The Pong you could program, possibly: the MOS 7600/7601

UPDATE: It is a microcontroller! Read more.

When people think microprocessors that MOS Technology made, they think of the 6502 and its many derivatives, as one should (which are of course frequent topics on this blog too) — but there might have been another one.

In the nethermists of time when polyester ruled the earth, G-d (or at least Al Alcorn) created the Pong machines. These started out as discrete logic that was hard-wired to play the game, both in the original 1972 Pong cabinet and then miniaturized for the Atari 3659 Pong-in-a-chip that was in the first home Pong console (sold through Sears) in 1975. By greatly reducing the component count Atari's new chip made the console cheaper to produce and assemble, significantly aiding mass production. Here at Floodgap orbiting headquarters we have an original Atari Ultra Pong Doubles with the later C010765, referred to as the "ultimate" Pong with 32 game variations for up to four players, and the last and mightiest of the Atari first-party consoles circa 1977. Still, like the original 1972 Pong, it was nevertheless controlled by hardwired logic; ultimately it just played Pong, and that was it. But that's not what this entry is about.

The Pong console wasn't the first home console; that was of course the 1972 Magnavox Odyssey, developed by Ralph Baer as a side project for defense contractor Sanders Associates, and licensed to Magnavox for sale. (For that matter, it wasn't even the first home Pong console: that was arguably the Universal Research Labs Video Action II, which ran ads for Christmas 1974 in an attempt to sell unused inventory made for ailing Pong licensee Allied Leisure and beat most of the others to market in 1975 by a couple of months.) However, it was the first Pong-in-a-chip, something the other fledgling semiconductor companies had yet to duplicate. Texas Instruments was supposed to be developing a single chip implementation for Magnavox's sequel two-game Odyssey 100, but it wasn't ready until later that year for the Odyssey 200, and the 100 ended up with four chips instead despite its simplified games. But that's also not what this entry is about.

The other thing Magnavox had besides their early market advantage was Ralph Baer's patents, and a plausible legal case. Atari themselves capitulated in 1974, determining they lacked the resources to invalidate his patents, and paying a settlement and licensing them instead. Other vendors followed suit. One of these companies was General Instrument, who devised their own black-and-white Pong-in-a-chip called the AY-3-8500, also in 1975. Ralph Baer was aware of its development from the licensing process and had a previous informal relationship with toy company Coleco's president Arnold Greenberg. This exchange was all business: Baer's patents represented a significant source of income to himself and the company he worked for, and if GI could move a lot of chips there'd be a big piece of the action in it. Coleco became GI's first customer and built the AY-3-8500 into the 1976 Coleco Telstar, which went on to sell roughly a million units. The AY-3-8500 was thus firmly established in the market and even Magnavox used it for subsequent Odyssey consoles to the great chagrin of Texas Instruments, including the 1977 Odyssey 3000 which we also have at Floodgap Orbiting HQ. But that's not what this entry is about either.

The AY-3-8500 naturally had its competitors, and Magnavox-Sanders-et amis got a piece of them too, though the design was so cheap and easy to work with that General Instruments crowded most of them out of the market. Besides TI, National Semiconductor introduced the colour MM-57100N, but despite its capabilities was too expensive to dislodge the market leader (and even its graphics advantage was eroded by the AY-3-8515, which was a bolt-on colour encoder for the 8500 and incorporated into later chips). However, there was one company at that time that was very good at making cheap chips cheaply, and it needed money fast. That company was MOS Technology, and finally that's what this entry is about.

In late 1975 when the Motorola 6800-pin-and-bus-compatible MOS 6501 emerged, a serious threat to the 6800 due to MOS' high manufacturing yields, Motorola was furious and promptly filed suit. Like the lawsuit Magnavox filed against Atari, Motorola had a decent case and well-documented patents, and the discovery process wasn't going well for MOS. The lawsuit prevented the 6501 from ever being sold and consumed much of the small company's capital. While the "lawsuit compatible" 6502 was being reworked before the company ran out of money, in early 1976 MOS management looked for other ways to raise cash with their production technology and noticed the big sales in home video games. Thus was the NTSC MOS 7600 (and its PAL variant, the MOS 7601) born.

