Why the Commodore 64 power brick fails, and how it takes the computer with it

The Commodore 64 does not make its own power. A brick on the floor beside it supplies two things through a 7-pin DIN plug: regulated 5 V DC for the chips, and 9 V AC from a transformer winding, which serves as the reference for the time-of-day clock and, once rectified, runs the SID and the Datasette motor. C64-Wiki puts the healthy range for the 5 V line at 4.95 to 5.10 V and notes that the machine uses that line almost directly to power its chips. According to the repair technician Ray Carlsen, neither the computer nor the original brick has a fuse on it.

What is inside the brick

Carlsen repaired consumer electronics from 1964 and retired in 2009 from a technician's job at the University of Washington in Seattle. He describes a linear design. A step-down transformer, a rectifier, electrolytic filter capacitors and a regulator IC turn mains into 5 V. Commodore rated it at 1.5 A, but Carlsen writes that it runs hot inside with a load of less than 1 A, because the original black and white bricks are potted in epoxy. The epoxy keeps fingers off the mains wiring, keeps the heat in, and makes the supply impossible to repair. Commodore later shipped C64 supplies in vented cases held together with screws, which could be opened; Carlsen says their electrical design was the same.

The protection that does exist does not cover the dangerous fault. The regulator folds back its output if the load passes 1.5 A, and a thermal fuse in the transformer cuts the mains if the transformer overheats. Nothing limits the voltage.

Three ways it fails

Carlsen lists the ways a linear supply dies, and two of them spare the computer. When the filter capacitors dry out, ripple appears on the DC output and the voltage sags; the computer refuses to start or crashes, but nothing is damaged. When a component fails open, the output drops to zero. The most common failure of all, he writes, is a C64 whose red power light stays dark because the 5 V is low or missing.

The third mode kills. If the regulator shorts internally, its input passes straight through to its output: more than 11 V, by Carlsen's figure, on a line meant to carry 5. He calls it a silent killer, because nothing tells the owner that the supply was to blame. The README for bwack's C64 Saver adds a second cause, broken solder joints on the regulator. C64-Wiki describes a slower drift toward the same end, saying the original supplies tend to produce higher voltages as they wear out, and it recommends measuring the 5 V output once a year.

Heat can be the trigger. Carlsen reports bricks that always worked when cold and put out excessive voltage once warm; they damaged several computers because their owners did not realise the fault would repeat. Breadbox64 makes the same point: a supply that measures correctly before use can still fail after it warms up.

Why the RAM goes first

C64-Wiki says that in most cases where a supply killed a C64, the 5 V line had risen above 6 V, and that the RAM is usually damaged first. Carlsen agrees on the order: the eight 4164 RAM chips in an early C64 are the most sensitive parts on the board. A repairer on the Lemon64 forum described the same sequence in March 2020: RAM first, then the ROMs and, on the long boards, the PLA, with the SID going after longer exposure.

The exact limit is less settled than it is usually written. Carlsen's 2014 and 2019 notes call 5.5 V the RAM chips' "absolute maximum rating"; his 2005 tester notes say 6 V. The data sheets for two 4164-type parts of the period put 5.5 V at the top of the recommended operating range, 5 V plus 10 percent, and set the absolute maximum on the supply pin higher: 6 V for Texas Instruments' TMS4164 and 7.5 V for Samsung's KM4164B. Either way, 5.5 V is the edge of the range the chips were designed to run in. A shorted regulator delivers about twice that.

Carlsen writes that shorted RAM always gives a blank screen and often gets hot to the touch, while healthy RAM runs barely warm. Shorted RAM usually points to a failing supply, which has to be checked before the repaired computer is switched on again.

Recognising a failing supply

These are the bands C64-Wiki uses for a reading of the 5 V line.

A quick reading is not enough: Carlsen's tester notes warn that a failing supply may still read 5 V with nothing connected. He loads each output with a 12 V lamp or a power resistor, runs the supply for about half an hour until the case is warm, and watches the meter. Below about 4.9 V suggests failing capacitors; a reading that climbs past about 5.2 V, loaded or not, means the regulator is going. The 9 V AC output is either present or absent, and C64-Wiki says it seldom fails. RetroMake's project Test a Commodore 64 power supply before it reaches the computer sets out the full procedure from these sources.

Replacement supplies

Electroware, based in Tychy, Poland, sells its Commodore C64 PSU for EUR 48.99 including VAT at c64psu.com, listed as in stock when RetroMake checked on 26 October 2026. The company says the current model, its eleventh revision, delivers 5 V at up to 4.0 A from an industrial AC/DC module and 9 V AC at 1.1 A from a linear transformer, with overvoltage protection on the 5 V line and fuses on both the AC and DC sides. Its history shows why such claims deserve checking. In a 2018 review, breadbox64.com measured 4.73 V or less from an early unit. Electroware told the reviewer that its crowbar circuit triggered at around 5.2 to 5.3 V and that the output aimed low on purpose; in 2019 the company said it had added a trimmer and tuned the output to 5 V on a C64C board.

Keelog, in Wroclaw, lists supplies for the C64 alone from USD 65.99 (EUR 57.99) to USD 101.99 (EUR 88.99), depending on the case and on whether an OLED display shows the output voltages. Keelog's specifications list a 5 V, 2.0 A switching module with an anti-ripple filter, a transformer for the 9 V AC, overvoltage and overcurrent protection, and fuses on both sides.

Carlsen builds supplies to order. The C64 OS buyer's guide quotes his description of the range, from simple wall-adapter designs for a C64 alone to units that power several devices at once, and says orders go by email. His site lists no prices. It has moved to portcommodore.com/rcarlsen, where he notes that his old TDS email address is being discontinued; the old site at personalpages.tds.net did not respond when RetroMake tried it on 26 October 2026. One of Carlsen's supplies read 4.98 V DC and 9.42 V AC in breadbox64's 2018 tests.

Protectors for an original brick

Carlsen's Computer Saver, which he makes as an internal module, an in-line cable and a small box, watches the 5 V line with a 4.7 V Zener diode and a transistor and drops a relay to open the line when the voltage exceeds about 5.4 V. It is not a crowbar circuit; it opens the line, and it resets itself when the fault clears. He sets the trip point that high because some supplies jump briefly when the computer is switched off, and calls 5.4 V still under the RAM's rating. The Console5 TechWiki publishes a version to build at home, credited to his design, whose parts include a 5 V relay rated for at least 2 A, a 4.7 V Zener and a 1.5 A fast-blow fuse. His site gives no price for the Saver.

The C64 Saver 2, by the designer known as bwack, is open hardware under the CERN Open Hardware Licence v1.2, published on GitHub. The base board, through-hole apart from one MOSFET and also offered as an SMD design, fits a Hammond 1551G case. An optional add-on board brings a small OLED display, measurement on both outputs, an overcurrent cut-out and a choice of automatic reset after a trip. It can also detect the 100 Hz ripple of a supply with dry capacitors, which would otherwise make a simple saver switch on and off. The current versions, Saver 2.6 and add-on board 1.4, are dated 2 April 2023, and the README warns that the latest versions of the savers are not tested. It is a design to build; RetroMake could not confirm a price for assembled units.

A saver guards against one fault. It does nothing for a supply whose output sags, and Carlsen writes that a relay buzzing continuously signals an overload or a supply whose output is falling. Both are reasons to stop and measure. The cheapest protection is still a meter: a reading taken once a year, under load and with the brick warm, separates a tired supply from one about to destroy a set of RAM chips.