Nintendo Game Boy Family Repair Reference

This page covers the whole Nintendo handheld line: the original Game Boy (DMG-01), the Game Boy Pocket (MGB-001), the Japan-only Game Boy Light (MGB-101), the Game Boy Color (CGB-001), the Game Boy Advance (AGB-001), and both flavours of the Advance SP (AGS-001 front-lit and AGS-101 back-lit). The Game Boy Micro (OXY-001) gets a short mention at the end, because my documentation on it is thin.

I work these on the bench constantly, so most of what follows is how I actually diagnose them, not a spec-sheet recital. A few things I lean on documented references for rather than my own repair count, and I flag those where it matters. Where a procedure is generic (corrosion cleanup, the SMD recap approach, bench testing method) I link out instead of repeating it.

The single organising idea for this whole family: the board revision, not the model name, decides the repair. Screen-kit fit, recap parts, even whether a console is worth fixing at all, all track the revision silkscreened on the mainboard. So almost every model section below starts with how to read the revision without opening the shell.

Safety first: the power trap

Read this before you plug in anything. It is the fastest way to destroy one of these, and the trap is nastier than it looks because a plug that fits proves nothing.

The line splits into two power generations, and they do not share a jack or a polarity:

Three things to burn in:

• The DMG is centre-negative, and its barrel is not unique to Nintendo. The DMG takes 6 V on a 3.5 x 1.35 mm plug, centre negative. The catch: the NEC TurboExpress uses the same 3.5 x 1.35 mm barrel at 7 V centre positive. The two adapters physically interchange and reverse-feed each other, and nothing mechanical stops the mistake. I have not bench-tested a TurboExpress brick into a DMG myself, but this is confirmed on both halves by two independent power-supply references each, and 7 V is already the top of the DMG’s own 3.5 to 7 V input window even with correct polarity. Label every bench supply and every dongle by polarity, not by plug size.
• The two 3 V models are the opposite polarity from the DMG, but on a different (smaller) barrel, so that pair at least is mechanically keyed. The MGB, CGB and Light are all centre-positive. One wrinkle the source material is clear about: the Pocket prints the voltage but not the polarity on the shell, so you cannot read it off a Pocket. Meter it or trust the table.
• Never apply DC-jack power with batteries installed. On every barrel-jack model the jack carries a normally-closed switch that lifts the batteries out of circuit when a plug goes in. When that switch is corroded or bent, the isolation is gone, and community references are blunt that alkaline cells can vent or explode when jack power hits them with cells still in. The GBA has no jack, so external power gets injected at the battery contacts instead, which is its own thing to be careful about.

The bench rule stands on its own regardless of my table: meter the barrel of any unknown adapter before the first power-on, every time.

Unlike a Game Gear, none of these will shock you. A DMG is electrically harmless to touch. If you want the console that can actually bite you, that is the Game Gear with its +34 V bias rail and kilovolt backlight inverter.

Know your board first

Every model has a no-disassembly revision tell. Learn these and you can grade a unit before you own it.

• DMG-01. Open the battery door and look through the little hole for the battery-cover clip. The board revision is printed there. Revisions -01 through -06 are normal packaged chips. Revisions -07 and -08 are cost-reduced boards where the SoC and both RAM dies are bare-die epoxy blobs (glop-top) soldered straight to the PCB, so they are unreplaceable. A -07 or -08 with dead logic or dead RAM is a parts donor, full stop. This is the highest-value thing to check on a DMG before you buy it for repair.
• Game Boy Pocket / Light. The Pocket revision is readable under the left battery spring. Early silver Pockets omitted the power LED (MGB-CPU-01) and Nintendo restored it in 1997 (the LED boards are MGB-LCPU-01/-02). An early silver Pocket has a unique silver screen lens; later ones use the same dark grey lens as every other colour. The Light is a single board revision (MGL-CPU-01) and is easy to spot: its A and B legends are embossed into the buttons themselves rather than printed on the shell, and it runs AA cells with a soft cyan electroluminescent lit screen.
• Game Boy Color. The revision number is under the left battery spring: “02” for CPU-02, “03” for CPU-03, and so on. The original CGB-CPU-01 has no number there, so an absent marking is the ID. This one matters electrically: CPU revisions A through C (early boards) have audio-register quirks, higher background noise, and are the ones flash-cart compatibility complaints cluster around; revision D fixes most of that; revision E (CPU-06 only) is the cleanest. A CGB-CPU-04 can carry either a C or a D chip and you can only tell by opening it.
• Game Boy Advance. Same battery-clip-hole trick. The robust rule: first digit “0” means a 40-pin LCD, first digit “1” means a 32-pin LCD. So -01 through -04 are 40-pin, and -10 is 32-pin (and draws less current, so it lasts longer on a set of AA). This decides which screen kit or ribbon adapter you order, so it is the first thing to check before quoting an AGB screen job.
• Advance SP, 001 vs 101. This is the one people get wrong because aftermarket shells are almost all labelled “001,” so the rear label lies on a reshelled unit. The shell-swap-proof test: press the brightness button. If the light goes fully off, it is an AGS-001 (front-lit). If it only gets brighter and never turns off, it is an AGS-101 (back-lit). Inside, AGS-001 boards read C/AGS-CPU-xx and the AGS-101 reads C/AGT-CPU-01 (note the AGT). The two boards are not interchangeable because of the lighting differences, so do not plan a board-swap repair across them.

