The Commodore 65

These machines are, for an unreleased prototype, very well-known. Now, if only they were more common ...

.jpg Image of the Commodore 65 (53K)
.gif Image of the Commodore 65 (31K)
Andre Kaesmacher's Commodore 65 Page

Complete Views of the Commodore 65 (.jpg, courtesy Moise Sunda, except Board with RAM Expansion)
Portrait (60K) | Rear Ports (18K) | Rear Ports Enlargement (48K) | Side Ports (23K) | Expansion Door (12K) | Version 2B Mainboard (59K) | Version 5 Mainboard Enlargement (136K) | Version 2B Board Closeups: CPU ("Victor") VIC-III ("Bill"), DMAGIC (80K), SIDs, ROMs, ELMER (56K), Board Notations (24K), Board with RAM Expansion (51K) | Power Supplies: 110V (40K), 240V (44K)

Screenshots of the Commodore 65 (.jpg, courtesy Moise Sunda)
65 Mode (42K) | 64 Mode (64K) | Using the Burn-In Board (50K)

Peripheral Cards, Development Boards (.jpg, courtesy Moise Sunda except RAM Expansion)
RAM Expansion (41K, courtesy Riccardo Rubini) | Burn-In Board (68K) | Widget Board (48K)

C65 In The News
A mention of the system in ZZap!64 December 1991 (117K .jpg courtesy Iain Black).

Jesper Sortoft also reports an appearance of what was probably an early incarnation of the 65 in Datormagazin #3 1988, as the "C64D." Although the article indicates it was scheduled to appear at CeBIT, Commodore Sweden denied its existence; the probability of it being an early 65 concept is increased by the case and 3.5" disk drive.

Gazette 9/89 also has a mention of the 65, referred to as the "64GS" which was never an official name (see the real 64GS). This very developed entry captures most of the specifications the final incarnation of the 65 would support, but notes Commodore claimed that the IEC serial bus would no longer be supported, on which Gazette correctly calls bull ("uh, excuse me," says editor Lance Elko, "pass the bologna"). It states the MSRP was approximately $300-$350.

aka C64DX, C64DX Development System (not the DX-64)
Introduced Never officially. Prototypes escaped development hell when Commodore was liquidated in 1994.
Hardware Commodore engineer Fred Bowen states that the CPU is a CSG65CE02 (CSG = Commodore Semiconductor Group). However, the actual chip in the models that escaped is a CSG 4510 ("Victor" chip, named for designer Victor Andrade), a modified 65CE02 carrying two 6526 CIA's. The clock speed is known for certain to be 3.54MHz. 128K RAM, max 8MB(!), 128K ROM (Commodore 64 ROMs, BASIC V10, a monitor and CBM DOS); integrated 1581; RAM expansion port; cartridge port (weirdly, based on the +4/16's port, not on the 64's), and a useless RF port that doesn't work on any of the models that have escaped. 6551-like ("6511" in some docs is probably a typo) UART up to MIDI data rate, with built-in DMA (powered by custom 51-DMAGIC 390957-01) and blitter.
Graphics and Sound Powered by CSG 4567 (VIC-III) ("Bill" chip, named for designer Bill Gardei); standard 64 modes (40x25, 320x200, etc.) plus 80x25 text (with blink, bold, underline); true bitplane graphics: VGA-style 320x200x256 (from a palette of 4096 colours), 640x200x256, 640x400x16, 1280x200x16, 1280x400x4; reportedly can display Amiga OCS IFF. Special facility for address resolution (Display Address Translator [DAT]) of that wacky 64 pixel coordinate system. PAL only, no NTSC (but the RGB output works just fine with the 1084S). Double SID chips for stereo sound (8580R5).
Eventual Fate Very interesting story -- the project was eventually canned in 1991 by then-chairman Irving Gould. Despite this, when Commodore was liquidated in 1994, prototypes escaped, eventually sold by U.S. companies Grapevine Group and Software Hut. Motherboards also exist. Estimates for the number that eventually made it to the outside world range from as low as 50 (clearly wrong; serial number 000062 has been independently verified) to as high as 2,000, including complete units, motherboards, etc. A working 65 is the Holy Grail of the Commodore 8-bit world these days. :-)

Comments
Commodore had very big ideas for its new 8-bit flagship, as evidenced in the specification document written by Fred Bowen back in March 1991:

"The purpose of the C65 is to modernize and revitalize the 10 year old C64 market while still taking advantage of the developed base of C64 software. To accomplish this, the C65 will provide a C64 mode of operation, offering a reasonable degree of C64 software compability and a moderate degree of add-on hardware and peripheral compability. Compability can be sacrificed when it impedes enhanced functionality and expandability, much as the C64 sacrificed VIC-20 compability.

