SGI Indigo² Troubleshooting Guide

A teal Indigo² and three purple Indigo² IMPACT machines (unicron1bot, CC BY 2.0)

Fault-finding on the SGI Indigo², from the troubleshooting chapters of SGI's Indigo2 Workstation Owner's Guide (007-9096-050) and Indigo2 IMPACT Workstation Owner's Guide (007-2849-004), with repair reports from the IRIX Network forums. SGI's procedures end at the replaceable part. Of those parts, the power supply is the one owners report failing most often today.

Power-on sequence

A healthy Indigo² powers on like this:

1. The LED on the front blinks green, then turns amber, and the power-on diagnostics start.
2. The diagnostics run for about five seconds, then the disks spin up, about 15 seconds for each disk.
3. The LED turns green when every test has passed.
4. IRIX boots, or the System Maintenance menu appears if the machine is set up that way.

If the LED lights but the monitor stays dark, connect an ASCII terminal at 9600 baud to serial port 1 to read the PROM's messages.

Indigo² power-on symptoms (SGI Table 7-1)

Symptom	System state	Cause given by SGI
No LED	No power to unit	Power supply failure (check the fan at the rear); check that the system disk LED flashes; possible CPU failure; backplane not seated
Solid amber LED	–	CPU board or graphics board failure
Flashing amber LED	No memory found	SIMM failure; read the error message for the faulty SIMM
No boot tune	No audio	Faulty speaker or CPU board; volume set too low or to zero
Solid green LED	System menu displayed on booting	Diagnostics successful
No LED on the system disk at any time	No system boot	Drive not fully in its slot, or drive failure
Diagnostic error messages, no login icons	Diagnostic failure, no graphics	CPU, graphics board or monitor failure

No LED

SGI's steps when the LED never lights and nothing appears on the screen:

1. Unplug the mains lead, remove the top cover and press down firmly on the top of the EISA backplane, the gold-coloured metal wall behind the expansion slots.
2. Check that the internal power connectors are fully on the system board.
3. Refit the cover and check every external cable, the outlet and the monitor's brightness, and make sure the power switch has been pressed.
4. If the symptoms remain, check the two fans, one at the front on the left and one inside the power supply. If neither runs, the power supply is faulty or overloaded; SGI notes that early supplies may not be adequate for newer configurations.

A machine that comes out of long storage with its fans running but no LED, and that cannot be switched off with the power button, was cured on two IMPACT machines by pressing the midplane and the CPU module firmly back into place.

Amber LED

Blinking amber

A blinking amber LED points at memory.

• No message: reseat every SIMM and power on again. If it still blinks with a blank screen, SGI's next step is to replace all four SIMMs of the bank.
• No usable memory found. Make sure you have a full bank (4 SIMMs): reseat the SIMMs; if the message returns, they are faulty.
• Check or replace: SIMM#: the number names the SIMM that failed. Reseat it, and replace it if the message returns.
• Memory is not usable. Check or replace all SIMMs.: check that the SIMMs are seated and in the right sockets, then replace them all.

Solid amber

The memory has passed but another part has not.

• A dissonant piano chord from the speaker with no message means the graphics board failed its test. Reseat it, and replace it if the fault returns.
• No chord and no message: check the monitor and its brightness and that the system disk is fully in its slot, then suspect the graphics board.
• Check or replace: Graphics board., Check or replace: CPU module or Check or replace: CPU baseboard: reseat the part named, and replace it if the message returns.
• Lines or "measles" on the screen mean a faulty graphics board.

Power supply

Owners report these failures:

• The fans twitch for a second when the power button is pressed and nothing else happens. In one report a replacement supply brought the machine back.
• The machine runs for 30 to 60 minutes and then trips out.
• An IMPACT machine chimes, clicks and restarts in a loop. The owner measured every rail within specification and saw nothing wrong on an oscilloscope, but the system boards ran normally on a non-IMPACT supply, which placed the fault in the IMPACT supply. Recapping its low-voltage side, where only the red 330 µF 6.3 V capacitors were out of specification, did not cure it.
• The machine switches off whenever the room passes about 80 °F (27 °C) and will only start again once it has cooled and been unplugged, which forum members put down to a thermal cut-out.
• A machine that starts, runs its boot checks and then locks up so that the power button no longer works. weblacky's view is that unexplained freezes, once memory is ruled out, usually come from a supply with too much ripple or a rail at the edge of tolerance.

Testing a supply

weblacky's method is to power a suspect supply through an incandescent bulb as a current limiter, using a 300 W bulb or two 150 W bulbs so that the supply can start. A bulb that throbs or blinks while the supply appears dead shows that the supply is restarting over and over in overcurrent protection. On one supply he rebuilt, that was caused by two failed diodes on the low-voltage side. Do not try to run the system boards from bench supplies instead: the rails have to come up in the right order, and an IC that loses one of its supplies can latch up and destroy itself.

A supply that has stopped working usually has a blown semiconductor as well as aged capacitors, so a recap alone often does not revive it; see SGI Indigo² Capacitor Replacement Guide.

Audio board

An Indigo² gave a pop and a burning smell a few seconds after switch-on. The cause was a shorted 47 µF 20 V tantalum capacitor, C57, on the audio board, which had also burnt out inductor L517 and one of its pads. The machine ran normally with the audio board removed. The owner found a report of the same capacitor failing on an R8000 Indigo² that had not been switched on for months.

Clock and NVRAM

The Dallas DS1286 clock has its battery sealed inside. With a flat battery the Indigo² still boots, though IRIX may complain about an impossible date. See General Maintenance for fitting a replaceable cell.

Diagnostics

IDE diagnostics

The diagnostics take about 30 minutes, plus 15 to 30 minutes for each extra disk, and Esc stops them.

1. Shut the system down and choose Restart. If the system cannot be reached with the mouse, keyboard or network, press the Reset button behind the front door.
2. At the startup notifier click Stop for Maintenance or press Esc. If the System Maintenance menu cannot be reached, SGI's guide says the system is faulty.
3. Click Diagnostics or type 3. A banner such as SGI Version 5.3 IP22 IDE field appears; if it does not, the disk may be faulty.
4. The LED blinks slowly while the tests run, and the screen goes blank and shows test images during the graphics tests.

The confidence tests in the System toolchest check the audio, the dial and button box, tablet, genlock and Spaceball, CD-ROM, floppy drive, keyboard, monitor, mouse and tape drives.

Keyboard and mouse

With a green LED but a dead keyboard or mouse, check that each is plugged into its own connector and not the other's, then power on again; if it still fails, SGI's guide treats the keyboard or mouse as faulty.

Forgotten PROM password

If IRIX still runs, log in as root and type /etc/nvram passwd_key "". Otherwise shut down, remove the cover and the 5.25-inch drive tray, and pull the jumper at the front centre edge of the CPU baseboard.

Related pages

• SGI Indigo²
• SGI Indigo² General Maintenance
• SGI Indigo² Capacitor Replacement Guide
• SGI Indigo² Documentation
• SGI Power Indigo²
• Capacitor Failure Symptoms