The machines covered here, and how they differ in one line each:

• DMG-01, four separate PCBs (mainboard, LCD board, DC-DC converter, jack board), which is why it has more revision axes than anything later.
• Pocket / Light, the DMG collapsed onto one board with the video RAM absorbed into the SoC. The Light adds an EL-panel inverter (and a high-voltage warning that goes with it).
• Color, an RGB TFT panel with the drivers built in, so no contrast pot and a different LCD-bias arrangement from the DMG generation.
• Advance, a reflective colour panel with no light source at all (which is the entire reason the backlight/IPS mod market exists), plus the dual-mode cartridge slot.
• Advance SP, the first with a Li-ion pack, a real charger, and no headphone jack (audio comes out the EXT2 charge connector instead).

Common problems and fixes

Ordered roughly by how often each one actually comes up, not by how interesting it is.

Vertical dead lines or columns (DMG and Pocket panels)

This is the signature Game Boy fault. One or more vertical columns of the LCD go permanently blank or permanently black. Nintendo knew about it at launch: service bulletin #90291 directed that all early DMG units (serial G02800000 or lower) get their LCDs replaced under warranty for exactly this. In 2026 we reflow instead of replace.

The ten-second confirming test: power the unit, wind up the contrast, and press on the rubber strip below the LCD. If the dead columns come back briefly, that is the diagnosis: the LCD’s column driver bonds to the glass through a heat-seal ribbon, not solder, and that bond has lost contact pressure on those columns. The fix is to reflow the ribbon, not replace anything.

Here is the material fact that governs the whole procedure, and it is where people ruin panels: the ribbon backing is low-temperature polycarbonate and its traces are pure carbon with binders, not copper on Kapton. That is why an iron held still burns straight through it, and why you cannot solder it. Reported working technique, converging across several independent write-ups: iron around 250 to 260 C (or hot air around 300 C with a small nozzle), a chisel tip (pointed tips fail, the heat is too localised), keep it moving, and run the unit powered so you watch the columns return. The endpoint test is that you can run the iron across the entire bottom of the LCD without lines reappearing. Hot air is repeatedly reported as more durable than an iron.

Keep this distinct from a Game Gear. The heat-seal reflow is Game Boy lore specifically. The Game Gear joint is a different construction and wants a different technique, which I cover on that page. Carry the wrong one across and you destroy the part. One more ceiling to respect: if a McWill LCD kit is anywhere in the job, its pads die above 280 C, which is below the hot-air figure above.

Splotchy black areas rather than clean columns are screen rot, a different and unfixable fault. That is a panel replacement, not a reflow.

Alkaline leak damage

Every AA and AAA model, and the DMG worst of all because it holds four cells and is the oldest. Green or white crust in the battery bay, and a console that is dead, intermittent, or drops out when you squeeze it. This is often paired with a dirty power switch on the same unit.

Diagnosis is visual, then continuity from each terminal to its board pad. If a terminal reads open or high after cleaning, the plating is gone and it needs replacing, not more cleaning. I handle the cleanup the same way I describe in my general restoration and testing writeup: neutralise, clean thoroughly, and if you use distilled vinegar to neutralise, wash it off afterward with isopropyl, because vinegar left on a board is itself corrosive. Terminal topology varies by model (the DMG has more slide-in shell terminals than the later ones), which just changes how much desoldering the job takes.

