SGI Indy Troubleshooting Guide

Fault-finding on the SGI Indy, from the troubleshooting chapter of SGI's Indy Workstation Owner's Guide (007-9804-060) and from repairs reported by owners on the IRIX Network forums, Andrej Bukošek's collection notes and the TechPubs Wiki.[1][2][3] SGI's procedures stop at the replaceable part: power supply, system board, CPU module, graphics board, SIMMs. Below that level the notes come from community repairs. The two faults most often met on an Indy today are a failing power supply and a flat battery in the Dallas timekeeper module.[4]
Power-on sequence
[edit | edit source]A healthy Indy powers on like this:[5]
- The LED on the front lights amber, and the boot tune plays.
- The power-on diagnostics run for about five seconds with the LED amber, then the disk spins up, which takes about 15 seconds.
- The LED turns green when every power-on test has passed.
- IRIX boots, or the System Maintenance menu appears if the machine is set up that way.
SGI's summary of what the LED means when it does not turn green:[6]
| Indicator | Symptom | Cause given by SGI |
|---|---|---|
| No LED | No power to system | Power supply failure; check for fan airflow at the rear and that the power cable is connected |
| No boot tune | No audio | Faulty speaker, system board or CPU module; volume set low or to zero |
| Blinking red LED | Memory diagnostic failure | SIMM failure; read the error message for the faulty SIMM |
| Solid red LED | A part of the workstation is faulty | Graphics board, system board or CPU module |
Messages with a red LED
[edit | edit source]SGI's first step for each of these is to switch off with the power button, switch on again, and only then act on the message if it returns.[7]
- Blinking red, no message, or
No usable memory found. Make sure you have a full bank (4SIMMs): check that every SIMM is pushed fully home and that each bank holds four SIMMs of the same size. - Blinking red with
Check or replace: SIMM S#: the message names the SIMM socket, S1 to S8, that failed. Reseat or replace that SIMM. - Solid red with
Check or replace: Graphics boardorCheck or replace: CPU module: run the IDE diagnostics, then replace the part named. - Solid red and no message: one of the parts has failed and SGI's guide goes no further.
With no picture at all, connect a serial terminal and watch the PROM's messages there. dexter1 suggests then stripping the machine to one bank of four SIMMs, with no graphics board and no disks.[8]
Power supply
[edit | edit source]No LED and no fan
[edit | edit source]SGI's procedure: check the mains lead and outlet, press the power button again, and listen. If the boot tune plays and the disk spins up, feel for airflow from the fan at the back of the power supply; if it is running, the LED itself may be broken. If there is no airflow and no fan noise, the power supply has probably failed.[9] A forum reply on a machine that played the boot tune but never showed a green LED suggests the LED had failed; it is inside the power supply and can be tested with a meter on diode test once the supply is opened.[10]
Not a fault: the Sony supply's fan does not start until the system board's temperature sensor reaches about 45 °C, so a cool Sony-equipped Indy can run with the fan stopped. The Nidec fan always turns.[3]
Telling the supply from the system board
[edit | edit source]The supply starts whenever mains is applied unless the system board holds the run line (pin 3 of the 20-pin auxiliary header) at ground, so a supply disconnected from the board runs by itself.[3] weblacky's method for separating a supply fault from a board fault:[11]
- Unplug both the main power connector and the smaller auxiliary cable from the system board. Plug the supply into a switched outlet and switch on. If the fan starts, the supply believes it can run. Do not leave it running unloaded for long.
- Watch the main connector voltages with a meter. In protection mode the Nidec supply looks as if it is running, but the voltages rise and fall rhythmically as it restarts over and over; weblacky reads this as a damaged or shorted recovery diode on the secondary side.
- If the supply runs on its own but not when connected, the system board is holding it off. With the power off, check the board's power connector pins to ground on the diode range for an obvious short.
- Only once both are proven, try the supply on the main connector with the auxiliary cable left off. This bypasses the board's control of the supply, so nothing will shut it down if there is a fault.
