SGI: Discussion

vishnu wrote: How dusty was it inside the poor thing? Octanes are notorious dust collectors...

Not too bad - a fluffball stuck above one of the top grills, but most of the honeycomb looked clear - and the system module seemed pretty clean as well.

Only one R10K showing up tonight and the MXI has disappeared from hinv

How do the compression connectors look? Warped or oxidized? Any evidence of overheating, charred PC boards in the system unit or power supply? Non-Vpro graphics are pretty cheap these days when they show up on Ebay. Or you could pay the premium price for a brand new one at Ian's depot...

vishnu wrote: How do the compression connectors look? Warped or oxidized? Any evidence of overheating, charred PC boards in the system unit or power supply? Non-Vpro graphics are pretty cheap these days when they show up on Ebay. Or you could pay the premium price for a brand new one at Ian's depot...

Bit of a hiatus while I took over the study as a computer room, put up some shelves and added a second desk - more elbow room for tinkering, and s second monitor!

Compression connectors looked OK, though I've only photos for comparison never having seen one "live" before.

Current status:
graphics - after a bit of research on running dual head, I added a "+xinerama" option to the Xservers file and got a two-screen desktop. Unfortunately the LH half shows up on the monitor to the right, but that's just a matter of cabling (or playing with "-hw board= ..." options). So those CCs are OK!

CPUs - PROM hinv now shows two, but IRIX hinv shows one. Need to run the diagnostics again, but maybe it is a software/kernel/config issue? This machine once had a CADduo card in a cardcage, I am sure - there is a "Xservers.2key" file which would have configured the two heads for two kbd/mouse inputs (/dev/input and /dev/input1). Since CADduo was intended for dual-CPU machines, I am wondering if there is some legacy configuration that I need to clean out in the absence of the CADduo card to reinstate normal "single-seat" operation, or perhaps I should create a new kernel file (though that might not help if IRIX can only see one CPU - is there a "force 2-cpus" switch?). This may be a red herring - AIUI CADduo did not dedicate one CPU to each user - so other ideas welcome. Does PROM hinv get system info in a different way and cannot spot that a CPU is "dud"? Does IRIX set a lower fault threshold?

I think the first thing to do would be to check the installed software and remove any CADduo driver(s) you see there...

vishnu wrote: I think the first thing to do would be to check the installed software and remove any CADduo driver(s) you see there...

Even before that, I'd do a < printenv > from the prom, write down anything interesting, then do < enableall > and < resetenv >. That way you're starting out with a clean foundation.

hamei wrote:

vishnu wrote: I think the first thing to do would be to check the installed software and remove any CADduo driver(s) you see there...

Even before that, I'd do a < printenv > from the prom, write down anything interesting, then do < enableall > and < resetenv >. That way you're starting out with a clean foundation.

PROM env was vanilla aside from "netaddr". "enableall" was rejected (IP35 only?) but "enable 1" told me that CPU1 had been marked as "enabled" but in fact "disabled" (not apparent from PROM hinv). So I ran PROM diagnostics, which reported nothing. May need to try again with a serial console attached, then remove/reseat the PM.

In IRIX, nothing calling itself "CADduo" (or similar) in swmgr's list of installed, and nothing obvious in /var/sysgen. Given the PROM "enable" status, I think it's hardware after all. The DUO notes in Techpubs indicate that switching from DUO to dual head is simply a matter of changing the /var/X11/xdm/Xservers file - no mention of an additional software package (though the Dual Head manual does refer to a CD - but not what was on it!).