Nintendo 64 Repair Reference

This page covers the Nintendo 64 control deck, the NUS-001, in every region it shipped in: NTSC-U, NTSC-J, PAL, the France-only board, and the Funtastic, clear-shell and Pikachu editions. It covers the controller too, because the single most common complaint on a used N64 is not a console fault at all. The iQue Player gets a short mention at the end and nothing more, because it is a completely different machine wearing the N64’s software library.

Two things about this page before you read any further.

There is no Nintendo service manual or schematic for the N64. None. Nintendo serviced these by swapping whole modules, so no factory document ever entered circulation. What exists instead is a community-recreated NUS-CPU-03 schematic drawn by a hobbyist in 2013, a high-resolution annotated board photo with legible silkscreen, the NEC and MIPS datasheets for the CPU (which is a stock part and therefore genuinely documented), and two independent community capacitor and IC lists that cross-check cleanly against each other. That is a thinner foundation than I have for a Game Boy or a Game Gear, and I would rather say so at the top than let a confident tone imply a factory BOM behind it.

I do not have deep hands-on hours with this machine. Most of what follows is worked out from documentation rather than from a long personal repair count, and I have written it that way — where a claim is somebody else’s, I attribute it, and where the documentation rests on one source I say that too. The pinouts on this page came off datasheets and source files I actually hold, and where I could not verify a pinout, the card says so and the table is empty.

The organising idea for this console: the board revision decides almost everything interesting. It decides which capacitor list you order, whether the composite-sync parts exist on the A/V connector, and above all whether adding RGB is a cheap afternoon with a small amplifier board or an expensive FPGA kit. So that comes first, right after the one thing that can actually hurt you.

Safety first: the console is harmless, the brick is not

The console itself runs on 3.3 V and 12 V. You can put your hands anywhere inside it with the shell off and nothing will bite you. If you want the machine on this site that can genuinely shock you, that is the Game Gear and its kilovolt backlight inverter.

The danger on an N64 lives entirely inside the external power brick, and it is real. Every one of these supplies has a mains-side high-voltage capacitor that holds a lethal charge after unplugging:

• NTSC bricks (110 V): typically C2, 100 µF at 200 V.
• PAL bricks (220 to 240 V): typically C2, 47 µF at 400 V — and on the Panasonic Type M design that position is C103, 47 µF at 400 V instead, so do not go looking for a C2 and conclude there isn’t one.

Discharge it and verify with a meter before anything else goes near that board. The community references are blunt about this and they are right to be.

Two more traps that are not shock hazards but will cost you money:

• Nothing about this console’s power is standard, and there is no barrel jack. The brick delivers two independent regulated rails on a proprietary 6-pin captive connector: 3.3 V at up to 2.7 A and 12 V at 0.8 A, simultaneously. Both are required. Lose 3.3 V and the power LED will not even light. Lose 12 V and the console will not attempt to start, because the board makes its own +5 V from the 12 V rail. This is also why there is no USB-C conversion for an N64 and why I do not sell a dongle for one — a single USB-C power profile physically cannot deliver two separate rails at those currents. Anything that claims to needs a real dual-output buck stage inside it, and I would meter both rails at the connector before trusting one. (The USB-C dongles I do build are for consoles that take a single DC rail. The N64 is not one of them.)
• The Expansion Pak goes in one way round. A forum report describes a reverse-inserted Pak shorting the RDRAM supply and reference rails to ground, blowing supply fuses and killing the coprocessor and the memory with it. I want to be precise about how well established this is: the mechanism is plausible and the rails named really are the RDRAM rails, but the damage claim traces to a single poster and no service document confirms it. The OEM Pak connector is keyed, so getting it backwards takes deliberate force. Some early open-source Pak recreations used a symmetrical connector that removed that protection, which is worth knowing if a console arrives with a third-party Pak in it.

Know your board first

The revision string is silkscreened along the top edge of the mainboard, reading something like NUS-CPU-03. Before you take a shell apart, though, there are three tells that cost nothing.