Dirty or oxidised power switch

This is the single most common no-power and erratic-behaviour cause across the whole line on an unmodified console, so I check it before any teardown. Symptoms by flavour: won’t power on, power LED flickers, the low-battery red LED trips early or randomly, the console resets when jolted, batteries drain fast, or the audio crackles (the switch grime injects noise into the ground reference). On an SP specifically it can show as the power LED stuck red on a fresh battery.

Work the switch 20 times and hit it with contact cleaner as a first pass. That is the quick, temporary fix. The reliable fix is to open the switch body and clean the contacts inside. On the GBA and SP the switch is the easiest to open (add solder to one side of the metal retaining shell, get a tweezer under it, lift as it melts, and mind the spring). Beyond saving, the AGB switch can be substituted with a DS Lite power switch, and flex replacements exist for the SP.

No power at all

Walk the path in order. Nintendo’s own DMG diagnostic table runs: dead batteries, then defective power switch, then defective DC-DC converter, then defective CPU.

Fuses. Most Game Boys carry two fuses, F1 in the battery path and F2 in the external-power path. Continuity-test both. One caution worth heeding: a fuse can beep on continuity and still be partly blown and preventing boot, and a blown fuse is a symptom, so find what blew it before fitting a new one. The AGB F1 is a 1.25 A fast-blow.

DC-jack detect switch stuck open (Pocket, Light, Color especially). Runs on the adapter, won’t run on batteries at all. The jack’s insertion-detect switch has failed stuck in the “plug present” state and is cutting the battery circuit. A USB-to-DC adapter proves the console runs on jack power and isolates the fault to the switch. Clean the jack or replace it. Do not leave a permanently bridged jack in a unit that will ever see an adapter, for the battery-explosion reason in the safety block.

DMG DC-DC converter board. The DMG makes its rails on a little daughterboard mounted vertically on the left: a self-oscillating flyback that turns the 6 V input into +5 V regulated (VDD, for the logic) and roughly -18 V unregulated (VEE, the LCD bias). Meter both at the converter output. VDD present but VEE missing gives you a blank or ghost-faint screen with working sound. Both missing points at the converter or its caps. Because VEE is unregulated it will read anywhere around -16 to -20 V and drift with the contrast setting, so do not condemn a rail on its exact value; condemn it for being absent or collapsing when the LCD is connected.

Watch that board for corrosion, not just for electrical faults. Because of where it sits, the DC-DC converter is one of the parts that catches alkaline leakage when the batteries vent, and I have had to repair lifted or eaten traces on it more than once. The good news is that it is a genuinely simple board, single or double sided with nothing multilayer to fight through, which makes it about the friendliest first patient you will find if you are learning to jumper a broken trace. Confirm the run is actually open with a meter, bridge it with fine wire, and retest the rails.

One converter-specific trap that bites USB-C dongle builders: the latest DMG converter revision (the DC-CONV2 rev D board) is the only one with an undervoltage-lockout circuit. It will cleanly refuse to start on a sagging supply where the older A/B/C boards would limp along. So a “dead” rev-D DMG on tired cells or a droopy dongle can be a perfectly healthy converter doing its job. Retest on a stiff 6 V bench supply before condemning anything, and validate any dongle design against a rev-D board, not an older one.

GBA and SP over-current lockout. The GBA runs a single regulator making all its rails. If any one output is overloaded, the regulator shuts all of them down, which shows as a quick flicker of the power LED then dead. The hunt is to inject about 1 V into each rail test pad in turn and find the shorted one rather than blaming the regulator. Two OEM numbers worth knowing before you open an AGB at all: the bicolour LED turns red below 2.35 V (normal, expected behaviour near end-of-charge, not a fault), and the console shuts down below 1.7 V. A unit that red-LEDs on fresh cells has a rail or contact-resistance problem, not a battery problem.

One AGB quirk that gets misread as a fault: turn it off and straight back on and it won’t start, but leave it ten seconds and it boots fine. In the off position the switch dumps the bulk capacitance through a 150 ohm discharge resistor (R13). A dirty switch or a damaged R13 leaves the rails charged and the console in an unbootable state until it bleeds down. Clean the switch or replace R13.