A supply that starts the moment it is plugged in, before the power button is pressed, also points at the system board, since only the board stops the supply from starting.[12][3]
Supply rails
[edit | edit source]| Rail | Rating | Main connector pins | Wire colour |
|---|---|---|---|
| +3.3 V | 7 A | 1, 2 | White |
| +5 V | 25 A | 3 to 8 | Red |
| +5 V standby | 1 mA on the Nidec label; 0.02 A on the TechPubs Wiki | 9 | Green |
| +12 V | 4.5 A | 17 | Yellow |
| −12 V | 0.75 A | 10 | Light blue |
| 0 V | – | 11 to 16, 18 to 20 | Black |
The 20-pin auxiliary header carries the control lines: pin 2 is the fan-control voltage from the board's temperature sensor (0 to 5 V), pin 3 is run (ground to stop), pin 4 is power good, and the remaining used pins go to the front-panel buttons, the red and green LED and the speaker.[3]
Shuts down after a few minutes
[edit | edit source]An Indy that ran for two or three minutes and then switched itself off, and later would not start at all, was diagnosed on the forum as a failing supply. weblacky's view is that the supply or the board detects a rail out of tolerance, usually because the supply's output sags or its ripple rises as it warms, and that this almost always comes from worn filter capacitors.[14]
Dead supplies that have been repaired
[edit | edit source]- Two Indys with dead supplies came back after a recap and a new R5, which read open in both.[15]
- A Nidec supply that flickered the red LED and clicked was dead; a Sony supply fitted in its place booted the machine.[16]
- weblacky finds that a supply which has stopped working nearly always has a damaged semiconductor as well as worn capacitors, and that the load resistors which run hot are worth replacing.[17]
See SGI Indy Capacitor Replacement Guide for the capacitors.
Clock, NVRAM and Ethernet address
[edit | edit source]The Dallas DS1386-8K-150 timekeeper holds the time, the PROM environment and the Ethernet address, with its battery sealed inside.[2][4] Symptoms of a flat battery:
- The date and time are lost whenever the machine is unplugged; one module weblacky tested reset itself to the year 2052.[18]
- The PROM reports a bad Ethernet address, or the address reads ff:ff:ff:ff:ff:ff, and the machine stops at the PROM monitor.[4][19][2]
- An owner whose Indy refused to boot confirmed from the serial console that the timekeeper was at fault.[20]
Replacing the module and restoring the address
[edit | edit source]Fit a DS1386-8K (the -150 original or the -120), or rebuild the old module with an external CR2032 holder.[2][18] Then, following Bukošek:[2]
- Press Esc at the startup notifier, then 5, to reach the Command Monitor.
- Type
resetenvto write a clean PROM environment into the new module. - Type
setenv OSLoadFilename /unix, whichresetenvdoes not always restore. - Read the Ethernet address from the label on the rear panel. It always begins 08:00:69.
Then write the six bytes of the address, one per command. For the address 08:00:69:c0:ff:ee:
fill -w -v 0x08 0xbfbe04e8 fill -w -v 0x00 0xbfbe04ec fill -w -v 0x69 0xbfbe04f0 fill -w -v 0xc0 0xbfbe04f4 fill -w -v 0xff 0xbfbe04f8 fill -w -v 0xee 0xbfbe04fc
Check the values with dump -w -x 0xbfbe04e8, set the clock with date, and type init to restart.
setenv -f eaddr 08:00:69:xx:xx:xx worked for one owner, but on the latest PROM the variable is read-only and the PROM refuses to change it.[19][2][21] The address written with fill only appears after the machine has been switched off and on again.[22]
A DS1386-32K is not a drop-in part. Its address lines A13 and A14 must be held at ground, but pin 3 of the Indy's socket carries the always-on 5 V auxiliary rail, so Adrian Chadd raised the module on pin headers with no pins at positions 3 and 28, and wired those two module pins to pin 16, ground, on its underside.[23]
System board faults
[edit | edit source]Shorted tantalum capacitors
[edit | edit source]The Indy system board uses tantalum capacitors, which fail short and can burn.[24][8]
- C602, on the underside of the board, burnt out on an Indy R5000 that lit both LEDs for a second and then switched off with a foul smell. Its value is 4.7 µF 20 V.[25]
- dexter1 reports that C16, in the row C15 to C17, sometimes blows, and has seen tantalum capacitors fail near the IndyCam connector on an Indy that then would not boot.[16][26]
- On the Challenge S, the server version of the board, a 10 µF 25 V tantalum capacitor between the GIO sockets blew while the machine kept running.[26]
To look for a short, remove the CPU module, SIMMs and boards and measure the resistance across each yellow tantalum capacitor. A very low reading suggests a short, though capacitors in parallel give the same reading, so compare with a good board where possible.[8] weblacky has seen tantalum failures on boards that had been run from a poorly regulated supply.[14]
Path test failures on the serial console
[edit | edit source]One Indy with a green LED, no boot tune and no picture reported this on the serial console:
Data path test *FAILED*
WD SCSI0 path test *FAILED*
RTC path test *FAILED*
weblacky's reading was that the parts named were unpowered, pointing to a short on the 5 V supply after the point where the timekeeper's supply joins the standby rail. He suggested checking for 5 V at the timekeeper socket, then at the SCSI controller, and looking for a warm part.[8]
CPU module
[edit | edit source]An R5000 Indy that passed its diagnostics but locked up while loading the IRIX installer worked normally with the CPU module from another Indy, which pointed at the module.[15] Before blaming a module, check that the PROM is new enough for it: R5000 modules need a PROM dated 12 February 1996 or later, and IRIX 5.3 supports the R5000 only in the edition titled for it.[2][15]
Diagnostics
[edit | edit source]IDE diagnostics
[edit | edit source]The Interactive Diagnostic Environment tests the hardware more thoroughly than the power-on tests.[27]
- Power off, then on. At the startup notifier click Stop for Maintenance, or press Esc.
