IBM RS/6000 Capacitor Replacement Guide

This guide documents capacitor diagnosis and replacement for the IBM RS/6000 family, covering the planar, the PSU, the IBM POWER GXT graphics cards, NVRAM TimeKeeper modules, and the drive and CD-ROM logic boards. RS/6000 capacitor topology varies significantly between eras — the Micro Channel POWER1 / POWER2 7012 and 7013 planars carry the same generation of surface-mount aluminium electrolytic capacitors that afflict the IBM PS/2 Model 70 / IBM PS/2 Model 80 planars (the "PS/2 plague"), while the later CHRP / PCI machines (7025 / 7026 / 7043 / 7044) use through-hole tantalum and through-hole aluminium electrolytics that age more slowly.

Safety Warning

All RS/6000 PSUs contain mains-rectified bulk capacitors that hold a lethal charge after power-off. Before any PSU work:

1. Power off and unplug the mains lead.
2. Wait at least 30 seconds.
3. Discharge the bulk capacitor through a 1 kΩ / 5 W resistor.
4. Verify with a multimeter.

Rack-class RS/6000s (7015, 7026 H-series) may carry dual or redundant PSUs and may be wired to −48 V DC instead of mains. Confirm the input topology before any service work.

Era Summary

RS/6000 planar capacitor topology by era

Era	Machine types	Planar caps	Failure mode
POWER1 / POWER2 Micro Channel	7011, 7012, 7013, 7015	SMD aluminium electrolytic + through-hole tantalum	SMD electrolyte leakage (same generation as PS/2 70/80)
PReP / Carolina	7248	Through-hole tantalum + small SMD ceramic	Tantalum short circuit
CHRP / PCI	7025, 7026, 7043, 7044	Through-hole aluminium + SMD ceramic + tantalum	Aged secondary electrolytics in PSU; tantalum decoupling failure on graphics cards

7012 / 7013 Planar SMD Electrolyte Leak — Primary Failure Mode

The POWER1 / POWER2 era 7012 and 7013 planars carry surface-mount aluminium electrolytic capacitors — typically 10 µF / 16 V and 47 µF / 16 V — scattered across both sides of the planar for IC supply rail bypassing. The aluminium electrolyte inside these caps is corrosive and the rubber bung seal degrades over decades, allowing the electrolyte to leak onto the planar.

Failure Progression

Identical to the PS/2 Model 70 / 80 plague:

1. Mild discolouration of the solder mask near the cap base (months).
2. Crusty deposits on adjacent vias and traces (1–2 years).
3. First intermittent POST faults — system passes cold, fails warm (2–3 years).
4. Permanent loss of specific functional blocks (NVRAM, graphics, SCSI) as traces corrode through (3–5 years).
5. Dead planar with corroded traces and lifted pads requiring extensive board-level repair (5+ years).

Detection

1. Power off, unplug, discharge PSU.
2. Remove the planar from the chassis.
3. Inspect both sides under a USB microscope or bright light.
4. Look for brown / green residue near any SMD electrolytic cap, discolouration of solder mask, crusty deposits on adjacent vias, lifted pads or corroded traces, "fishy" electrolyte smell.

If any SMD cap shows leakage, all of them should be replaced — they are all the same age.

Replacement Procedure

Identical to the PS/2 70/80 procedure:

1. Photograph the planar from both sides at high resolution. Record every cap's location, value (top marking), orientation.
2. Remove every SMD electrolytic with a hot-air rework station (set to 320 °C, 30 seconds per cap). Lift each cap straight up; do not pry sideways.
3. Clean each pad with solder wick.
4. Wash the planar with isopropyl alcohol (90%+). For severely contaminated planars, an ultrasonic cleaner with planar-safe detergent.
5. Inspect every pad for damage. Repair lifted pads or corroded traces with 30 AWG enamelled wire.
6. Fit replacements: through-hole 10 µF / 16 V tantalum (preferred) or SMD low-ESR aluminium 10 µF / 16 V; through-hole 47 µF / 16 V or equivalents.
7. Verify polarity on every cap before soldering.

No value table. The leak itself is well documented — it is the same generation of surface-mount aluminium electrolytic that fails on the PS/2 Model 70 and 80 — but the "10 µF × 30–50, 47 µF × 8–12, 22 µF × 2–4" breakdown that used to sit here had no source and its quantities were marked "approx". IBM publishes no component-level capacitor list for the RS/6000; its service guides treat the planar as a field-replaceable unit.

What to do instead: the leaking cans are visible, and the electrolyte track is visible under magnification. Read the value off each can before you remove it, photograph the area, and replace like for like. Tantalum or SMD low-ESR aluminium are both used as replacements for this generation of part.

The exact count depends on planar revision. POWER2 39H/397/590/595 planars carry more caps than POWER1 320/530 planars because the MCM has more supply pins.