Audio faults

The highest-yield audio diagnosis in the whole line: speaker dead but headphones fine. Every Game Boy headphone jack has a switch that mutes the speaker when a plug goes in. When corrosion or damage holds that switch open, the console thinks headphones are permanently plugged in and never un-mutes the speaker. Contact cleaner into the port fixes most of these. Replacement is the annoyance, because the part is Game-Boy-specific and new old stock is basically gone, so your options are a donor console or a modern replacement. For the DMG I make a drop-in headphone jack board for exactly this, which saves hunting down a donor. The DMG uses one type; the AGB, CGB and Pocket share another.

Crackling or crunchy audio. Clean the power switch first (see above), then the headphone jack, then look at the electrolytics. On a Game Boy Color the usual culprit is a specific cap: C38 is the audio cap (quiet or crackling), C32 is the DC-DC output cap (unit won’t power on), and C35 is LCD power (dim screen). On a GBC just replace all three and move on.

Low or fuzzy volume. Usually a dirty volume wheel, which cannot be disassembled: spray contact cleaner and work it through full travel repeatedly, or replace it. A scratchy DMG volume control is the single most common misattribution in the DMG audio pile: people replace caps when it is a worn $3 pot. Plug in headphones before you blame the pot, though. The wheel sits ahead of both outputs, so it controls headphone level as well as the speaker. If headphones sound normal and only the speaker is quiet, the pot is not your problem, and the fault is downstream of it: the speaker itself, its side of the amp, or the jack’s mute switch not fully releasing. Speaker and headphones both quiet together is what actually points at the wheel. (Thanks to Smaxx on Reddit for that split.)

Hum and hiss on the GBA and SP is a design characteristic, not a fault. The amps themselves are clean (lifted onto a bench and driven from a clean generator they amplify fine); the noise originates in the console, and the SP is the noisiest of the line. The free wins are cleaning the volume wheel and the power switch. The trap: bolting a louder aftermarket amp onto a stock unit makes the noise worse, not better, unless you address the underlying causes. Do not chase this as a repairable defect on a stock console.

Recap: which models actually need it, and which don’t

This is not a blanket-recap family like the Game Gear. Ranked by how often a recap changes the outcome:

• DMG-01. Real payoff, but the electrolytics are spread across four boards, so order by board or you’ll come up short. The audio caps and the DC-DC board are the ones that matter.
• Pocket / Light. Real. A small SMD set; the audio caps drive the classic low-or-no speaker output.
• Color. Real but targeted: three caps, and I listed which does what above (C32 no-power, C38 audio, C35 dim screen). C32 is the highest-yield single cap in the whole family: a dead-on-arrival GBC with clean battery terminals is C32 far more often than anything else.
• Advance. Moderate, and mostly audio-noise driven rather than power. CP4 is the audio-amp ground reference, which is why a humming GBA is worth a recap even when the rails look fine.
• Advance SP. Minimal. There is exactly one aluminium electrolytic on the whole SP mainboard (CP1, under the volume slider); everything else is ceramic or tantalum and does not age out. A “recap the SP” job is a one-capacitor job. If the fault survives CP1, it is not a capacitor fault.

Full values, part numbers and the DC-DC revision trap are in the parts section below.

Won’t boot a cartridge

Clean the cartridge edge and the slot connector. Nintendo’s doctrine was famously “do not clean the 32-pin connector, replace it,” which reflects how easily those contacts deform; modern practice is to clean gently and only replace if that fails.

A genuinely useful diagnostic split lives in the boot behaviour. There is no CIC-style lockout chip anywhere in this line, but there is a boot-ROM logo check. A DMG that scrolls the Nintendo logo down and then freezes or flashes has already proven its CPU, RAM, LCD chain and DC-DC well enough to render, so you are almost certainly looking at a cartridge-contact or cart problem, not a console fault. A unit that never draws the logo at all is a console-side fault. That single split saves a lot of time.

On the GBA specifically, a stuck or dirty cartridge-slot mode switch gives the signature “GBA carts work, GB/GBC carts don’t” (or the reverse), because that switch is what tells the slot whether to run at 3.3 V (GBA) or 5 V (GB/GBC). Check it before condemning a slot.

Inside a cartridge: the ROM is the only part that is unique

Once you have decided the fault is the cart and not the console, it is worth knowing how little is actually in there. Most Game Boy cartridges are two chips on a stock Nintendo board: a mapper at U2 and a mask ROM at U1, with one decoupling capacitor and a pair of resistors. The game lives entirely in the mask ROM, and that chip is marked with the game’s own product code.

Here are three from my shelf. Different publishers, six years apart, and the boards are identical.