- Choose Run Diagnostics, or type 3, at the System Maintenance menu. If the menu cannot be reached, SGI's guide says the system is faulty; if the IDE banner does not appear, the disk may be faulty.
- The tests take 30 minutes to an hour. The LED blinks slowly while they run, and the screen can stay blank for 10 minutes or more during the graphics tests. Esc stops them.
- A pass ends with
Diagnostic tests completed with no failures detected.A failure givesERROR: Failure detected on thefollowed by the part, circuit or subsystem.
The confidence tests in the System toolchest check the audio, keyboard, mouse, monitor, floppy or floptical drive, CD-ROM, tape drives and Indy Presenter.[27]
Forgotten PROM password
[edit | edit source]
If the System Maintenance menu asks for a password nobody knows, shut down, unplug, remove the cover and pull the jumper beside the power supply cable connectors on the system board. The PROM then shows "Password jumper has been removed. Not enforcing PROM password." To set a password again later, clear the old one with resetpw in the Command Monitor and refit the jumper.[28]
Peripherals and drives
[edit | edit source]- Internal floptical or second disk not seen: with nothing on the external SCSI port, the active terminator must be fitted there; reseat the ribbon and power cables; a floptical drive must be at SCSI ID 2.[29]
- External SCSI devices: switch them on before the Indy, do not use IDs 1 or 2, keep the total external cable length to 2 metres, connect no more than five devices, and terminate the last one with an active terminator.[29][30]
- No picture on a modern monitor: the Indy's 13W3 output uses sync on green, which not every monitor accepts.[4]
- Erratic mouse: clean the rollers and ball; see General Maintenance.[31]
Related pages
[edit | edit source]- SGI Indy
- SGI Indy General Maintenance
- SGI Indy Capacitor Replacement Guide
- SGI Indy Documentation
- Capacitor Failure Symptoms
References
[edit | edit source]- ↑ Silicon Graphics, Indy Workstation Owner's Guide, 007-9804-060, 1996, chapter 7, pp. 103–129 (File:SGI Indy Workstation Owner's Guide (007-9804-060).pdf).
- ↑ 2.0 2.1 2.2 2.3 2.4 2.5 2.6 Andrej Bukošek, "SGI Indy", Hardware collection, bukosek.si (retrieved 21 October 2026). DS1386-8K-150, MAC address procedure, read-only
eaddron the latest PROM, PROM dates for CPU modules. - ↑ 3.0 3.1 3.2 3.3 3.4 3.5 "Indy Power Supply Information", TechPubs Wiki, based on measurements posted by SGUG forum member Elf in 2019, tech-pubs.net (revision of 27 September 2025). Rail ratings, pinouts, run signal, fan control.
- ↑ 4.0 4.1 4.2 4.3 "Indy", TechPubs Wiki, tech-pubs.net (retrieved 21 October 2026). Power supply as the part most prone to failure; Ethernet address in battery-backed RAM; sync-on-green graphics.
- ↑ SGI, 007-9804-060, p. 108. Power-on LED sequence and timing.
- ↑ 6.0 6.1 SGI, 007-9804-060, p. 109, Table 7-1, "Troubleshooting Symptoms and Probable Causes".
- ↑ SGI, 007-9804-060, pp. 110–112. Blinking and solid red LED, with and without messages.