"It is anticipated that the many features and capabilities of the new C65 mode will quickly attract the attention of developers and consumers alike, thereby revitalizing the low-end home computer market. The C65 incorporates features that are normally found on today's more expensive machens, continuing the Commodore tradition of maximizing performance for the price. The C65 will provide many new opportunities for third party software and hardware developers, including telecommunications, video, instrument control (including MIDI), and productivity as well as entertainment software."

There are several boards known, the most recent being version 5 (the enlargement of the version 5 board appears virtually complete); the creators were nice enough to leave their initials and marks on it with the copyright message. (At least one of them ["FISH" -- Terry Fischer] is known to have worked on the 128's board as well; Gardei and Andrade also appear, along with system engineer Paul Lassa.) Several versions of the chips themselves are also known, but not all chip revisions are interchangeable. According to some testing done by Riccardo Rubini, revision 3, 4 and higher boards use Victor revision 4510R5, which is not pin-compatible with the previous 4510R3 and R4 (which *are* interchangeable), and is unfortunately undocumented. There is also a puzzling documentation issue with Bill revisions; there is both a 4567R7 and 4567R5 documented, both of which seem to be mutually incompatible. On Riccardo's revision 5 mainboard, however, was a 4567P, which seems to at least partially work in revision 2A and 2B motherboards (which would indicate it's part of the R5 family). So far, no samples of the allegedly incompatible 4567R7 have surfaced.

For that matter, there are multiple versions of the 65 firmware based on board revision, and some programs are known to work on some revisions, but not others. According to Anders Carlsson and Riccardo, the following board versions have the following firmware (not all are confirmed):

* Rev 1.1 + 910??? ROM
* Rev 2A  + 910114?ROM
* Rev 2B  + 910111 ROM
* Rev 3   + 910801 ROM (910601 ?)
* Rev 4   + 910802 ROM
* Rev 5   + 911001 ROM

The rear ports of the 65 are, from left to right, a disk drive connector (more on this in a second), 8-pin standard Commodore video, non-functional channel selector switch and RF output, RGBI (identical to the 128's), stereo outputs, user port (identical to the VIC-20, 64 and 128's), standard serial port and the cartridge port (labeled "expansion port", a misnomer because RAM expansion doesn't go there: it connects through a door on the bottom of the unit and plugs into the mainboard separately in the style of the, surprise surprise, Amiga). The disk drive connector appears to be a 8-pin Mini-DIN on both boards and is where the also-unreleased 1565 disk drive connected (version 0.9B.910710 of the 65 ROM actually looks for it and reports, invariably, "NO C1565 DRIVE").

BASIC 10 has many similarities to BASIC 7.0 on the 128, plus a few other interesting keywords like GENLOCK for video sync; MOUSE/RMOUSE for mouse manipulation; TYPE (which displays a SEQ file to screen -- very handy!); FIND/CHANGE for BASIC find and replace; FRE(0, 1, and 2) for bank 0 (BASIC program text) and bank 1 (variables) free space, as well as number of expansion banks available, respectively; PALETTE for defining colours and a mysterious DMA command for DMA operations. BASIC 10.0 clearly was unfinished; PAINT, PASTE, CUT, LOCATE, SCALE, WIDTH, QUIT, OFF, SET and VIEWPORT all give ?UNIMPLEMENTED COMMAND ERRORs. Like BASIC 7.0, the function keys are programmable, but the assignments are intriguing: F1 toggles 40/80 columns, F2 erases everything below the cursor, F3 is "DIR", F4 is 'DIR "*=PRG"', F5 puts the cursor at the beginning of the prior word, F6 is "KEY 6" (another programming gap unfilled?), F7 moves to the next word, F8 is "MONITOR" and HELP is, naturally, HELP.

The 4510 CPU -- so thus the 65CE02 -- has some nice tricks in it. One of the most useful is its ability to move zero page around (apparently

LDA #page
TAB ; this is a real opcode in the 65CE02

does the trick, and TBA brings it back). The stack pointer is 16-bit, so TSX moves the low byte to X, and TSY moves the high byte to Y. Guess what the opposites are.