Why this matters on the bench: a rare cartridge can be repaired with a common donor. If the board is cracked, the edge connector is corroded past saving, or the mapper is dead, you do not need another copy of that game. You need any cheap cartridge on the same board, and you move the mask ROM across. The expensive part of a rare cart is the one chip that is least likely to fail.

Match the board, not the game. This is not one universal board: it is a family, and the differences are the things that change the circuit: a cart with battery-backed saves carries the SRAM, the battery and a different mapper, and larger or later games use MBC3 or MBC5 instead of MBC1. Read the designation off the donor board and match that. Two carts that look identical from the outside can be different boards inside.

When the cartridge’s own solder joints are the fault

There is a fault that sits between “the contacts are dirty” and “the mapper is dead,” and it is worth setting out carefully, because the version of it in circulation is more specific than the evidence behind it.

The claim is that the solder joints on a cartridge’s own chips crack with age and that reflowing them brings a dead game back. The broad version holds up. Several independent first-hand accounts describe a cart that would not boot, or that showed a white screen after the Nintendo logo, coming back after the chip legs were reflowed: the Game Boy Resource wiki, a UK repair guide, an iFixit thread, and Hungry Goriya’s write-up all report it working.

What none of them reports is how often it works. No sample counts, no success rate, and nobody says how many carts they opened where the joints turned out to be fine. Real, fixable, frequency unknown.

The mechanism, as asserted. Game Boy cartridge boards are 0.9 mm and flexible. Every insertion and removal flexes the board and stresses the IC legs, and the fractures that result are, in the wiki’s words, “just small enough to not be visible to the naked eye but big enough to cause electrical connectivity issues.” Three independent sources assert that mechanism. None demonstrates it. There is no cross-section, no failure analysis, no measurement anywhere. Treat it as the best available explanation rather than a settled one.

One useful thing does fall out of the flex model. If flex is what opens the joint, an unflexed board will not show the fault. The wiki makes the point directly: inspect with the game assembled and inserted in a console, because on the bench the board is not flexing and the fracture is closed. That is also what produces the works-if-I-press-it symptom, which is the strongest tell you get.

The same source gives the one search method I would actually trust: gently prod each leg on each chip and watch for movement. A properly soldered pin does not move. That is a mechanical test for a mechanical fault, and it beats staring at the joint, because with the board relaxed the fracture is closed and optical inspection is nearly useless.

Which chips, and the part I have to correct. I have heard this fault described, more than once, as the pins along the top edge of the board being the ones that go. I went looking for that and could not support it. Every repair source localises by chip and by board quadrant, never by pin number, pin row, or package edge: the mask ROM at U1, the mapper at U2, and on save-equipped boards the SRAM. The one account with enough positional detail to test the idea points the other way, since reflowing the top-left chip changed nothing and the bottom-right one fixed the cart.

The sources do not even agree on which chips. One names the mask ROM and the SRAM; the wiki names the mask ROM and the MBC, and the SRAM and MBC for corrupted sprites. That disagreement is the informative part. Nobody here is working from measured data, so treat “which chip” as a starting guess and let the symptom narrow it.

The only pin-specific instruction I found anywhere is to pay attention to pins 19 and 20 on an MBC3, and the reason given is functional rather than mechanical: that is where the real-time-clock crystal connects. If an MBC3 cart has lost its clock, that is a sensible place to look. It says nothing about mechanical stress.

One thing not to do with the pinout on this page. The 32-pin table in the pinout section is the gold edge fingers, per Pandocs. It is not a map of the chips. There is no published per-pin signal map for U1 or U2 on the DMG cartridge boards, and numbering a chip’s legs off the edge-connector table will have you probing the wrong pin entirely.

The edge-connector table does explain the symptom, though. Twenty-four of those 32 pins are address (6 to 21) or data (22 to 29) lines, so an open or intermittent connection anywhere on that bus corrupts fetched bytes rather than killing power. The boot ROM reads the logo back and compares it, which is why this failure class shows up as a garbled logo, or a logo that scrolls and then locks, instead of a dead console.

