-houdini-list] houdini crashes...
Mark A. Bolstad
mbolstad at arl.army.mil
Tue Jan 10 13:21:09 EST 2006
I'm running a 3.4 GHz P4 Xeon EMT (x86_64) with a QuadroFX 1300 (PCI-
Express), RHEL4, driver version 81.78, h8.0.475
I ran your mplay test for an hour (643 iteration to be precise) with no
problems (I went to lunch, it's really an annoying test).
On Tue, 2006-01-10 at 11:02 -0500, Mark Alexander wrote:
> > The only common demoninator is that we're using non quadro cards - but is that really going to alter the stability that much?
> I wouldn't think so. I don't think Nvidia even creates separate drivers
> for Quadro and GEForce cards on Linux - at least choosing Graphics
> Driver / Quadro or GEForce / Linux IA32 gives me the same link on the
> Nvidia driver site. There are some hardware differences between the
> GEForce and the Quadro cards, but this doesn't seem to be the culprit
> (as my QuadroFX card also exhibits the problem).
> The information from users I have so far seems to indicate that a Linux
> / AMD / Nvidia combo may be part of the X freezing issue. So, does
> anyone out there experience this problem:
> - with a Pentium CPU? or,
> - with a PCI-Express card?
> I am currently testing the 8178 drivers with an AMD / Linux / QuadroFX
> AGP card by continually running:
> mplay Mandril.pic ; sleep 4 ; killall -2 mplay
> in a loop. From previous experience, it should lockup after 50 or so
> iterations (I've never actually counted). This doesn't use a fragment
> shader at all, but does use a 256x256 RGB texture to display the image.
> > Also, sesi could help by recommending a hardware setup that works(tm). (and isn't three years old).
> I will certainly try to recommend a hardware setup once I determine if
> it is indeed a hardware issue. If it is a driver issue, hopefully Nvidia
> will fix it or we can find a workaround (again). If it is a software
> issue, hopefully we can fix it :)
> For now, try Peter B's suggestion of disabling Fragment Shaders. I would
> have thought that using well-supported and optimized textured mapped
> quads instead of the old, poorly supported 2D pixel pipeline would have
> improved stability for mplay, but I suppose that would have made my work
> too easy.
> Sidefx-houdini-list mailing list
> Sidefx-houdini-list at sidefx.com
More information about the Sidefx-houdini-list