Jump to content

SGI Fuel Troubleshooting Guide

From RetroTechCollection
SGI Fuel rear panel (Rules666)
Rear of a Fuel: the L1 port (USB-B) is below the Ethernet socket, beside serial ports 1 and 2 (Rules666, CC BY-SA 4.0)

The Silicon Graphics Fuel reports its state through the light bar on the front bezel and through its L1 system controller. Most faults owners meet start with the power supply or with the chips that monitor voltages, temperatures and fan speeds; a failing supply can damage those chips.[1] SGI's diagnostic reference manual and owner's guide cover the light bar, the L1 and the diagnostics; the supply and monitoring-chip faults below come from owners' repair threads.

Order of work

[edit | edit source]
  1. Connect a terminal to the L1 before anything else; owners on the IRIXNet forums ask for its messages first.[2][3]
  2. Check the supply's outputs with the L1 env command, and its 12 V rail in particular.[4]
  3. Watch the light bar through power-up and compare it with the table below.
  4. If the system starts, run the offline or online diagnostics before replacing parts. SGI's first steps are to swap the keyboard and mouse ports, reseat cables, DIMMs and PCI boards, then reset and run the diagnostics again.[5]

Connecting to the L1

[edit | edit source]
  • The internal DB-9 connector on the system board carries the L1 and power-on diagnostic output. Set the terminal to 38,400 baud, 8 data bits, no parity, 1 stop bit, RTS/CTS flow control, and use a null-modem cable. Press Enter for the 001a01-L1> prompt, Ctrl+D for the diagnostic console and Ctrl+T to return to the L1.[6][7]
  • The L1 port on the rear panel is a USB-B socket. It needs a second system running SGI's L2 controller software (/stand/sysco/bin/l2).[6] kubatyszko notes that SGI's SGIconsole software in a virtual machine running old Red Hat Linux can serve as that system.[3]
  • The IRIX console is on serial port 1 at 9600 baud, 8 data bits, no parity, 1 stop bit, hardware flow control.[7]
  • pwr up powers the system up from the L1 and shows where it stops; env lists voltages, temperatures and fan speeds with their warning and fault limits; leds reads the virtual LED value that SGI's diagnostic manual decodes in Table 3-6.[6][3][4]

Light bar

[edit | edit source]
Fuel light bar patterns
Pattern Meaning SGI's action
Off (perhaps after a brief LED sequence) System off Check the power cord and mains; replace the supply or the IP34 system board
Blinking white Power button pressed; system powering on or off Normal during power-up and shutdown; see below if the system power-cycles on its own
Solid white PROM booted successfully, or IRIX running With no display: replace the system board or graphics board
Solid red System board failure: the PROM could not be read at power-on Replace the system board or PIMM
Blinking red During boot: memory failure. With IRIX running: kernel panic Reseat or replace the DIMMs, or replace the IP34
Blinking red and white, alternating Graphics configuration error; can also be caused by the keyboard or mouse Try another keyboard and mouse; reseat or replace the graphics module
Blinking red, red, white, white Faulty secondary cache Replace the processor module or IP34

[8][6][9][10]

SGI explains the keyboard case: with the console set to ioc3, the PROM stops if it finds no keyboard.[10] weblacky saw solid red on a board whose PROM later read back correctly, and has asked whether a processor module of the wrong speed for the board gives the same pattern; this is unconfirmed.[11]

Environmental shutdown

[edit | edit source]

SGI's guide says the Fuel powers itself off if any fan falls below 80% of nominal speed, any temperature sensor reaches 70 °C or any voltage moves 20% from nominal; check for these first when it shuts down unexpectedly.[8] weblacky's L1 limits for the 12 V rail were a warning at 10% (10.80 to 13.20 V) and a fault at 20%.[12] indigofan once found that an apparent monitoring fault was a fan that needed reseating.[2]