• Unplug the brick and shine a light into the empty power-connector recess. The revision silkscreen is visible inside. This is the one to use on a transparent Funtastic shell you would rather not open, and it is documented on ConsoleMods rather than being folklore.
• Read the serial prefix. NS1 (USA) or NUJ1 (Japan) indicates an early board, revisions -01 through -04. NS2 or higher indicates -05 or later. Every source that quotes this trick also says it is a strong hint and not proof, and I would treat it the same way — it is good enough to price a unit at a swap meet and not good enough to quote a mod job on.
• Read the bottom label on a PAL machine. “EUR” is an ordinary PAL console. “FRA” is the France-only board, and that one behaves like an early NTSC board for video purposes. It is the single most valuable label on any N64.

The definitive answer, once the shell is off, is the marking on U4, the video DAC. That chip is the whole story:

Only the early VDC-NUS hands raw red, green and blue to a separate encoder chip, which means the RGB is already sitting there on the board and just needs buffering out to the connector. Every later part folds the encoder in and takes that opportunity with it. The absence of a chip at U5 is the fastest visual check if the DAC marking is hard to read.

One thing which is not mapped anywhere I have found: which of DENC-NUS, AVDC-NUS and MAV-NUS lands on which specific later NTSC revision. The set is documented; the per-revision assignment is not. Read the chip.

A correction worth carrying: NUS-101 is not a board revision

You will see “NUS-101(EUR)” described online as a later, cost-reduced PAL board revision. It is not. NUS-101 is the Pokémon Pikachu Edition case SKU — a genuine Nintendo model number, also sold as NUS-101(JPN) and NUS-101(USA) — and Pikachu consoles carry perfectly ordinary NUS-CPU(P) mainboard revisions inside, reported across P-01 through P-03. A case SKU is not a board revision.

This one misleads buyers as much as repairers. If you are shopping for a PAL Pikachu console specifically because you read that it has a distinctive board, it does not, and the board it does have depends on the year it was built like every other N64. Check the silkscreen.

The regions, and how they differ

Limited-edition shells — Funtastic transparent colours, Clear Blue, Pikachu, the Color Editions — are cosmetic. Whatever revision the factory was building that year is what is inside.

Japanese and PAL cartridges share the same “world” notch shape, so a cartridge fitting the slot does not mean it will boot. The lockout is electronic and it is covered below.

Common problems and fixes

Ordered by how often each is the actual root cause in refurbishment work, according to the documentation, rather than by how interesting it is. The first two are the ones worth internalising, because between them they explain most “broken N64” listings.

Red LED on, no picture and no sound

Before you open anything, reseat the Jumper Pak or Expansion Pak in the top slot and clean its contacts with high-purity isopropyl. Then do the same to the cartridge.

Here is why this matters more than it sounds. The N64’s main memory is Rambus RDRAM on a channel that must be terminated, and the module in that top slot is what terminates it. The Jumper Pak contains no memory whatsoever — it is purely a bus terminator. So a console with an empty expansion slot is not a console with less RAM. It is a console that will not boot at all, and it presents exactly like a dead board: red power LED, black screen, no sound.

More than half of “dead N64” reports resolve right here, at a Pak that has worked loose or gone dirty, or one that is simply missing because a previous owner pulled it out. If you buy these to resell, check the slot before you check anything else, and never ship a console with a bare slot.

While you are in there: never blow on a cartridge. Breath moisture oxidises the brass. Clean the contacts with isopropyl on a lint-free swab, the same way I describe in my general restoration and testing writeup.

The analog stick

The other dominant symptom, and it is mechanical rather than electrical. The stock stick is a plastic bowl with a small gear set and a centring spring turning two optical encoder wheels. All three parts wear, and the result is a dead zone, a loss of range, drift, or just a loose feel. ConsoleMods describe it as essentially universal on unrefurbished sticks and I have no reason to doubt them.

There is a wrinkle in how the stick reports position that explains a symptom people misread as electrical. The encoders report relative movement, and the controller chip integrates that into an absolute position. So the stick’s zero is wherever it happened to be at power-on. Start a console with the stick held over and you get a permanent offset that looks like a hardware fault. The re-zeroing tricks people pass around — unplug and replug, or hold L, R and Start while the stick is at rest — are exploiting exactly this.

Repair tiers, cheapest first:

• Regrease and reseat. Temporary. Buys time, fixes nothing, and if you sell a unit this way you should disclose it.
• Replace the gears, or the gears and the bowl. Cheap and a genuine improvement.
• Drop in a self-contained module. The modern default is a drift-proof magnetic module — the Retro-Bit hall-effect unit with a GameCube-style gate, or 8BitDo’s TMR module. There are also all-metal mechanical rebuilds (Steel Sticks 64) and optical ones (Sharpshooter, which ConsoleMods note has a small deadzone and stiffer travel than an OEM stick).

