SGI Crimson Troubleshooting Guide
.jpg)
Faults on the SGI IRIS Crimson are found mainly with the status panel behind the front door, a serial terminal and SGI's IDE diagnostics. Most of the serious faults owners report are in the 1050 W power supply, which can damage the boards if it is left running while faulty.[1][2] SGI's documents describe the indicators and the diagnostics; the power supply information below comes from owners who have repaired these units.
Order of work
[edit | edit source]- If the system switches itself on with the front switch off, or any over-voltage LED lights on the status panel, switch off at the breaker and test the power supply on the bench before going further. jan-jaap's advice is not to put the boards at risk until the supply is known to be good.[1]
- Connect an ASCII terminal to serial port 1 (the port SGI's guide calls 1A) at 9600 baud, 8 data bits, 1 stop bit, XON/XOFF. SGI recommends a terminal for running diagnostics, and a server has no other console.[3]
- Watch the status panel LEDs and the hexadecimal display through power-on.
- At the System Maintenance Menu, use option 5 for the command monitor and type
hinvto list the hardware the PROM has found; use option 3 to run diagnostics.[3]
Status panel
[edit | edit source].png)
The PS3 status panel carries the operating controls on its outside and the diagnostic indicators inside the front door.[4]
| Indicator | Normal | Meaning when abnormal |
|---|---|---|
| Green run LED | On whenever the power switch is on and DC levels are normal | Off: no DC power, or levels out of range |
| Yellow fault LED | On at power-on, off once the operating system has booted | On continuously: software failure, or the hex display shows something other than 0 or 1 |
| Hexadecimal display | Alternates between 0 and 1 | Any other value is a status or fault code |
| Voltage status LEDs | Not given by SGI | Show the state of the +5, +12, −5 and −12 V rails and chassis temperature; owners report separate over- and under-voltage lights |
| Thermal or voltage circuit breaker | Plunger in | Tripped when exhaust air exceeds 60 °C (±5%) or any supply voltage moves more than 40% from nominal |
To reset a tripped breaker, press the plunger back in. It will only reset once the exhaust air has cooled to 40 °C for a minute or more.[4] jan-jaap recalls that the display counts down from F to 0 during the power-on tests before alternating at the PROM monitor, and that an F that never changes most likely means the processor has stalled.[1] The PS3 also has voltage test points.[5]
Power-on modes
[edit | edit source]An eight-way DIP switch behind the status panel selects the power-on mode. SGI ships it with every switch closed (00).[4]
| Setting | Mode | Effect |
|---|---|---|
| 00 | Default | Normal power-on tests and boot |
| 81 | Power-on (PON) | The processor drops into power-on mode, with its stack in the first-level cache |
| 82 | SYMMON | The PROM drops into SYMMON, the CPU debugging tool |
| 84 | NODIAG | The power-on diagnostics are skipped |
| 88 | MPDEBUG | The processor drops into the debugger and prints the power-on test results on serial port 1 |
Power supply faults
[edit | edit source]The Crimson uses a Power-One SPM5 or a Cherokee supply. SGI gives no procedure for either beyond replacing the unit.[1][6]
| Symptom | Likely cause | Action |
|---|---|---|
| System comes on when the rear breaker is switched on with the front switch off; IP17 LEDs lit, hex display stuck at F, over-voltage LEDs for 5 V and 12 V | Power-One control circuitry on the supply's backplane not responding to the inhibit signal; capacitors in the 5 V module leaking | Switch off at once. Bench-test the supply; CB_HK replaced the input module, backplane and 5 V module from a donor SPM5[1] |
| Card LEDs all blink rapidly together; system powers up and down | Power supply fault | Power down immediately and test the supply[5] |
| Fans start slowly with the front switch off, keyboard and network LEDs flicker, acid smell when switched on; hex display stays at F | 5 V output module soaked in electrolyte from its capacitors | Recap and clean the 5 V module (34 capacitors replaced in total by dexter1)[7] |
| Outputs switch on and off rapidly as soon as mains is applied, but are good once the system is switched on | Not the input rectifier or main filter capacitors: jan-jaap found the 300 V bus steady on a supply with this fault | Investigate the control and bias circuits; dexter1 saw the same oscillation before recapping his supply[8][7] |
| Supply fan does not run and no outputs, mains present | Failed auxiliary bias supply (UC3844 controller and BUK456 switching transistor), blown fuse, failed fan or corroded fan connector | Check the fuse first, then the bias supply and the fan wiring[2][7] |
| System runs for about 20 minutes and then reboots by itself, after a supply repair | Aged electrolytic capacitors elsewhere in the system | CB_HK replaced every electrolytic he could find, including those on the backplane[7] |
