IBM 3270 PC/GX Troubleshooting Guide

Revision as of 00:16, 16 July 2026 by Josh (talk | contribs) (Expand troubleshooting: RIFA/tantalum + Dallas/planar battery (161/163) + POST codes; cited)

This guide documents fault diagnosis for the IBM 3270 PC/GX (machine type 5371). POST is identical to the IBM 3270 PC and IBM 3270 PC/G — standard XT 1xx–19xx codes plus 28xx (3278/79 Emulation Adapter) and 32xx (Display Adapter / APA / PSS). PC/GX-specific failures manifest in the IBM 5378 Display Attachment Unit, the IBM 5379 19" Color Display, or the Graphics Control Program (GCP) layer.

Reference Documents

  • IBM GA33-3141-0Introducing the IBM 3270 Personal Computer/G and /GX Workstations (May 1984).
  • IBM SA38-0037-00Personal Computer Family Service Information Manual (July 1989), Chapter 10.
  • IBM 1502336PC 3278/79 Emulation Adapter Technical Reference (October 1983).

Initial Diagnosis Workflow

A working PC/GX boots:

  1. Standard IBM PC XT POST.
  2. Keyboard Adapter exposes the Display Adapter's video BIOS.
  3. Standard 3270 PC Display Adapter output is routed through the 5378 to the 5379.
  4. IBM PC DOS 2.0 / 2.1 boots.
  5. Graphics Control Program (GCP) loads.
  6. GCP initialises the 5378.
  7. Mainframe sessions render on the 5379 at 1024 × 1024 APA resolution.

If any of these does not happen, stop and diagnose at that stage.

POST Codes

Identical to the IBM 3270 PC Troubleshooting Guide:

  • 1xx–19xx — standard XT chassis.
  • 28xx — 3278/79 Emulation Adapter.
  • 3201–3250 — Display Adapter.
  • 3261–3279 — PSS card.
  • 3280–3289 — APA card.

The 5378 is not POST-tested (it operates downstream of the Display Adapter); 5378 faults present as GCP errors or no display on the 5379 rather than POST codes.

5378 Attachment Unit Diagnosis

The 5378 is larger and more complex than the /G's 5278, and is the most likely PC/GX-specific failure point:

  • No display on 5379 but POST passes — 75-pin cable between 5378 and 5379 disconnected or with pin damage; or 5378 internal PSU dead.
  • 5378 PSU dead — 5378 has its own mains lead and internal PSU. Verify the mains lead is connected, the rear-panel power switch is on, and the 5378's status LEDs (if any) are illuminated.
  • CGA-compatible mode works but 1024 × 1024 APA does not — the 5378's high-resolution vector graphics processor or extended frame buffer has failed; CGA-compatible mode comes through the Display Adapter directly.
  • Mainframe graphics rendered correctly but very slowly — 5378's local processor may be running in fallback mode (e.g. host-side rendering instead of local pan / zoom).

5379 19" Display Diagnosis

  • No raster on 5379 — flyback failure (high anode voltage means flyback insulation is the most common end-of-life mechanism on 19" 5379s); horizontal output transistor failure; PSU dead.
  • Raster but no characters / graphics — video signal from 5378 not arriving; check 75-pin cable.
  • Visible phosphor burn-in — display has been used for sustained periods on static graphics. No field repair; replacement requires a 5379 donor.
  • Convergence drift (colour misregistration) — adjustment via the 5379 service controls; if unfixable, replacement.
  • Geometry distortion (pincushion / barrel) — deflection coil or yoke issue.

Graphics Control Program (GCP) Diagnosis

Same as the IBM 3270 PC/G Troubleshooting Guide — GCP runs on PC DOS, loads after standard 3270 PC Control Program init, and manages the 5378's local processor via the system unit's Display Adapter card.

GCP-specific failures on the /GX:

  • GCP loads but reports "Display Attachment Unit not found" — 5378 not detected. Verify the cable between the 5371 and 5378 (different connector from the 75-pin cable to the 5379).
  • GDDM application crashes when sending high-resolution graphics — 5378 frame buffer issue; reseat the frame buffer RAM modules in the 5378.
  • Mouse not tracked correctly — IBM 5277 cable issue or GCP misconfiguration.