The install is the same for every self-contained assembly and needs no soldering: seven shell screws, three assembly screws, unplug the ribbon, drop the new module in. Note the controller uses JIS #1 (or Phillips PH1) screws, not a gamebit — the console shell wants a 4.5 mm gamebit and cartridges want a 3.8 mm, so you need all three on the bench.

Completely dead, no LED at all

Swap in a known-good brick before you pick up a screwdriver. A dead N64 is far more often a dead NUS-002 than a dead mainboard, and every source I hold leads its troubleshooting with that swap.

If a good brick fixes it, the brick is the fault and it is usually repairable — see the parts section. If it does not, probe the two rails at the 6-pin connector under load: 3.3 V should read 3.3 to 3.6 V and 12 V should read 10.8 to 12 V. Then check the on-board regulators. The board has exactly two:

• U13, a 78M05, makes +5 V from the 12 V rail. The board hands you test points for this, silkscreened VI (+12 V) and V0 (+5 V) either side of the chip. This part is trivially replaceable with any modern 78M05 or L78M05.
• U12, a Sharp PQ7VZ5, makes the RDRAM termination rails off 3.3 V — Vterm at about 2.56 V and Vref at about 1.92 V, also silkscreened as test points. This one matters and it is widely mislabelled: at least one parts catalogue calls the PQ7VZ5 a “5 V regulator,” and the recreated schematic shows plainly that it is not. Do not go hunting for 5 V here. A failed U12 gives you garbage video or no boot while every ordinary rail measures fine, so it belongs early in a no-boot hunt rather than late. It is also obsolete with no documented drop-in, so a donor board is the realistic answer.

One cross-check before you condemn a supply: a shorted controller cable can pull the 3.3 V controller bus to ground and trip the brick’s protection, which looks exactly like a dead PSU. Unplug every controller and try again.

Vertical jailbars, and an audible buzz

These travel together and they are a real fault, unlike the console’s stock softness (which is not — see below). Aged electrolytics in the power supply put ripple on the 3.3 V rail, and that ripple modulates both the video and the audio.

Scope the 3.3 V rail. One source gives the thresholds as healthy under 50 mV peak to peak, and over 200 mV peak to peak meaning a tired supply; that is a single source and I have not measured it myself, so treat it as a picture of what bad looks like rather than a specification. The fix order is the brick first, then the mainboard capacitors if it persists.

Two footnotes. If you recap the mainboard and the jailbars come back, check the polarity of the capacitors near the video section — reversed parts there reintroduce the same noise. And on an already-RGB-modded console, jailbars have a completely different cause: a ground loop or a poor install, with the analog video routed alongside digital lines or the mod board sharing the noisy 3.3 V rail. That is fixed by separating the wire bundles, adding series resistors on the data lines, and generating clean local 3.3 V on the mod board, not by recapping anything.

Powers on intermittently, or drops out when jostled

The slide power switch oxidises. It gates both incoming rails, so a bad one takes the whole console down. Contact cleaner and thirty to fifty cycles fixes most of them; replace it if that does not.

No boot, garbage screen, or the cartridge is not detected

Inspect the 50-pin cartridge slot under magnification for bent pins and cracked solder joints, then reflow. The reason a single bent pin gives you garbage rather than a clean failure is worth understanding: the cartridge bus is a multiplexed 16-bit address and data bus. The address is latched in two halves by ALE_H and then ALE_L, and only then does RD or WR strobe the actual data over the same sixteen wires. One intermittent line corrupts addresses and data both.

The full pinout is in the data section below, and it is one of the better-sourced tables on this page: the address/data lines, the latch strobes and the read and write strobes are confirmed independently by two references that agree pin for pin.

No video, or one audio channel, and it changes when you wiggle the cable

Cracked solder at the A/V Multi Out. That connector takes real mechanical stress every time a cable goes in or out, and it is a straightforward reflow. Wiggle-test the cable with the console running to confirm before you touch it.

The Multi Out is the same 12-pin Nintendo connector as the SNES and GameCube, so cables interchange mechanically. What comes out of it does not fully interchange, and pin 3 is where the two machines differ most — on an N64 it is composite sync (on the revisions that populate it at all), where on a PAL SNES the same pin carries a +12 V SCART switching rail. Read the SNES page before you move an RGB cable between the two.