| Supply fan runs, but nothing happens at the front switch | Dirty power switch on the status panel | Work contact cleaner into the switch and test it with a meter, or replace it[1] |
| Cherokee supply: charred board near an optocoupler and LM339 comparator; after repair, all rails go to under-voltage after a few seconds | Fault in or around what jan-jaap identified as the supply's feedback circuit | Geoman replaced the supply with an LH Research Mighty-Mite unit[5] |
The Power-One's auxiliary bias supply is the part jan-jaap has found failed most often; he reports that in most SPM5 supplies he has seen, its switching transistor had failed.[2]
Bench-testing a Power-One supply
[edit | edit source].jpg)
Warning: the Power-One's input module, internal 300 V bus and parts of its control circuit are not isolated from the mains, and about 70 V can remain on the filter capacitors ten minutes after power is removed. Use an isolated oscilloscope probe; GanjaTron powered his test rig from an isolating transformer and variac.[2][7]
- Take the supply out of the system. Check that the mains link on the input terminal block matches your supply; jan-jaap expects a unit wired for 110 V to have a link between terminals 1 and 2.[1]
- Apply mains. The fan should run and the internal power bus should read about 300 V DC, while the outputs read 0 V. dexter1 measured 317 V on his.[1][7]
- Link pins 10 and 11 of the white 3 × 4 connector on the input module to switch the supply on. This connector is isolated from the mains. The fan should speed up and each output terminal should show its voltage, even without a load.[1]
- If the 300 V bus is missing, replace the input module. If 300 V is present but the fan does not run, look at the auxiliary bias supply, the fan and its wiring.[1]
- Only then look at the output modules. jan-jaap ran his rebuilt supply for hours with a 40 A load before refitting it, and GanjaTron used 5 W load resistors on the outputs.[2][7]
When reinstalling, check the supply fan connector and the inhibit wiring from the status panel. Recapping is covered in SGI Crimson Capacitor Replacement Guide.
Overheating
[edit | edit source]If the thermal breaker trips, check that every blower in the fan tray turns. A RealityEngine system needs the six-blower tray and the fan controller jumpered for full speed; without the jumper the fans stay at low speed.[9] The over-temperature circuit is on a small board with a blue heat sink above the backplane, wired to a temperature sensor above the card cage. jan-jaap had this circuit fail on a Power Series machine, where it tripped the breaker on its own.[5]
Settings and console faults
[edit | edit source]- "Non-volatile RAM checksum is incorrect" at power-on: the battery in the DS1216 socket on the IO3B is exhausted. Replace or modify it and set the variables again with
setenv.[10] See SGI Crimson General Maintenance#Settings battery. - No picture, and
printenvshowsgfx=dead: the console is set to graphics (console=g) and no working graphics board was found. Setconsole=dto use the serial port.[10] - A RealityEngine system can take up to 60 seconds to show anything on screen at power-on. SGI says not to press reset, which only delays the boot.[3]
- Wrong date after rebooting: the Crimson is not year-2000 ready. Set the date or use a time server.[11]
SCSI faults
[edit | edit source]The IO3B has two Western Digital WD33C93A SCSI controllers, one for each channel; CB_HK's board had AMD-made equivalents.[11][10] CB_HK found the second channel dead. jan-jaap advised against replacing the controller chip, because the thick multilayer board is easily damaged; cable all the drive bays to the working channel instead. If the boot channel is the one that failed, the PROM boot variables have to point at the other channel.[10] If a newer disk on an SCA adapter gives errors, try an older drive: Geoman's IBM DCHS worked where a later IBM DDRS and an 18 GB Quantum did not.[12]
Diagnostics
[edit | edit source]SGI's IDE diagnostics for the Crimson come on tape or CD-ROM. The first run copies them to disk: the IP17 tests to /stand and the graphics tests to /usr/diags.[3] jan-jaap notes that ide.IP17 can also be loaded by hand from the CD-ROM, in the same way as sash, if IRIX is not installed.[10]
- Run them from option 3 of the System Maintenance Menu. At the
DIAGS:prompt,mlists the test menus and?lists the commands.[3] - Reset the system before running the diagnostics a second time; SGI warns that it will otherwise hang.[3]
- Some disk tests write to partition 1 and only run once disk write mode is turned on.[3]
- The graphics test chapters in SGI's guide list the tests for Entry, XS, XS24, Elan and the MG1 board, for VGX and VGXT, and for RealityEngine's GE8, RM4 and DG2 boards.[3]
For failed electrolytic capacitors in general, see Capacitor Failure Symptoms.