- ↑ 8.0 8.1 8.2 8.3 "SGI Indy green light but not booting and no chime", IRIX Network forums, June to July 2021, forums.irixnet.org. Serial console path-test failures, stripping the machine down, testing tantalum capacitors for shorts.
- ↑ SGI, 007-9804-060, p. 109. No visible light from the LED with no error message.
- ↑ "Indy - boot tune but no green LED?", IRIX Network forums, 15 June 2025, forums.irixnet.org. Failed LED, tested in the opened supply.
- ↑ "dead indy", IRIX Network forums, weblacky, 9 July 2022, forums.irixnet.org. Bench test with the cables unplugged, protection-mode cycling, checking the board connector for shorts.
- ↑ "My Indy keeps date/time, but changes its MAC address (only the last 3 octets)?", IRIX Network forums, weblacky, 17 January 2022, forums.irixnet.org. Supply starting as soon as it is plugged in; the board is what holds it off.
- ↑ Andrej Bukošek, photograph of an Indy Nidec power supply label (File:SGI Indy Nidec power supply label (Andrej Bukosek).jpg). 5 V auxiliary output 1 mA.
- ↑ 14.0 14.1 "Indy not powering on", IRIX Network forums, 7–8 March 2023, forums.irixnet.org. Shutdown after a few minutes; weblacky on sagging rails and tantalum failures.
- ↑ 15.0 15.1 15.2 "Indy bad cpu?", IRIX Network forums, 9 August 2026, forums.irixnet.org. Two supplies restored by recapping and replacing R5; R5000 module fault; IRIX 5.3 R5000 edition.
- ↑ 16.0 16.1 "Indy will not power on", IRIX Network forums, January to March 2019, forums.irixnet.org. Flickering red LED and clicking, cured with a Sony supply; dexter1 on C16.
- ↑ "Repair / Refurbish Indy power supply", IRIX Network forums, weblacky, 20 July 2023, forums.irixnet.org. Failed supplies usually have semiconductor damage; hot load resistors.
- ↑ 18.0 18.1 "An unusual drill mod for a DS1386-8k-120 Dallas RTC...", IRIX Network forums, weblacky, 22 January 2022, forums.irixnet.org. Clock resetting to 2052; external CR2032 holder.
- ↑ 19.0 19.1 "problems with the etheret address of my SGI indy r5000", IRIX Network forums, 29–30 April 2019, forums.irixnet.org. "Bad ethernet address", fixed with
setenv -f eaddr. - ↑ "DALLAS DS1386 replacement?", IRIX Network forums, 22 June 2022, forums.irixnet.org. Indy that would not boot, with the serial console pointing at the timekeeper.
- ↑ "ec0 bad ethernet address", IRIX Network forums, 23 July 2023, forums.irixnet.org. PROM 5.3 Rev B7 refuses
setenv -f eaddr;fillworks. - ↑ "SOLVED!! MAC address not being set", IRIX Network forums, 30 April to 2 May 2019, forums.irixnet.org. Read-only
eaddr;fillcommands take effect after a power cycle. - ↑ Adrian Chadd, "Installing a 32k DS1386 into an SGI Indy that expects an 8k DS1386", 15 December 2025, adrianchadd.blogspot.com.
- ↑ "hello network, lots of questions", IRIX Network forums, weblacky, forums.irixnet.org. Indy system boards use tantalum capacitors that can burn on poorly regulated power.
- ↑ "Value of capacitor C602 on Indy logic board needed!", IRIX Network forums, 30 May 2019, forums.irixnet.org. Burnt C602, 4.7 µF 20 V.
- ↑ 26.0 26.1 "Challenge S blew a tantalum cap", IRIX Network forums, 22–23 February 2019, forums.irixnet.org. 10 µF 25 V tantalum between the GIO sockets of a Challenge S; dexter1 on Indy failures near the IndyCam connector.
- ↑ 27.0 27.1 SGI, 007-9804-060, pp. 113–122. IDE diagnostics and confidence tests.
- ↑ SGI, 007-9804-060, pp. 127–128, Figures 7-12 and 7-13. Disabling the System Maintenance password.
- ↑ 29.0 29.1 SGI, 007-9804-060, pp. 106–107. Internal and external SCSI devices not working.
- ↑ SGI, 007-9804-060, p. 171. The Indy ships with a high-density active terminator.
- ↑ SGI, 007-9804-060, p. 105. Cleaning the mouse.