PSU Recap

RS/6000 PSUs vary by machine type. Common topology: single-board switching supply with primary / secondary stages.

• Primary side — bridge rectifier, X2 mains suppression cap, bulk capacitor (typically 220 µF / 200 V on smaller PSUs, 470 µF / 200 V on 7013 / 7015, 1000 µF / 250 V on H70 / H80 PSUs), switching transistor, switching transformer.
• Secondary side — rectifier diodes, low-ESR smoothing electrolytics, output chokes, optocoupler feedback.

7012 / 7013 PSU Cap Values (Typical)

No table here. The capacitor list that used to occupy this section was self-labelled "representative" or "typical", and across this wiki's verification drive that label has, without exception, marked a table with no source behind it. IBM published no component-level capacitor list for this machine — its hardware maintenance manuals treat the planar and the power supply as field-replaceable units and give part numbers, not component values. The table has been removed rather than corrected, because there is nothing to correct it against.

Read the values off your own board. Photograph both sides before desoldering anything, match capacitance exactly, and use a voltage rating equal or higher. If you transcribe a real list from a real board, please add it here with photographs — that is worth more than any amount of plausible-looking table.

The exact cap list per machine type is in the corresponding IBM service guide (SA38-0531 for 7013 500-series).

7013 500-Series Fan-Sense Interlock

A specific 7013 500-series PSU quirk: the supply refuses to stay on if it does not sense rotation of the cooling fan. Aged secondary electrolytics cause fan stutter at low duty cycle which causes the PSU to interpret a fan failure and shut down. Symptom: PSU runs for a few seconds, clicks, restarts in a loop. Recap fixes this.

7043 / 7044 PSU Cap Values (Typical)

The 7043 and 7044 PSUs have a 3.3 V rail added for the CPU and carry slightly more secondary filtering:

No table here either, for the same reason. IBM's RS/6000 service guides (SA38-0512, SA38-0531) give FRU part numbers for the whole supply assembly; they do not give component values inside it.

7026-H70 750 W PSU

The H70 PSU is FC 6290 (IBM PN 09P5429 / 24L0586). It is a hot-swap unit; community-published cap lists exist but vary by revision — refer to the per-revision label on the PSU shell and the corresponding IBM service publication.

Recap Procedure (All PSUs)

1. Discharge the bulk capacitor; verify with a multimeter.
2. Remove the PSU from the chassis.
3. Open the PSU housing.
4. Photograph the board. Mark each electrolytic's polarity with a paint pen.
5. Desolder each electrolytic with solder wick.
6. Fit low-ESR, 105 °C replacements, equal capacitance, equal or higher voltage rating.
7. Inspect the X2 mains suppression cap. If RIFA-branded or cracked / bulging, replace.
8. Reassemble. Verify rails on the bench with a multimeter under a 1 A resistive load before refitting to the chassis.

POWER GXT Graphics Card Capacitors

The IBM POWER GXT family of graphics cards (GXT110P, GXT250P, GXT255P, GXT500P, GXT800P, GXT2000P, GXT3000P) all carry tantalum decoupling around the graphics chip and VRAM.

Common failure: GXT800P / GXT3000P tantalum decoupling short on the +5 V rail. Symptoms:

• System will not POST with the card installed but POSTs when card removed.
• GXT card heats up unevenly with hot spot near a specific tantalum cap.
• PSU +5 V rail sags when the card is inserted.

No values are given here. The GXT cards vary by model and revision and IBM publishes nothing at component level for them. The symptoms above are what to look for; the values are on the parts. Removing bulk filter near the PCI edge connector.

Diagnostic procedure:

1. Set multimeter to diode test.
2. Probe each tantalum in-circuit: black probe to ground, red probe to the rail side. Good cap reads as open / high resistance; failed (shorted) cap reads close to 0 Ω.
3. Where a short is found, remove the cap to confirm.
4. Replace with a fresh tantalum or low-ESR ceramic of equal value and equal or higher voltage rating.

NVRAM / RTC Module Replacement

RS/6000 NVRAM / RTC modules are Dallas Semiconductor TimeKeeper parts with an internal lithium cell. The cell fails after 10–20 years.

Common Dallas parts in the family:

• DS1287 / DS1287A — 24-pin DIP, same as PS/2 50/70/80.
• DS1385 / DS1387 — DS1385 + integrated battery/crystal in an encapsulated module.
• DS1644 — non-volatile TimeKeeping RAM.
• DS1742W — 32-pin nonvolatile timekeeping RAM.
• DS1746 / DS1746P — 128 KB × 8 NVSRAM + RTC, Y2K-compliant.