Common Field Symptoms

  • POST passes but no 5379 display — 5378 PSU or 75-pin cable.
  • POST passes, DOS boots, GCP fails — version mismatch (GCP for /GX is a different image from GCP for /G).
  • Graphics work in DOS but not in mainframe sessions — 5378 vector graphics processor failure.
  • Random reboots when warm — PSU caps aged (any of the three PSUs: 5371, 5378, 5379); recap.
  • Smell of fish from any unit — RIFA X2 mains-suppression cap venting; power off immediately.
  • Severe phosphor burn-in on 5379 — CRT end-of-life.
  • Display geometry distortion appears gradually — deflection-board electrolytic ageing; recap the 5379 deflection board.

Diagnostic Workflow Summary

  1. Power on; observe POST.
  2. If POST fails, treat as standard 3270 PC chassis fault.
  3. If POST passes but no 5379 display, check 5378 (PSU, cable to 5379, cable from 5371).
  4. If 5379 shows DOS text but no high-resolution graphics, suspect 5378 vector graphics processor or expanded frame buffer.
  5. If GCP fails to load, verify version matches /GX configuration.
  6. If 5379 shows convergence drift or burn-in, suspect CRT end-of-life; donor required.

⚠️ Power-supply RIFA capacitor and tantalum shorts

Two age-related failures are near-universal on this era of IBM hardware:

  • RIFA mains-filter capacitors in the power supply are metallised-paper parts that crack and fail short with age, producing acrid smoke shortly after power-on. Replace them pre-emptively with modern X2-class parts.[1]
  • Tantalum capacitors on the planar (system board) and on ISA cards fail short with age. A shorted tantalum will prevent the power supply from starting (dead machine, PSU protection latched) — look for a cracked or discoloured tantalum and lift suspect ones to find the short.[1]

IBM PC/XT switching supplies also need a minimum load to start, so a bare supply on the bench may not run without a dummy load.[1]

⚠️ CMOS / RTC battery

This machine keeps its configuration in battery-backed CMOS, and the battery is a common failure. On AT-class boards the clock/CMOS is often a Dallas DS1287/DS12887 module with the cell sealed inside; it lasts about ten years and then dies, giving 161 / 163 CMOS and clock errors at POST (and sometimes spurious floppy-drive errors). PS/2 planars use a rechargeable barrel or pack battery that leaks and corrodes the board. Replace a dead Dallas module (or rework it with an external coin cell), and on a leaking planar battery remove it and clean the corrosion before it eats the traces.[2]

POST beep and error codes

The IBM Power-On Self Test signals faults by beeps and, where a display works, by a numeric code:

IBM POST beep codes
Beeps Meaning
1 short Normal POST — system OK
2 short POST error (numeric code shown on screen)
No beep Power supply or system-board fault
Continuous / repeating short Power supply or system board
1 long, 1 short System-board fault
1 long, 2 short Display-adapter fault (MDA/CGA)
1 long, 3 short Display-adapter fault (EGA/later)

Common numeric codes include 161/163 (dead CMOS battery/clock), 201 (memory), 301 (keyboard) and 1701 (hard disc). A code ending in the family prefix identifies the failing subsystem.[3]

References

  1. 1.0 1.1 1.2 minuszerodegrees.net — IBM failure symptoms; Repairing and Restoring an IBM XT; and Adam's Vintage Computer Restorations. Source for the RIFA mains-filter capacitor failing short (smoke) and the tantalum capacitors failing short and preventing the PSU from firing.
  2. Fixing a Flat Dallas DS1287 RTC, Classic Computers; and Reworking Dallas RTC Modules, Ardent Tool. Source for the Dallas DS1287/DS12887 internal-battery death (161/163 CMOS errors) and the leaking planar battery.
  3. Standard Original IBM POST Error Codes; and IBM — POST beep errors. Source for the IBM POST beep and numeric error codes.

References