Commodore 64 | Code Like It's 198x

At the Winter Consumer Electronics Show in Las Vegas in January 1982, Commodore showed two new colour computers built on the same new chips. One was the Ultimax, a membrane-keyboard games machine at $149.95. The other was the Commodore 64: 64K of RAM, a full-stroke keyboard and a price of $595. Compute! reported that Commodore “stole the show”, and that its marketing chief, Kit Spencer, “boldly predicted that Commodore would sell more computers in 1982 than were sold by all manufacturers in 1981.”

Commodore could price the 64 that way because it owned the chip maker, MOS Technology. The engineers there had spent 1981 designing a video chip and a sound chip with no computer to put them in. In late November 1981 Jack Tramiel decided they would go into a 64K home computer, betting, in Al Charpentier’s words, “that by the time we were ready to produce a product, 64K RAMs would be cheap enough for us to use.” The team had working prototypes by the new year. The first machines cost Commodore about $135 to build.

Shipping began in the late summer of 1982. Compute! heard that 12,000 left the factory in the first two weeks of full production, and that the problem after that was “building them fast enough to meet demand”. Commodore kept building them until it went into liquidation in 1994. Guinness World Records lists the C64 as the best-selling desktop computer model. How many were sold is less certain than that title suggests; the sales section below sets out the evidence.

Fast facts

CPU: MOS 6510, a 6502 with an on-chip I/O port at locations 0 and 1. The port switches the BASIC ROM, KERNAL ROM and I/O area in and out of the memory map and drives the cassette motor. The clock is 1.023 MHz on NTSC machines and 0.985 MHz on PAL ones, derived by the VIC-II from its pixel clock.
RAM: 64K, with 20K of ROM and 4K of I/O banked over it under software control. BASIC starts with 38,911 bytes free.
Graphics: VIC-II (6567 in NTSC machines, 6569 in PAL). 40×25 text, 320×200 high-resolution or 160×200 multicolour bitmaps, 16 colours, and eight hardware sprites of 24×21 pixels, with smooth scrolling, collision detection and raster interrupts.
Colour RAM: a separate memory four bits wide, at $D800–$DBE7 (55296 to 56295). Its 1000 locations hold one colour for each character cell on the 40×25 screen.
Sound: SID (6581; the 8580 on later boards). Three voices, each with four waveforms and its own ADSR envelope, plus a multimode filter, ring modulation, oscillator sync and a master volume from 0 to 15.
I/O: two CIA 6526 chips for the keyboard, joysticks, timers and serial lines.
Storage: the Datassette tape recorder, rated by Commodore at 300 baud with every program saved twice; the 1541 5.25-inch floppy drive, 170K per disk; cartridges.

Built as a synthesiser and a games chip

The chip project began at MOS in January 1981, with no computer in mind. In Charpentier’s words, “we decided to produce state-of-the-art video and sound chips for the world’s next great video game.” Bob Yannes designed the SID. He later explained: “I’d worked with synthesizers, and I wanted a chip that was a music synthesizer.” The problem with game sound, he told IEEE Spectrum, was that “there is either full volume or no volume at all. That really makes music impossible.” The SID gave each of its three voices an envelope (attack, decay, sustain, release), so a note could swell, pluck or fade the way an instrument does, and a filter could be swept across the sound while it played.

The VIC-II’s sprites did a similar job for graphics. Compute!’s first report on the machine explained them to “Atari fans” as the same idea as Atari’s player-missile graphics. Each sprite is a 24×21-pixel shape that moves over the background with its own colour, and the chip reports when a sprite touches another sprite or the background.

The chips shared a limit that would shape the whole platform: there was no software to drive them. The machine shipped with Commodore’s 8K BASIC 2.0, the same BASIC as the PET and VIC-20, which has no commands for sprites, sound or graphics. Yannes said later that “there was never any intention that the system would have built-in software to control the graphics or the sound.” Reviewing the machine in October 1982, Your Computer found it “let down by a rather limited Basic” but “an excellent machine which can be highly recommended.” Owners who wanted sprites or music from BASIC used POKE to write values straight into the chips’ registers, or bought an extension such as Simons’ BASIC.

The raster beam as a clock

On the Spectrum the defining constraint is spatial: two colours per character cell. On the C64 it is about time. Much of what C64 programmers did is organised around where the TV’s electron beam has got to.

The VIC-II can raise an interrupt when the beam reaches a chosen scanline, so a program can change the chip’s registers partway down the screen. It can switch the background colour at line 100, change character set halfway down, or move a sprite that has already been drawn down to a lower part of the screen and draw it again. That last trick, sprite multiplexing, is how games show more than eight sprites.