The fan numbers in the L1 output are: FAN 0 exhaust, FAN 1 hard disk, FAN 2 PCI (upper fan of the three in the PCI and graphics housing), FAN 3 XIO 1 (lower front), FAN 4 XIO 2 (lower rear) and FAN 5 power supply.[13]

Faults reported by owners

[edit | edit source]
Fuel faults and their causes
Symptom Likely cause Action
L1 reports the 12 V rail below specification and powers down, mainly after the system has been off for a long time; it works after a few attempts Power supply capacitors; megaimg measured the rail falling to 10.2 V Recap or replace the supply[4]
System runs for a few seconds, then shuts down; L1 shows missing or impossible sensor readings, or "Error initializing the ODYSSEY monitor, no acknowledge" Failed DS1780 monitoring chip on the system board, PIMM or graphics board ("ENV disease"), often after the supply went out of regulation Replace the failed DS1780. weblacky replaced all three on his system board before finding the faulty one was on the V10 graphics board[1][12][2]
Nothing at all from the L1 or the power button until the power connector is unplugged and replugged; recurs after mains is removed DS1742W L1 NVRAM and clock chip with a dead internal battery Replace the DS1742W, or fit a replacement board with a coin cell[14][15]
PROM settings corrupt after every cold start Flat SNAPHAT battery cap on the real-time clock Fit a new SNAPHAT cap, then reset the PROM variables[12]
Fans run, white light bar flashes and the system power-cycles or reboots by itself Power supply, its connections or the front power button SGI's advice: reseat the supply and its connectors, run with the front bezel removed in case it is pressing the button, then replace the supply[16]
Front power button does nothing, but the system starts with the L1 power up command The thread did not establish a cause; Raion suggested checking the switch circuit for continuity Start the system from the L1 meanwhile, and check the switch and its harness[2]
L1 date shows a date years out Flat clock battery. jwhat attributes it to the SNAPHAT cap; cz7asm describes the DS1742W as the L1's clock Check both batteries[2][15]
Memory sizes wrongly, one DIMM socket defaulting to 256 MB, with XBOW errors Debris or a fault on the DIMM socket lines Clean the sockets and inspect the board around them; Raion once cured this by cleaning the sockets[12]
No L1 output and no motherboard LEDs, only a relay click from the supply Standby supply or system board; possibly a failed monitoring chip Connect the L1; Raion advises replacing the original supply in any case[3]

Owners stress not to run the Fuel with environmental monitoring disabled to get around a failed monitoring chip. Raion has seen VPro boards burnt out this way, because the monitoring controls the fan speeds.[1][3]

Diagnostics

[edit | edit source]
  • Power-on diagnostics run at every power-on and reset. They test the processor, Bedrock ASIC, PROM, DIMMs, secondary cache, Xbridge, PCI slots, serial ports, SCSI, keyboard and mouse, graphics and Ethernet, and disable any part that fails.[8]
  • Offline diagnostics: choose Run Diagnostics from the System Maintenance menu, or type boot -f dksc(0,1,0)/stand/smdk/smdk --a at the PROM. The launcher tests the PIMM, secondary cache, DIMMs and system board; SGI's timing for a 500 MHz system with 512 MB is about 7.5 minutes.[8]
  • Online diagnostics: install them from the Customer Diagnostics CD (812-1122-001), then run ./runalldiags in /usr/diags/bin as root with the system idle. Basic mode tests memory and runs a 30-minute stress test; normal and extensive modes add I/O tests that disrupt the serial ports, USB and Ethernet.[8]
  • After a fault has been fixed, jwhat's advice for clearing old errors is enableall and update at the PROM command prompt.[12]
  • For the meaning of a stuck virtual LED value, see Table 3-6 of SGI's diagnostic manual.[6]
[edit | edit source]