Crashes and colour corruption after it warms up, fine when cold

Dried-out thermal pads. The N64 has no fan, so three silicone pads are the entire thermal path from the CPU, the coprocessor and the memory into the heatsink. Decades on, they desiccate and leave an air gap.

Replace all three with 1 mm silicone pads. Pads, not paste — paste is far too thin to bridge the gap the OEM design leaves, and “just use thermal paste” is probably the most commonly repeated bad advice about this console. One correction to a claim that circulates: the third pad sits on the memory, not on a regulator. Measured package temperatures reported by ModRetro put the CPU near 49 °C, the coprocessor near 64 °C and the memory near 60 °C, which is consistent with three real heat sources.

This is also the cheapest preventive there is against the expensive failure at the bottom of this list.

Audio gone, picture fine

Almost always the coupling capacitors on the mainboard, not the amplifier. The confirming test is to jumper the AMP-NUS output straight to the Multi Out audio pins; if the sound comes back, the capacitors are dead and the section needs a recap.

Worth knowing which capacitors those are, because sources disagree slightly. One troubleshooting reference names C28. The recreated schematic shows the actual series coupling parts as C25 and C26, feeding Multi Out pins 11 and 12 through 1 kΩ series resistors, with C28 sitting elsewhere in the network. I have gone with the schematic in the data below and flagged it here so you know why the numbers you may have seen elsewhere differ.

And if audio and video are both gone while the console otherwise seems alive, stop looking at the audio section: the amplifier runs on +12 V and the video encoder runs on the +5 V derived from 12 V, so one dead rail takes both out while the 3.3 V digital side keeps running happily.

Black screen and a red LED that is actually a region mismatch

This presents identically to a dead board and it is not one. On boot, the cartridge’s CIC chip streams a region nibble to the console’s PIF — 0x1 for NTSC, 0x5 for PAL — and if it does not match, the PIF halts the CPU over its non-maskable interrupt line. Black screen, red LED, nothing else.

The thing most people get wrong is where the console’s region lives: it is in the PIF’s internal ROM, not in the CIC. The cartridge only presents its region; the console decides. That is why region-free work means a cartridge-slot notch cut for the mechanics plus a matching or switchable lockout for the electronics, and never a video-standard change.

Try a known-region cartridge first. And if a previously-working modded console suddenly does this, suspect a cracked wire on the mod before you suspect the PIF.

The lockout keeps working after boot, too. If a game does not send the PIF a particular command within about five seconds, or if the running challenge and response between PIF and CIC breaks — pull a cartridge mid-game, for example — the PIF halts the CPU again.

A controller port is dead, or the console trips off when a pad is plugged in

Either cracked solder at the port (reflow it) or, as above, a shorted controller cable pulling the 3.3 V bus down. Test with a known-good controller to split the two. The port is only three pins because by this generation the pad answers a serial protocol instead of shifting bits out — Controllers: How Five Consoles Read a Button traces how that happened and what it costs in repairability.

Saves disappearing

Only affects cartridges with battery-backed SRAM — Ocarina of Time, Majora’s Mask, F-Zero X, the Pokémon Stadium games and similar. Those hold a CR2032 with a fifteen to twenty year life and they are all well past it now. Cartridges that save to EEPROM or Flash are unaffected. The Controller Pak has the same problem and reads “Note Empty” when its cell dies; FRAM replacements exist that remove the battery entirely.

Dump the save before you change the cell, or hot-swap it powered if you want to keep it.

The expensive one: cracked solder balls under the CPU or coprocessor

Both large chips are BGA-mounted, and decades of thermal cycling in a fanless console crack solder balls. The signature is random crashes, polygon and texture glitches, audio glitches or an outright no-boot, all worsening as the console warms.

This is the last suspect, not the first. Clear the Pak, the cartridge, the supply, the connectors and the thermal pads before you go here. When you do get here, a reflow is often only temporary — the cracks reopen within months — and a reball is the durable fix and needs real rework experience. On most consoles this is where the economics stop making sense.