Related pages
[edit | edit source]- SGI Crimson
- SGI Crimson General Maintenance
- SGI Crimson Capacitor Replacement Guide
- SGI Crimson Documentation
References
[edit | edit source]- ↑ 1.00 1.01 1.02 1.03 1.04 1.05 1.06 1.07 1.08 1.09 CB_HK, jan-jaap and dexter1, "Crimson Help - PSU (SOLVED)", IRIXNet forums, September 2018, forums.irixnet.org/thread-427.html. Self-powering system and over-voltage LEDs, hex display codes, bench test, mains link, dirty power switch.
- ↑ 2.0 2.1 2.2 2.3 2.4 jan-jaap and others, "Rebuilding a PowerOne PSU for a 4D PowerSeries/Crimson", IRIXNet forums, April to May 2022, forums.irixnet.org/thread-3477.html. Bias supply failures, residual charge, load test.
- ↑ 3.0 3.1 3.2 3.3 3.4 3.5 3.6 3.7 SGI, 108-7034-020, chapter 6, pp. 6-1 to 6-5 and the test tables that follow. Bringing up the system, terminal settings, System Maintenance Menu, loading and running IDE, RealityEngine start-up delay.
- ↑ 4.0 4.1 4.2 4.3 4.4 SGI, 108-7034-020, sections 4.1 and 4.2.1, pp. 4-3 to 4-9, Table 4-1 and Figure 4-4. Status panel controls and indicators, circuit breaker limits, power-on switch modes.
- ↑ 5.0 5.1 5.2 5.3 Geoman, jan-jaap and others, "Red Fever Trouble", IRIXNet forums, 2021 to 2022, forums.irixnet.org/thread-2828.html. Blinking card LEDs, Cherokee supply fault, over-temperature circuit, PS3 test points.
- ↑ Silicon Graphics, IRIS Crimson Installation and Configuration Guide, 108-7034-020, 1992 (File:SGI IRIS Crimson Installation and Configuration Guide (108-7034-020).pdf).
- ↑ 7.0 7.1 7.2 7.3 7.4 7.5 7.6 dexter1, CB_HK, jan-jaap and GanjaTron, "My Crimson PSU Troubles", IRIXNet forums, February to July 2020, forums.irixnet.org/thread-1896.html. Electrolyte-soaked 5 V module, 317 V bus, reboots after 20 minutes, test rig with load resistors and variac, corroded fan connector.
- ↑ jan-jaap, "SGI PowerOne PSU, round two", photo log, vdheijden-messerli.net/sgistuff/photos/2009.12.15-powerone/ (retrieved 2026-10-21). Output oscillation at power-up on a Power Series supply of the same type.
- ↑ Silicon Graphics, IRIS Crimson Graphics Upgrade Installation Instructions, 108-7035-020, section 3.9.2, pp. 3-23 to 3-24 (File:SGI IRIS Crimson Graphics Upgrade Installation Instructions (108-7035-020).pdf). RealityEngine fan speed.
- ↑ 10.0 10.1 10.2 10.3 10.4 CB_HK, jan-jaap and GanjaTron, "NVRAM Checksum Issue (Solved)", IRIXNet forums, October and November 2018, forums.irixnet.org/thread-555.html. NVRAM checksum error, PROM variables, failed SCSI channel, loading ide.IP17.
- ↑ 11.0 11.1 Gerhard Lenerz, "Crimson", sgistuff.net, sgistuff.net/hardware/systems/crimson.html (retrieved 2026-10-21). WD33C93A SCSI controllers, Y2K.
- ↑ Geoman and jan-jaap, "Crimson Harddisk Installation woes", IRIXNet forums, September 2022, forums.irixnet.org/thread-3681.html. SCA drives on adapters.