The VIC-III apparently has its own MMU. This means, most likely, that the CIAs no longer have their old job of fooling the vid chip into seeing different tracts of memory. In fact, in 64 mode, location $0000 is provided by the VIC-III, not by the 4510. The memory mapping/banking scheme of the 65 is quite complicated because the 4510 is still at its heart addressing just 64K. The reason for the segment/offset system being as bizarre as it is, says Fred Bowen, is because the CPU mapper restricts programmers to a single offset for each 32K half of a 64K segment. For a ROM to bank in another chunk of memory with a different offset, it must be in the other 32K half of that segment. As a consequence, all the standard banking layouts for the 65 put I/O and Kernal interfaces in $C000-$DFFF to force currently executing code to be in one 32K zone only and allowing it to bank the other half around at whim. How this plays out in practice is that the OS (Kernal, etc.) lives in the top 32K with the interfaces in the usual position, and the lower 32K is the offset that is actively being swapped around. For the monitor, the OS will bank out BASIC and bank in the monitor; for BASIC, the reverse. None of these higher- level applications interface with each other. The exception is DOS; for DOS services the Kernal overlays itself with the DOS and puts the DOS RAM addressing space in the lower 32K. To do this trick I guess there must be a snippet of code in the other 32K to do the actual Kernal-DOS switch. Can anyone elaborate?

There are hooks in the $C000-$DFFF memory space for an REC (marked as optional in the map), but because the cartridge port is not 64-compatible, you can't just plug in a 1750 and expect its REC to map there. In most cases, though, chips do map to the same locations (like the on-chip CIAs still map to $D[CD]00, just like on the real 64). The 65 also appears to have two busses, D and E, with 64 kilobytes of RAM on each. The VIC-III can access the D-bus while the CPU accesses the E-bus, and then they can swap around, making the whole 8MB address space effectively both chip RAM and fast RAM. This might also imply that the 65 has no more badlines, which probably spells doom for custom 64-mode hacks like FLI.

Because the 65's DOS is based on the old IEEE 8250 disk drive (PET-era!), the DOS subsystem can only deal with two drives -- and that includes the internal one. However, programmer Dennis Jarvis says he preferred to use the 8250 DOS as a base because it was "a whole lot less buggy." Amongst his improvements for the 65 are better wildcard handling such as DIR A*S?G, which is not possible on the older disk drives.

The RAM expansion board, as shown on Todd Elliott's unit (above), sits on a robin's-egg blue PCB and plugs into the bottom through an Amiga-like door. The board reads "COPYRIGHT 1990/COMMODORE ELECTRONICS AWK 313384/C65 RAM EXPANDER ASSY 313381" and the initials of the designers: "PL/BG/JA/VA/HD/NINES". By the way, Todd's is also one of the earliest C65s extant and certainly the earliest unit I've ever seen; his unit is serial number 000008, and several of the chips are EPROMs and hand-labeled to boot.

Moise has several interesting peripheral and testing cards that should be commented on. Both appear to connect through the bottom expansion port door and it isn't known if they were intended as production peripherals (likely not, at least for the burn-in card) or just in-house equipment. The widget board, although it is rather simplistic in appearance, seems especially useful for debugging and testing as it has all the bus lines exposed for easy access.

The name 64DX is easily confused with the DX-64, a double-drive version of the SX-64, because according to Doug Cotton, Commodore was penny-pinching and decided to actually use the stickers (and therefore the name) from the ultra-rare DX-64 unit for the 65 (deciding it would make as good a model number as any). None of the actually extant 65s, however, have the DX sticker on them.

Moise's Burn-In Board doesn't seem to work on all units, according to Riccardo. Even though the screenshot is of Moise's 2B unit, 2B motherboards will crash right after sprite testing (assumed the same on 2A, although Riccardo's board is admittedly not 100%). Only on revision 4 and 5 motherboards does it seem to work correctly (revision 3 and 1.1 boards yet to be tested).

The 65's RGB connector pinout is as below, with pin 1 at the top left of the computer's male socket, and pin 9 at the bottom right:

 Pin
 ===
  1  GND
  2  n/c
  3  R
  4  G
  5  B
  6  n/c
  7  composite sync
  8  horizontal sync
  9  vertical sync

Interestingly, the 4510 survived the 65 project; it is also found in the (similarly never released) Amiga CDTV CR. After Commodore's demise, 4510 designer Victor Andrade went on to design the K7 for AMD.