Myths worth not chasing

• “All N64s have jailbars.” No. A healthy stock unit is clean. Jailbars mean aged supply capacitors or a bad RGB install.
• “The blurry picture is a fault.” It is not, and this one is important enough that it has its own section under mods. Do not try to repair it.
• “Reflowing permanently fixes a BGA.” Partially, and usually not for long.
• “Use thermal paste under the heatsink.” No. Pads.
• “There is a separate reset chip.” There is not. Reset is handled by the PIF itself, so a dead reset button is a dirty switch, a broken trace, or a cracked PIF joint.

Inside the family: what differs

The N64 kept one external design for its whole life while the board inside it was cost-reduced repeatedly. Here is what actually changed, in the order it happened.

• NUS-CPU-01 to -04 (1996 to 1997). The launch architecture: a VDC-NUS video DAC feeding a discrete ENC-NUS encoder, with analog RGB present internally. The early Japanese boards and the first US board carry the well-known VR4300 floating-point multiply bug in the CPU stepping, fixed in later parts (the stepping is software-readable — early parts report processor revision 0x10, later ones 0x22). On -04 specifically, the composite-sync components on Multi Out pin 3 ship unpopulated, so a -04 has no sync there at all. That last detail started life as one troubleshooting wiki’s observation, but ConsoleMods' mod guide names the same three parts and the same exception, and borti4938’s open-source board documentation independently states pin 3 is only connected on -01 through -03. Three lineages, so I am comfortable treating it as settled.
• NUS-CPU-05 to -07 (1997 to 1998). The video DAC changes and integrates the encoder, which is the change that ends easy RGB. Two capacitor positions move or disappear (C33 drops from 68 µF to 10 µF, C145 goes). The plastic around the bottom expansion port was made smaller, so a Doctor V64 no longer seats, which only matters if you have one.
• NUS-CPU-08 to -09-01 (1999 to 2001). The most integrated boards, and the ones most often found in the Funtastic series. Two more capacitors drop out (C34, C128). The stock picture is noticeably sharper and more aliased than the early boards, which people notice and sometimes report as a fault.
• PAL, NUS-CPU(P)-01 to -03. PAL started where NTSC ended up: even the 1996 P-01 uses an integrated DENC-NUS at U4 with no U5 fitted at all, so no PAL board except one has easy RGB. P-01 and P-02 present S-video on the Multi Out but need a cable with particular parts to display it correctly; P-03 does not present S-video at all, though it is restorable.
• PAL France, NUS-CPU(R)-01 (1997). The exception to everything above. It keeps the discrete arrangement, it is RGB-capable by completing footprints already on the board, and its S-video is restorable the same way a New-Style SNES is. The bottom label reads FRA. If you are hunting a PAL console to RGB-mod cheaply, this is the one.

A few structural facts that make the fault list above make sense:

• Everything routes through the coprocessor. The CPU does not talk to memory or to the cartridge directly — the RCP contains the memory controller, the cartridge interface, the video interface and the audio interface. That centralisation is why a single cracked BGA joint produces such a scattered, unhelpful symptom set.
• Video timing and memory refresh are the same clock domain. The console issues one RDRAM refresh per horizontal sync pulse. Two subsystems that look unrelated are not.
• Region timing is set by a crystal, not a jumper. The video clock generator runs a 14.32 MHz crystal on NTSC and 17.734475 MHz on PAL, selected further by a pin on the generator. A “region-free” console can still show you a 50 versus 60 Hz mismatch on a CRT after the lockout is defeated, because the lockout and the timing are separate things.

The iQue Player, briefly

The iQue Player (China, 2003) is not an NUS-001 and nothing on this page applies to it. It is a system-on-chip reimplementation built into the controller: a MIPS core at 140.625 MHz instead of 93.75, a coprocessor integrated into the same die, 16 MB of ordinary DDR SDRAM instead of RDRAM (so no Vterm, no Vref, no Jumper Pak), games downloaded to encrypted flash instead of cartridges, and therefore no cartridge bus and no CIC. Do not carry an N64 rail, clock, capacitor list or tap point onto one.

Mods worth knowing

I am not reproducing anyone’s install guide. This is an orientation to what is worth doing on this machine and where the traps are.

The RGB gate is the whole conversation

Everything above about board revisions exists to answer one question: what will it cost to get RGB out of this console? There are two completely different answers.