Where reflow belongs in the order: near the end. The wiki’s own does-not-boot list runs to eight steps and puts broken joints at seven. Dirty contacts, on the cart or in the console’s slot or both, are named the single most common cause. So:

1. Try another game in the same console, and this cart in another console. That split alone tells you which side you are on.
2. Remove and reinsert. Reseating fixes a surprising share of these.
3. Clean the cartridge edge fingers, and clean the console slot. Both.
4. Inspect for visible damage. Corroded or broken gold fingers are usually a dead cart.
5. If the board has a save battery, take it off and look underneath before anything else. Electrolyte creep attacks traces and vias, and the legs can look perfect while the copper under the battery footprint is gone.
6. Check continuity with a meter before you reach for the iron. A documented dead Super Mario Land turned out to be a fractured trace, cured with a jumper wire and no reflow at all. The owner proved it by holding the wire on with tape and watching the game boot.
7. Only then suspect the joints.

If you do reflow it. Take the battery off first if one is fitted. The documented method is an iron rather than hot air: a knife or J tip, generous flux, and reflowing the solder that is already on the legs. Nobody sourced here prescribes adding fresh solder, doing both rows as a matter of course, or removing the chip. There is one hot-air account, at 225 to 250 C for at least 30 seconds behind a metal shield with a hole cut for the target chip, and its author explicitly warns readers off trying it. I would take that warning seriously: a blanket heat pass over a whole package is a blunt instrument on a part you cannot buy again.

Verifying it, which is where every account gets thin. Not one source describes an electrical check after the repair. Every documented success was declared on the cart booting once. That is a weak bar, and it is weakest for exactly this fault, because the fracture you are chasing is the kind that closes when the board is sitting still. Prod every leg again and confirm nothing moves, then reassemble, insert, and try to provoke the fault by pressing and flexing the cart in the slot. A cart that survives that is a repair. A cart that booted once on the bench is a coin toss.

A note on what is actually on these boards. The two-chip layout shown above is the simple case. A save-equipped board carries four: mask ROM at U1, mapper at U2, SRAM at U3, and a battery supervisor at U4. Read the designations off your own board rather than assuming, and check the cell while you are in there. Gekkio’s hardware database records the DMG-A02 board as taking a CR1616, not the CR2025 or CR2032 people tend to reach for.

SP won’t charge

On the SP, “no power” most often means “no charge.” The path is EXT2 port, then an EMI filter (EM8), then a fuse (F1), then the charger. Field experience is that EM8 most commonly fails on the ground side, disconnecting the EXT2 ground pin. Test order: insert the charger and look for ~5 V at the charger-IC 5 V pad; if it’s absent, continuity-check EM8 and then F1. Note for resale honesty: aftermarket SP cells are typically only 400 to 700 mAh whatever the label claims, and an SP won’t boot much below about 3.4 V at the pack.

Inside the family: what differs

A short tour of what the hardware is doing, which makes the faults above make sense. What I don’t list here is shared closely enough that it isn’t worth repeating per model.

• Two power generations. The 6 V DMG generation makes +5 V logic and roughly -18 V LCD bias from a flyback converter. The 3 V generation (Pocket, Light, Color) makes the same rails, or in the Color’s case +5 V plus a +13.6 V / -15 V pair, from half the input voltage. The GBA and SP move to a multi-rail switch-mode regulator with rails down to 1.8 V. Every generation uses a jack that mechanically lifts the battery when a plug is inserted, which is why an intermittent that only appears with a plug seated points at that switch.
• Video is direct-drive on the DMG generation. The SoC drives the panel’s shift registers directly, with no frame buffer or display controller in between. That is exactly why dead columns are a connection fault (a lost ribbon bond) and not a driver fault, and why no LCD bias (VEE) gives a white/blank screen even with perfectly healthy logic. The Color changed to an RGB panel with built-in drivers, so it has no contrast pot and no external bias ladder to blame. The Advance panel is reflective with no light source, which is the whole reason the mod market exists.
• The cartridge 5 V / 3.3 V boundary. The DMG, Pocket, Light and Color cartridge slots are 5 V logic. The GBA slot is 3.3 V in GBA mode and switches to 5 V in GB/GBC mode, selected mechanically by the cartridge’s own notch profile pressing a switch in the slot. This is the boundary that decides whether a flash cart needs level shifting, which is a topic I cover elsewhere.
• No region lockout, anywhere in this line, in the NES/SNES sense. Cartridge compatibility is purely about architecture (GB, GBC-only, or GBA), never region. Shell colour is the marker: a black cart shell is GB/GBC dual-compatible, a clear shell is GBC-only.
• The master clock is 4.194304 MHz on the DMG/Pocket/Light and, interestingly, on the Advance too (the AGB multiplies it by 4 for GBA mode and by 2 for GB/GBC mode). The Color runs an 8.4 MHz oscillator. One real DMG service trap: the crystal load capacitors are tied to the CPU die revision, so if you swap a CPU across revisions and leave the wrong load caps, you get marginal cold-start oscillator behaviour (works when warm).