The cost is that the VIC-II and the 6510 share one bus. A PAL line lasts 63 CPU cycles; NTSC lines last 64 or 65, depending on the chip revision. On the first pixel line of each text row, called a badline, the VIC-II has to fetch that row’s 40 character codes. It stops the processor for 40 to 43 cycles, leaving it about 20 on a PAL line. Each sprite shown on a line takes 2 more cycles, and the chip warns the processor up to 3 cycles before it takes the bus. So the budget on one PAL line looks roughly like this:

Line	Cycles left for the 6510 (approximate)
Ordinary display line, no sprites	63
Ordinary display line, all 8 sprites	about 44
Badline, no sprites	about 20
Badline, all 8 sprites	almost none

Badlines come every eighth line inside the display area. The vertical scroll register decides which lines they are, so a program that scrolls the screen moves its own badlines. Code that runs comfortably for seven lines stalls on the eighth. Timing-critical code has to know where the badlines fall, or turn the screen off and take the cycles back.

Writers at the time worked out the total. The Home Computer Advanced Course told its readers that the VIC-II’s memory fetches “slow up” the 6510 “by as much as 15 to 20 per cent”. Because those pauses are unpredictable, it recommended blanking the screen with POKE 53265,11 during tape loading and saving, then restoring it with POKE 53265,27. The cycle counts above are where that percentage comes from.

The price war

Commodore cut the price quickly. By June 1983 Compute! was reporting the 64 at $399, or $299 for buyers who traded in an old computer or games console for a $100 rebate. At the Summer CES that year Commodore cut its wholesale price from about $360 to $200, and Compute! expected shop prices of “$250 or less by mid- to late summer”. Ahoy! reported that 1.2 million were sold in 1983.

The cuts did not keep Jack Tramiel at Commodore. He resigned as president on 13 January 1984, days after that year’s Winter CES, and within months several more senior executives had followed him out.

In Britain the machine was first shown at the Commodore Computer Show in London in June 1982, with Your Computer quoting a price of £400 and availability from October. By that autumn Commodore was advertising it at £299 plus VAT, and dealers were listing it at £345.

How many were sold

The number most often repeated is 17 million, and it is the weakest. Figures quoted in the period grew over the years, and later estimates disagree:

Ahoy! (April 1984): 1.2 million sold in 1983.
Compute!’s Gazette (August 1986): an installed base of “more than two-and-a-half million 64s”.
INFO (September/October 1988): “Commodore has sold 10 million 8-bit computers. There are now eight million C64s and two million C128s.”
Guinness World Records quotes Jack Tramiel’s estimate of 22 to 30 million, and 17 million as “Commodore’s official estimate”.
Michael Steil (2011) estimated about 12.5 million by analysing motherboard serial numbers.
Dave McMurtrie of the Commodore International Historical Society (2021) worked from Commodore’s annual reports and the internal figures of its former treasurer, Don Greenbaum. He found about 12.35 million C64s and C128s combined, and about 10.6 million C64s. He also found that Commodore’s 1993 annual report, often given as the source of the 17 million figure, does not contain it.

Even the lowest of these is a very large number for a single model, which is why the Guinness title survives the uncertainty. A single figure quoted without its source is best read as one estimate among several.

Why it mattered

Price. Owning the chip maker let Commodore cut the price hard and fast. Within eighteen months of a $595 launch the C64 was heading towards $200 in American shops, a price that put a 64K colour computer in homes that would otherwise have had none.
Hardware you could reach. With no graphics or sound commands in BASIC, C64 owners met the VIC-II and SID registers almost straight away. The Programmer’s Reference Guide documents every one of them.
A music culture. The SID gave composers such as Rob Hubbard, Martin Galway, Ben Daglish and Jeroen Tel an instrument with its own recognisable sound. Their game soundtracks became part of why people bought the games.
The demoscene. The C64 was one of the main machines of the demoscene, whose techniques for pushing the VIC-II past its specification, such as FLI and opening the borders, are still being developed.

Models and variants

C64 (1982): the original “breadbin”. Compute! noted that it used the VIC-20’s case and keyboard in a different colour, so visitors to Commodore’s stand “had trouble at first picking it out of rows of VIC-20s”.
SX-64 (1984): a portable “luggable” version with a small built-in colour screen and a 1541 drive, shown as the Executive 64 in 1983. K-Power listed it in December 1984 at 20 lb and $995.
C64C (1986): announced at the Chicago CES in June 1986. It has a lower cream case in the style of the Commodore 128 and was sold with the GEOS graphical operating system. Compute!’s Gazette reported that “the new 64C is the same 64 internally”, and that Commodore had twice stopped C64 production in late 1985 only to restart it because orders kept coming. Later boards used the redesigned 8580 SID, whose filter sounds noticeably different from the 6581’s. Which SID a C64C has depends on its board revision, not its case.
C64GS (1990): a cartridge-only games console version with no keyboard, launched for Christmas 1990. By 1991 ZZAP!64 was calling it a failure.