On an early board (NUS-CPU-01 to -04, or the France R-01), it is cheap. The VDC-NUS already emits red, green and blue into the encoder; a small THS7374-class amplifier board buffers those three lines out to the Multi Out and you are done. The documented tap is not the DAC legs but the three vias on the board underside beside R8 (red), R9 (green) and R10 (blue) — and the warning that comes with it is worth repeating, because those vias sit directly under the DAC: keep the wire stub through the hole extremely short or it will short the DAC’s pins. The direct-to-chip fallback is DAC pins 17, 19 and 21, confirmed by two independent sources.

Sync is the other half. Composite sync lives on Multi Out pin 3, and clearing the existing parts off it (C22, R1 and R14) is the standard step — except on a -04, where those parts were never fitted. Pre-assembled amplifier boards with a sync stripper are worth the money over a bare DIY board; the sources are consistent that they use better amplifier silicon and solve sync properly.

On any later board, RGB requires a kit that taps the coprocessor’s digital video and regenerates analog RGB. A CPLD or FPGA board, not an amplifier. borti4938’s open-source N64RGB (with an adapter that takes the digital tap off the RCP) and Tim Worthington’s commercial kit both do this, and both come with switchable deblur as a bonus. It costs several times what the early-board mod does, which is exactly why reading U4 before quoting anybody a price matters.

Digital video, and why it sidesteps the whole problem

The HDMI kits tap the coprocessor’s digital video directly, so they work on every board revision and are completely independent of the DAC lottery.

PixelFX’s N64Digital is the current one worth buying: HDMI out while simultaneously still driving analog RGB or component through the Multi Out, an on-screen menu, and a proper deblur. It is explicitly an advanced install — a flex soldered to the coprocessor, more connections to the PIF (their instruction is to solder to PIF legs only, never vias, and they note some late boards route one line to a via instead of the expected pin), and a case cut for the mini-HDMI unless you fit a no-cut bracket. On boards -01 through -04 the optional RGB flex wants C22 and R14 removed first, which are the same sync parts the analog RGB mod removes.

UltraHDMI is the older kit, discontinued and not sold to consumers. It matters now mainly as something you may have to service in an already-modded console. Know two things about it: its firmware update is done by displaying a special image from a flash cartridge, takes about six minutes, and bricks the unit if interrupted; and the second hardware revision added analog RGB and component to the Multi Out where the first did not.

There are open-source equivalents (borti4938’s N64Advanced and N64Advanced2) if somebody wants a repairable path rather than a finished product.

The blur is not a fault, and only one kind of it can be fixed

The N64’s soft picture comes from two independent stages, and the distinction decides what you can accurately claim in a listing.

1. The coprocessor’s anti-alias and dither filter, controlled per game through a register the game code writes. Because it is software, it can be defeated in software: GameShark codes, ROM patches, or precompiled patch sets. No hardware risk, works on any board through a flash cartridge. Some games do genuinely look worse with dither fully off.
2. A horizontal 320 to 640 resample applied after the coprocessor. This one cannot be disabled by any software or hardware mod. The only thing that fixes it is a video processor that resamples with corrected horizontal timing — the deblur feature in UltraHDMI, N64Digital and Tim Worthington’s kit, or an OSSC optimal-timing profile.

So: do not sell “deblur” on the strength of a software patch. It addresses half the problem.

Region-free

Two independent barriers, and defeating one does nothing on its own.

The mechanical barrier is the shape of the plastic guides in the cartridge slot, removed with a careful trim. It is cheap and it is irreversible on that shell, so only do it on a console you are actually selling as region-free, and disclose it.

The electronic barrier is the PIF-versus-CIC region check described earlier. A flash cartridge with CIC emulation sidesteps it for backups. An original out-of-region cartridge needs a region-switchable PIF replacement, because the console’s region byte lives in the PIF.

Expansion Pak

Trivial, and it should be on every console you sell. It replaces the Jumper Pak and takes the console from 4 MB to 8 MB, which Donkey Kong 64 and Majora’s Mask require outright. If a donor Pak is not available, a full open-source 1:1 recreation exists as a KiCad project.

Ship something in that slot regardless. A Jumper Pak is the minimum, and a console with an empty slot does not boot.

S-video restore on late PAL boards

Niche, but a real premium on the right unit. A PAL NUS-CPU(P)-03 does not present S-video on the Multi Out and it can be restored with board-specific rework; the France R-01 restores the same way a New-Style SNES does. NTSC boards output S-video natively and need nothing.