Mods worth knowing

I do not reproduce anyone’s install guide here. This is an orientation to what is worth doing, and a few traps that cost real money on a resale build.

Fleet rules before any screen kit

Two rules that keep a build from eating its margin. Dry-test every screen kit before install, always. Socket the kit’s ribbon, power the bare board, run an actual cartridge, and inspect, because shipping damage is frequently invisible until first power-on and vendors void the return the moment you apply adhesive, peel film, or bend a ribbon. And assume the screen adhesive is permanent: on DMG and Pocket installs a botched alignment means buying a new panel, so use the alignment bracket and prefer Kapton over the supplied adhesive where the guides allow it.

Screens: the panel is the real taxonomy

Kit marketing names churn, but the underlying salvaged donor panel does not, and the panel tells you the image size, the scaling factor and most of the quirks. The current mainstream is the “Q5” panel (from a BlackBerry Q5), which gives a 4x integer scale on GB/GBC and supports convincing pixel-grid emulation. The older “9380” (BlackBerry Curve) panel was the long-time GBA gold standard but its stock is drying up. The cheap DSi-salvage (“ITA”) panels are 1x native, low power, and have a pronounced screen-door look like a stock AGS-101.

Two cross-cutting things to be able to explain to a buyer, because both generate returns:

• FRM (frame blending). GB and GBC games fake transparency by flickering sprites, relying on the awful response of the OEM panel to blur it. Modern panels are fast, so the sprite just visibly flickers on kits without frame blending. Titles it hits include Zelda: Link’s Awakening DX, F-Zero and ZAS. State whether a kit has FRM in the listing.
• The runtime hit. Every backlit kit costs battery. A DMG Q5 kit is roughly 25 to 50% of stock runtime depending on brightness, and the power-hungry GBC drop-ins are worse (one is called out as about a third of stock at minimum brightness). Plan a Li-ion or lithium-cell conversion into the same build, not as an afterthought.

For the GBA, identifying 40-pin vs 32-pin first is mandatory, because it decides either what you buy or how you install it. For the SP, all 32-pin AGB kits are electrically compatible with the SP through a passive 32-to-34-pin adapter, but the AGS-001 front-light and AGS-101 back-light supplies are not interchangeable, so an in-console light swap is variant-specific. One GBC-specific install trap cracks panels: flush-cut the cartridge-connector pin tails (and watch the battery-clip BT+ pin) before a GBC IPS install, or an untrimmed pin destroys the new ribbon on reassembly.

Backlight and bivert (DMG, Pocket)

The cheap path that keeps the OEM panel’s look. You remove the rear reflective film, put an LED panel behind the LCD, and add a “bivert” (a 74HC04 hex inverter tapping the two LCD data lines) to sharply raise contrast. One thing worth knowing if you sell one: at least one common bivert module ships with its unused inverter inputs left floating, which wastes a few milliamps and makes heat. Unlikely to damage the console, but tie the unused inputs off or use a properly terminated board. A bivert does nothing useful behind a replacement IPS panel (it will only invert the colours), so don’t sell it as an upgrade on an IPS build.

Audio, power and video-out

RetroSix’s CleanAmp line amplifies the original audio to modern levels, but on a GBA their own warning is that CleanAmp alone makes hum and hiss noticeably worse, so budget CleanAmp plus their Dehum board together, not the amp alone. For power, the mature paths are USB-C: a drop-in Li-ion module for most models, a straightforward EXT2-pin USB-C charge board for the SP, and dongle or port-replacement options for the barrel-jack models. On video-out, the ranking is that a dedicated consolizer (a real AGB mainboard plus an FPGA board out to HDMI) beats every in-console HDMI kit on scaling and audio.

Flash carts and saves

For resale honesty, two things matter. Clone flash carts are everywhere (sold as “GB Pro+” and similar), usually identifiable by an Altera CPLD instead of the Lattice FPGA in a genuine EverDrive, and updating a clone with official firmware can brick it. If a unit ships with a flash cart, disclose whether it’s genuine. And dump the cartridge save with a GBxCart-class tool before you replace a dead save cell, so the save survives the battery swap.