Quirks worth knowing

The 1541 is slow. The C64 talks to its disk drive over a serial bus designed for VIC-20 compatibility. Charpentier explained that the team fixed the fault in the serial shift register for the 6526, “but we couldn’t use it, because the disk drive had to be compatible.” Transfers are therefore done one bit at a time in software. Cartridges such as The Final Cartridge advertised disk loading “6 times faster”, and many commercial games brought their own fast loaders.
Tape is slower still. Commodore’s own format saves every program twice, so it can check for errors, and runs at a nominal 300 baud. Turbo loaders dropped the second copy and tightened the timing; The Final Cartridge claimed tape loading “10 times faster”.
Two colours per character cell in the standard text and bitmap modes, as on the Spectrum. Multicolour mode gives four colours per cell at half the horizontal resolution. Sprites carry their own colours, which is one reason they do so much of the work in C64 games.

Modern legacy

Emulation: VICE is the standard emulator; MiSTer offers an FPGA recreation.
Hardware: the Ultimate 64 replaces the motherboard with an FPGA board, the 1541 Ultimate replaces the disk drive with a cartridge that reads SD cards, and the MEGA65 is a new machine built as a successor rather than a copy.
Community: demo parties such as Datastorm, X and Revision still premiere new C64 productions.

Sources

Compute!, issue 22, March 1982, p. 26: David D. Thornburg’s report on the January 1982 Winter CES; the Ultimax at $149.95, the Commodore 64 at $595, Kit Spencer’s prediction.
Compute!, issue 27, August 1982, p. 60: Tom R. Halfhill, “A First Look At The Commodore 64”; 8K PET BASIC 2.0, sprites compared to Atari player-missile graphics, the VIC-20 case, 170K 1541 drives.
Compute!, issue 30, November 1982, p. 8: 12,000 shipped in the first two weeks of full release.
Compute!, issue 37, June 1983, p. 6: $399, or $299 with the $100 trade-in rebate.
Compute!, issue 39, August 1983, pp. 35-36: Summer CES report; wholesale price cut from about $360 to $200; the Executive 64 portable.
Tekla S. Perry and Paul Wallich, “Design case history: the Commodore 64”, IEEE Spectrum, vol. 22 no. 3, March 1985, pp. 48-58 (read in the version republished at spectrum.ieee.org): the chip project from January 1981, the November 1981 decision, the 64K RAM bet, the $135 build cost, Yannes on the SID and on built-in software, Charpentier on the serial bus.
Commodore Magazine, vol. 3 no. 2, April 1982, pp. 104-105: “Unsung Heroes”, the MOS/Commodore engineering group, naming Bob Yannes and “Design Manager Al Charpentier”.
Commodore 64 Programmer’s Reference Guide (Commodore, 1982): colour memory (p. 101), sprites of 24 by 21 dots (p. 131), SID volume 0 to 15 (p. 186), the 6510 I/O port and the 20K of ROM (p. 260). Commodore 64 User’s Guide (1984 printing): the start-up screen, “38911 BASIC BYTES FREE”.
Christian Bauer, The MOS 6567/6569 video controller (VIC-II) and its application in the Commodore 64 (1996): clock rates, cycles per line, badlines and sprite fetch timing.
The Home Computer Advanced Course, issue 61, p. 1217: VIC-II DMA costs 15 to 20 per cent; screen blanking during tape I/O.
Commodore, C2N/1530/1531 Datassette service manual (preliminary, October 1984): 300 baud, programs saved twice.
Your Computer, July 1982, p. 16: UK showing at the Commodore Computer Show, 3-5 June 1982, £400, due in October. Your Computer, October 1982, pp. 32-33: review, “let down by a rather limited Basic”; p. 105: Commodore advertisement, “£299 plus VAT”.
Ahoy!, issue 4, April 1984, p. 6: Tramiel resigned on 13 January 1984; 1.2 million C64s sold in 1983. Ahoy!, issue 8, August 1984, p. 7: the continuing departures of Commodore executives.
Compute!’s Gazette, issue 38, August 1986, p. 23: the 64C announced at CES in June 1986, “the same 64 internally”; production stopped and restarted in late 1985; more than 2.5 million 64s installed.
INFO, issue 22, September/October 1988, p. 35: “eight million C64s and two million C128s”.
Compute!’s Gazette, issue 29, November 1985, p. 69: The Final Cartridge advertisement, disk turbo “6 times faster”, tape turbo “10 times faster”.
K-Power, December 1984: SX-64, 20 lb, $995.
ZZAP!64, issue 67, November 1990: the C64GS launched for Christmas; issue 78, October 1991: “a failure like the C64GS”.
Guinness World Records, “Best-selling desktop computer” (web, secondary): production dates; Tramiel’s and Commodore’s estimates.
Michael Steil, “How many Commodore 64 computers were really sold?”, pagetable.com, 1 February 2011 (web, secondary).
Dave McMurtrie, “How many C64 (and C128) were actually sold?”, Commodore International Historical Society, 5 July 2021 (web, secondary).
C64-Wiki, “Motherboard” (web, secondary): the 6581 on the early C64C board, the 8580 on the later short board.