66MHz 3.3V PCI Graphics cards?

Robert G. Brown rgb at phy.duke.edu
Tue Jun 18 13:43:40 PDT 2002

On Tue, 18 Jun 2002, Nicholas Brealey wrote:

> Hello
> I have only built one machine so far (the master in a 4U case).
> It worked fine except that it would hang before you could
> get to the setup menu with a SiS PCI graphics card.
> An old Matrox Millenium G200 AGP graphics did not display
> anything at all. (I should double check I might have done
> something stupid like not plug in the monitor - it was rather
> late when I tried it).
> A new NVidia graphics card worked OK.
> I have upgraded the motherboard BIOS to the latest beta 4.01u
> but is still hangs when using the SiS PCI card.

I wish I could help beyond what I already said, but as you've discovered
there are cards that work and cards that don't, and you'll just have to
experiment to find out which is which. I don't know if it has anything
to do with the BIOS setting -- there is a place where you can set AGP vs
PCI for the video console, but it worked for our PCI cards either way if
it worked for them at all.

Another thing to let you know about is that they now make a 2x64 bit,
1xAGP riser that has an AGP ribbon to extend the AGP bus to a horizotal
riser slot.  With it one can in principle run an AGP video card and two
64/66 PCI cards.  We're likely going to test this riser on our next
cluster purchase, but on this one I didn't get any video cards at all.
I just use the one Matrox (PCI) card that works to set up the serial
console, and thereafter use only the serial console.  Saves a bit of
money, a bit of power, and a bit of heat NOT to have a video card in a


> Regards
> Nick
> "Robert G. Brown" wrote:
> > 
> > On Mon, 17 Jun 2002, Nicholas Brealey wrote:
> > 
> > > Hello
> > >
> > > Does anyone know of any 66MHz, 3.3V, 32 or 64 bit PCI graphics cards.
> > > Or even a 33MHz, 3.3V, 32 or 64 bit PCI graphics card?
> > >
> > > A have just got two test Tyan Tiger S2466N-4M motherboards which
> > > I want to be able to put into 2U rackmount cases. They would not
> > > get to the BIOS setup with a 5V 32bit 33 MHz PCI graphics card
> > > in a 33MHz slot. They worked fine with an NVIDIA based AGP
> > > graphics card. I am looking for a card I can use to get into
> > > BIOS setup so that I can enable console redirect so that I
> > > don't need a graphics card. I am intending to build a cluster
> > > with 16 boards in 2U cases.
> > 
> > This cluster is "just like" the one I'm building.  What we did is take
> > the case top off, put a card like an ATI Rage or a Matrox Millenium in a
> > 32 bit slot, and boot it up into the BIOS.  "Most" PCI graphics cards
> > (and there are still a number out there) will work on this motherboard,
> > but really old ones (I tried one of my manifold 4 MB S3's -- maybe a
> > Trio64) do not work.
> > 
> > Once in the BIOS, remember to set the keyboard to ignore error on boot
> > as well as set your comm parameters in the advanced/console section.
> > Then minicom suffices to control the system, with luck.  The one
> > "problem" one faces is that one cannot do a three-finger salute-boot (it
> > boots the wrong system:-) -- you have to hard boot the case any time you
> > want to start over.
> > 
> > We have had the best luck with the console in vt100 mode and minicom in
> > ANSI mode, both 115200 N81.
> > 
> > We have had a very indifferent experience thus far with these systems.
> > Of the maybe twelve I've tried to bring up and install so far, three
> > blew power supplies on a powerup, one isn't working and I don't know why
> > yet, and two boot up into the BIOS but cannot seem to find the PXE chip
> > (and yes, I've checked the jumpers) -- one BIOS indicates two CDROM's in
> > its boot list, the other two removable device entries, where neither
> > system has either CDROM or removable device.
> > 
> > Sure, these are diverse causes -- maybe -- but it is a lot of "bad luck"
> > if nothing else, and sometimes bad luck like this indicates a marginal
> > design, and combined with our experiences with 2460 systems I'd have to
> > say that this is a very marginal design.  Be prepared to "mess" with it
> > to get it installed and working.  The (7) nodes that have made it
> > through the install phase and a few days of burnin seem stable and are
> > working away nicely, though.  Inshallah I will have the other 9 going in
> > a day or five.
> > 
> >    rgb
> > 
> > >
> > > Nick
> > > _______________________________________________
> > > Beowulf mailing list, Beowulf at beowulf.org
> > > To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
> > >
> > 
> > Robert G. Brown                        http://www.phy.duke.edu/~rgb/
> > Duke University Dept. of Physics, Box 90305
> > Durham, N.C. 27708-0305
> > Phone: 1-919-660-2567  Fax: 919-660-2525     email:rgb at phy.duke.edu

Robert G. Brown	                       http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567  Fax: 919-660-2525     email:rgb at phy.duke.edu

More information about the Beowulf mailing list