Replacement options:

1. Fresh-date NOS or refurbished Dallas/Maxim part.
2. Glitch Works GW-1742-1 (DS1742 substitute) — modern board with SRAM + RTC + CR1225 cell.
3. Glitch Works GW-1387-1 (DS1387 maintainable repair board).
4. Necroware nwX287 (DS12887 replacement using a CR1225 + modern SRAM + RTC chip).
5. DIY epoxy-case opening: cut the Dallas plastic case open, find the internal cell, replace with an external CR2032 holder.

Replacement procedure:

1. Power off; discharge PSU bulk capacitor.
2. Locate the Dallas DIP — typically near the RTC crystal or the boot ROM.
3. Desolder using solder wick and a fine-tipped iron at no more than 350 °C; limit each cycle to 5–7 seconds.
4. Fit a socket (recommended) before installing the replacement.
5. Power on; enter SMS (F1 / F4); re-enter boot list, IP / SLIP config, date/time.

Drive Logic Board Capacitors

Original IBM SCSI drives in long-running machines do carry aluminium electrolytics on the drive logic board, and a drive that works cold and fails warm is a classic symptom of one. The "47 µF / 16 V spindle motor driver, 10 µF / 25 V head amp, 22 µF / 16 V sector buffer" list that used to be here has been removed: the identical three-line list appeared on four different IBM capacitor pages on this wiki covering four different decades of drive, which is not how real drives are built. Read the values off the drive you actually have.

The practical alternative, and usually the better one, is to replace the drive with a solid-state SCSI emulator — no capacitors, no stiction, and it removes a substantial load from an ageing supply.

CD-ROM Drive Logic Board

The SCSI CD-ROM drives of this era carry electrolytics on their logic boards, and a drive that refuses to eject or fails to spin up is a candidate. No value list is given here — the one that used to be had no source and the drives varied by supplier and revision.

Replacement Parts Summary

No family-wide parts summary is given, because every value in it would have been invented. What is real, and is set out above, is where to look: the SMD electrolytics on 7012 and 7013 planars, the supply secondaries, and the drive logic boards.

Polarity Reference

The RS/6000 planar silkscreen marks the + (positive, rail) side of each electrolytic with a small "+". The cap body shows the negative side with a stripe and "−". Match these conventions. The convention is the same as on the IBM PS/2 line.

Polarity reference for IBM motherboard tantalum capacitors. The convention applies to the RS/6000 planar SMD positions when fitting tantalum replacements. (Image: minuszerodegrees.net)

A failed tantalum often shows no visible damage.

Failed tantalum on an IBM motherboard. The small hole on the body is often the only visual indication. (Image: minuszerodegrees.net)

Post-Recap Verification

1. Bench-test the planar with a known-good PSU and no peripherals.
2. Verify rails at the planar power connector.
3. Refit; power on; watch the LED panel through BIST and IPL.
4. Run AIX diag to verify all diagnostic categories pass.
5. Run a memory test pass for several hours via AIX diag → Memory Test.
6. Run a SCSI surface scan to confirm the storage subsystem is healthy after controller recap.

If any test fails after a recap, re-inspect the polarity of every replaced cap before suspecting another fault — reversed polarity is the most common error.

When Not to Recap

If the RS/6000 boots, POSTs cleanly to AIX, errpt shows no hardware events, and the operator panel reaches the login prompt code (typically 551), the caps are within tolerance.

Always recap if:

• Any SMD electrolyte leak visible on a 7012 / 7013 planar.
• Fluid leak visible from any electrolytic in the PSU or drive logic boards.
• PSU smoke, fishy odour or audible whine.
• 7013 500-series PSU running in click-restart loop.
• GXT graphics card pulling down +5 V or running hot.
• Drive becomes unreliable when warm.
• System passes BIST cold but fails when warm.

Related Pages

• IBM RS/6000
• IBM RS/6000 Maintenance Guide
• IBM RS/6000 Troubleshooting Guide
• IBM PS/2 Model 70 Capacitor Replacement Guide — same generation of SMD electrolytics (POWER1 / POWER2 era)
• Capacitor Failure Symptoms

References

Further reading

• IBM SA38-0531-00 — RS/6000 7013 500-series Installation and Service Guide. Authoritative PSU + planar service.
• IBM SA38-0512-03 — RS/6000 7043 / 7248 Service Guide. CHRP / PReP service.
• Commonly Failing Electronic Components, minuszerodegrees.net. Reference for SMD aluminium electrolyte leakage and tantalum failure modes.
• Glitch Works — DS1387 rebuild. Reference for Dallas TimeKeeper failure mode and modern repair boards.
• Glitch Works GW-1742-1. Drop-in DS1742 replacement.
• Glitch Works GW-1387-1.
• Necroware nwX287 DS12887 replacement.
• Analog Devices DS1742 datasheet.
• Analog Devices DS1746 / DS1746P datasheet.