What actually changes value

The stick module is the cheapest reliability win per console and the first thing a buyer notices, so do it on every controller. RGB is a strong add on an early board and impossible-to-cheap on a late one, which belongs in a listing either way. A digital HDMI install is the biggest premium available on this platform and it preserves the analog output, so the buyer keeps both. And a recap, on this machine, is baseline maintenance rather than a selling point — which brings us to the parts.

Recap and parts

The N64 is not a recap platform the way a Game Gear is. This is the single most useful thing to know before ordering anything. The mainboard uses small surface-mount aluminium electrolytics that age slowly, and a dead N64 is far more often a dead power brick than a bad mainboard capacitor. Recap the mainboard when you see age-related video noise, dot crawl, intermittent glitches or instability — or as belt-and-suspenders on a console you are selling — and not reflexively.

My priority order on a console destined for resale:

1. Test, and if necessary repair, the power brick.
2. Replace the controller stick module.
3. Clean and reseat the cartridge slot; reflow the Multi Out if it is at all loose. Mechanical work, not parts.
4. Replace the three thermal pads.
5. Recap the mainboard only if it is symptomatic.

The mainboard capacitor list is per revision and the differences are real. Eighteen capacitors on -01 through -04, seventeen on -05 through -07, fifteen on -08 through -09-01, with C33 changing value rather than just appearing or disappearing. Do not stuff an early kit into a late board. Pre-kitted “all revisions” kits are supersets, so you finish with leftovers rather than shortages, which is the correct direction to be wrong in. Full per-revision maps are in the data below, cross-checked between two independent community lists and the recreated schematic’s own on-sheet value labels.

The power brick is where the real work is. Several different OEM designs ship under the same NUS-002 number — the Panasonic LSEP family, a Mitsumi design and the “Zebra” boards — and the PAL bricks are a different set again. Read the label and order against it. The per-brick capacitor lists are in the data section, and two rules apply across all of them: use 105 °C radial parts, and do not use low-ESR here, because it causes high inrush (one source calls that out specifically for the largest secondary capacitor).

If the brick has a blown fuse, capacitors alone will not fix it. The fault is usually a shorted bridge rectifier or the primary MOSFET, and the original parts are effectively unobtainable — but tested modern substitutes are documented for each brick type and they are in the parts table below. One PAL-specific fault worth knowing: on the Type M design, two 0603 resistors beside the 3.3 V rectifier fail open under the silicone potting, which presents as a brick that powers up with a rail missing.

Two on-board parts to know about before you order anything. U13, the 78M05, is a stock part you can buy anywhere. U12, the Sharp PQ7VZ5, is obsolete with no documented drop-in, so plan on donor salvage. On a fleet, your worst board is your parts stock.

Getting the picture onto a modern TV

Separate from repair, but it comes up constantly, so here is the full version.

A stock N64 outputs composite and S-video through its Multi Out, and composite is what most people have a cable for. My composite AV-to-HDMI converter handles that directly — its input is composite RCA and the N64 provides composite through any SNES, N64 or GameCube A/V cable, since all three share the connector. It runs off any USB port. What it will not do is make a composite signal look like RGB, and I have not bench-measured its latency yet, so I am not calling it a competition display until I have.

My SCART-to-HDMI converter is the better picture, with a real condition attached: it only pays off on an N64 that has actually been RGB-modded. Feed it a SCART cable carrying composite from a stock N64 and it will work, but you have paid more for the same signal. Feed it RGB from a modded early board and it looks genuinely sharp. If your console is stock and you have no plans to mod it, the composite box is the right answer.

If you would rather buy a console that has already had this work done, everything I restore is in the shop.

The hard data behind all of the above — the per-revision capacitor maps, the power brick lists, the non-capacitor consumables, the connector and chip pinouts, and the measured figures worth probing against — is tabled in the sections that follow. It is organised by board revision, because that is what decides the job. Long per-item notes fold away; expand any of them for the full detail.

Capacitor lists

Per board revision, and here the power bricks get equal billing with the console, because on this machine the brick is the part that actually fails. Read the silkscreen or the brick’s label before you order — the console lists differ by revision and the brick lists differ by manufacturer.

Replacement parts

The non-capacitor parts an N64 repair actually consumes, plus the two that are donor-harvest only.

Chip & connector pinouts

The component library: each chip 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. On this console more cards than usual are deliberately empty — Nintendo published nothing, so where no source I hold gives a pinout, the card says so rather than offering a guess.

Schematic facts