This Old SGI wrote:The Mysterious Cursor Hardware Hack
A number of early posts refer to a hardware hack that was apparently popular within SGI itself, and which made it possible to have a two-color hardware-supported cursor on the PI. I am not currently aware of what added value this modification actually has, but I'll include it here for anyone who wishes to experiment. The gist of the modification is the addition of a Brooktree chip (Bt431KC) to the GR1.x Graphics board. If you remove the GR1.x board, you will notice an empty IC location (I8G6) next to the a Bt431KC which is already on the board. You simply install a second Bt431KC in this location. Note that the IC pads at this location will be filled with solder, so you'll have to CAREFULLY remove the excess solder from each pad hole. If I were you, I'd strongly consider using an IC socket, if the board clearance allows it. I have NOT tried this modification myself, nor have I been able to locate a source for spare Bt431KC chips. Apparently, IRIX versions 4 and above detected the extra hardware cursor support and could make use of it. I don't know if IRIX 5.x retained this support.
Recently I was able to obtain two Bt431KC chips and I decided to put this myth to the test
As you can see in the last screenshot, 'gfxinfo' recognizes it, and I actually have a dual color cursor! This is IRIX 4.0.5 running on an IP10 (4D/25).
Even though the Bt431KC is in a socket, there is enough clearance to install the Z buffer board again.
As the Mythbusters would say: Myth confirmed
EDIT: the paragraph in "This Old SGI" is probably based on this message: http://groups.google.com/group/comp.sys ... e919cd89be
Good thing I installed the second Bt431KC in a socket, I don't want to loose the option of running 3.3 on this box.