What actually changes value

The screen kit is the headline, so name the panel and whether it is laminated, because that is what enthusiast buyers filter on. State FRM and state the runtime hit. Reversibility sells: a solder-free, cut-free build on an OEM shell is worth more than a trimmed original, and if you do use an aftermarket IPS-ready shell, keeping the original shell with the unit is the higher-value move. And describe the recap accurately: a “fully recapped” claim on a model that doesn’t need one (the SP, for instance) is the kind of thing that gets noticed.

Recap and parts

The one rule to carry away: identify the exact board before you order, because the parts are not uniform even within a model.

The DMG DC-DC converter is the one place a generic cap kit gets you the wrong part. The converter’s capacitor complement genuinely changes across revisions: the early A/B boards use a 10 uF output bulk cap where the later C/D boards use 100 uF, a tenfold difference. Read the cans on the board before ordering. If you stock 33 uF 25 V, 100 uF 10 V, 22 uF 25 V and 10 uF 16 V you cover every DMG converter revision. The rest of the DMG is spread across the LCD board (note the five 1 uF caps that sit under the LCD module: don’t lift the module, solder replacements to the cut legs, and modern kits ship ceramics for these five), the mainboard, and the passive headphone jack board.

For the SMD models, the values by model:

• Pocket / Light, a handful of SMD aluminium caps. Watch that the Light’s C29/C32 do not match the Pocket’s despite sharing designator numbers, so never order Light caps off a Pocket list.
• Color, three: C32 (100 uF, power), C35 (22 uF, LCD), C38 (100 uF, audio).
• Advance, four (CP1 to CP4). Two worth flagging: CP2 wants a low-ESR part because it’s a boost-converter output, and CP3 is 470 uF, whose off-the-shelf aluminium equivalent is unusually tall for a thin console, so a low-profile polymer or tantalum is the routine field answer there. On resale work be aware tantalum fails short and these sit on power rails, so I lean toward aluminium or polymer on a console I’m selling.
• Advance SP, one (CP1, 100 uF).

The pragmatic path for throughput is a pre-matched kit per model, because the vendor has already solved the footprint problem. The standing rule, worth restating because it’s where most amateur recaps go wrong: replace with equal capacitance and equal-or-greater voltage, and prefer the higher temperature rating.

Some parts are effectively unobtainium new: the DMG’s Sharp audio amp (IR3R40) and LCD regulator (IR3E02) ASICs have no modern drop-in, so it’s salvage from a donor or a board-level replacement (RetroSix’s CleanPower replaces the whole DMG converter board, CleanAmp replaces the audio path). On a fleet, your worst board is your parts stock. Full BOMs, real distributor part numbers and footprint dimensions live in my bench notes; the summary above is enough to order against once you’ve read the board.

A note on the Game Boy Micro (OXY-001)

My documentation on the Micro is thin: no service manual, no schematic, no board scan, and only a handful of surveyed units. So treat this as provisional rather than bench-proven. The one fact that changes every buying decision: the Micro plays Game Boy Advance cartridges only. It dropped backwards compatibility entirely, so no original Game Boy or Game Boy Color games at all. It brings back a real 3.5 mm headphone jack (unlike the SP), combines its link and charge ports into one proprietary connector, and uses a small, fine-pitch backlit screen whose OEM replacements are rare and expensive enough that a damaged Micro panel is usually not worth fixing. Some Micros, particularly silver ones, suffer the same brittle-plastic frame problem as late AGS-101 shells, so check the frame before quoting any work that flexes it.

The hard data behind all of the above (the per-board cap maps with the substitutes I fit, the non-cap consumables, the chip and connector pinouts, and the schematic facts) is tabled in the sections that follow. It is organised by board revision, because that is what actually decides the job. Long per-item notes fold away to keep it readable; expand any of them for the full detail.

Capacitor lists

Per board revision, not per model. The DMG splits its electrolytics across four separate PCBs with independent revision runs, so order by board or you will end up short, and read the DC-DC board’s silkscreen before you order, because C5 differs by a factor of ten between revisions.

Replacement parts

The non-cap parts a Game Boy repair actually consumes, and the handful that are donor-harvest only.

Chip & connector pinouts

The component library: each IC and connector defined once, with an interactive pin diagram and a folded pin table. The revision badge on each card shows which board revision(s) it applies to. Where I could not verify a pinout against a schematic or manual on my disk, the card says so and the table is empty.