References

[edit | edit source]
  1. ↑ 1.0 1.1 1.2 blacksmith, Raion and weblacky, "Dead Fuel", IRIXNet forums, October 2021, forums.irixnet.org/thread-3175.html. DS1780 failures and supply damage; do not disable monitoring.
  2. ↑ 2.0 2.1 2.2 2.3 2.4 stuart, Raion, indigofan, Trippynet and jwhat, "Fuel not powering-on from front-panel button", IRIXNet forums, January to February 2021, forums.irixnet.org/thread-2673.html. Fan reseat, start from the L1 with power up, wrong L1 date and the SNAPHAT battery.
  3. ↑ 3.0 3.1 3.2 3.3 3.4 SkypeForever!, Raion and kubatyszko, "SGI Fuel doesn't turn on", IRIXNet forums, January to February 2026, forums.irixnet.org/thread-4810.html. L1 connection, pwr up, SGIconsole, failed monitoring chip.
  4. ↑ 4.0 4.1 4.2 megaimg, jwhat and weblacky, "SGI Fuel PSU Replacement Options or Service the Original", IRIXNet forums, November 2021 to March 2022, forums.irixnet.org/thread-3233.html. 12 V rail sag, L1 env output.
  5. ↑ Silicon Graphics, Silicon Graphics Fuel Visual Workstation Diagnostic Reference Manual, 108-0350-002, July 2002, section 1.6, p. 1-11 (File:SGI Fuel Visual Workstation Diagnostic Reference Manual (108-0350-002).pdf). Before you replace components.
  6. ↑ 6.0 6.1 6.2 6.3 6.4 SGI, 108-0350-002, sections 3.4.1 to 3.4.4, pp. 3-10 to 3-14. Table 3-5 light bar patterns, L1 diagnostic port, DB-9 connector and terminal settings, virtual LED table.
  7. ↑ 7.0 7.1 SGI support note sg4941, "How to connect a serial console to an SGI Fuel?", 2009, same collection. Serial port 1 at 9600 baud; internal L1 port at 38,400 baud.
  8. ↑ 8.0 8.1 8.2 8.3 8.4 Silicon Graphics, Silicon Graphics Fuel Visual Workstation Hardware User's Guide, 007-4480-001, February 2002, chapter 5, pp. 109-117 (File:SGI Fuel Visual Workstation Hardware User's Guide (007-4480-001).pdf). Environmental fault monitoring, Table 5-1 light bar signals, power-on, offline and online diagnostics.
  9. ↑ SGI support note sg1796, "Decode lightbar status (diagnostic LEDs) for a non-booting Fuel", 2004, in the Internet Archive collection HPE Documentation - Silicon Graphics & IRIX Troubleshooting Documentation. Actions for each pattern, secondary cache pattern.
  10. ↑ 10.0 10.1 SGI support note sg4187, "Fuel - Alternate blinking red/white LED", 2002, same collection. Keyboard check, graphics reseat.
  11. ↑ weblacky, "What's the Fuel's front LED sequence when using a mismatched (slower) processor PIMM?", IRIXNet forums, August 2023, forums.irixnet.org/thread-4040.html.
  12. ↑ 12.0 12.1 12.2 12.3 12.4 weblacky, jwhat and Raion, "The start of a LONG Fuel repair thread...", IRIXNet forums, October 2021 to January 2022, forums.irixnet.org/thread-3185.html. DS1780 replacement, ODYSSEY monitor error, memory sizing fault, SNAPHAT, L1 limits, clearing errors.
  13. ↑ SGI support note sg1130, "Locating Fan location by FAN Number on SGI FUEL Workstations", 2008, same collection.
  14. ↑ cz7asm and weblacky, "Fuel needs to be power cycled to start", IRIXNet forums, April 2021, forums.irixnet.org/thread-2797.html.
  15. ↑ 15.0 15.1 cz7asm and others, "DS1742 NVRAM Chip Replacement", IRIXNet forums, January 2022, forums.irixnet.org/thread-3360.html.
  16. ↑ SGI support note sg1735, "Fuel - Blinking white LED, system power cycles or reboots on its own", 2002, same collection.


Retrieved from "https://retrotechwiki.com/index.php?title=SGI_Fuel_Troubleshooting_Guide&oldid=29307"