What's odd about the MOS 7600/1 is that everyone seems to have a clear idea of what it can do, but there's yet to be any hard proof like an available datasheet to substantiate it. Most of the sites that talk about it (including, I must admit, my entry at the Secret Weapons of Commodore) make reference to the other sites in terms of its specifications, which circularly point at other sites which point back at the originals. In any event, what allegedly made the 7600 unique compared to the TI, NS and GI silicon-Pong designs was that instead of hardwired circuitry it supposedly had a mask ROM containing the game programming and a primitive internal CPU to run it (though it did use internal discrete circuitry for the graphics and audio; more in a moment). This internal ROM is confidently and consistently described as 512 "words," despite no surviving MOS or CSG spec sheet to back that up, nor any idea what the actual word size is (possibly eight bits, but at that time could have been four). On the other hand, given the constraints MOS was working under, it isn't inconceivable that they might have used some of the pre-existing work from the 6501/2 and combined it with a ROM and some internal static RAM and colour video circuitry to get a superior working design up quick. In that sense the mask ROM would have been an advantage, as it could be customized by/for any potential vendor. And MOS knew how to make cheap chips, another obvious advantage in a market with low margins.

Once again, Ralph Baer advised Coleco of the new development, and his involvement in the design of Coleco's next console during 1976 with the 7600 strongly implies MOS had also negotiated a license with Magnavox and its recent new parent company Philips. By November 1976 Commodore Business Machines had bought out MOS Technology to secure their semiconductor supply chain. To keep cashflow as Chuck Peddle persuaded Jack Tramiel to develop an actual 6502 computer, Tramiel continued to sell some of the products MOS was already working on, most notably the in-production KIM-1 but also a game console concept MOS themselves had developed around that other chip. The console became the PAL-only TV Game 2000K, which like the AY-3-8500-based consoles played basic Pong games and had a target practice mode, but did so in colour. It used the PAL 7601. Here's mine, a later manufactured unit from 1977.

This machine didn't come with a manual or the light gun, and the cover over the Styrofoam took some water damage, but otherwise this is the kit.

Inside are four paddles, all separatable by long wires, though two are permanently affixed. The console is labelled in English, with a couple rookie misspellings ("AMATUER"), and German. The paddles connect to jacks on the front panel.

The TV cord comes out the rear, along with 9V power and a plug for the light gun.

Battery compartment (6 AA batteries) and backplate. The patents molded into the back aren't Commodore's: they're actually the Sanders Associates-Magnavox patents. Among others listed the 3,829,095 patent was Ralph Baer's famous 1970 "METHOD OF EMPLOYING A TELEVISION RECEIVER FOR ACTIVE PARTICIPATION," and the presence of this case-molded listing is more proof that Commodore was a full Magnavox licensee. However, the patent numbers listed are the American and Japanese ones — which are both NTSC countries. It is entirely possible that this and its sibling unit (to follow) were only released in Europe to avoid the expense of FCC certification, which was quite strict in those days, even though their presence also suggests that (at least) the 2000K was originally considered for American sale as well.

The backside of the logic board. You can see that there are only a few ICs, plus an internal speaker; the 7600/1 didn't transmit sound to the television set, which was typical of most home Pongs along with my Atari Ultra Pong. This makes them easy to test: power them on and start a game, and see if it makes noise. If it does, the basics are undoubtedly working.

Rather than risk damage to the logic board or to its soldered wire connections trying to dig it out and turn it over, I got out my flexible USB camera probe and wormed it underneath to get a look at the main core. It's silkscreened with MPS 7600 001 3177, meaning a plastic DIP, variant 001 (the basic Pong-and-target chip), manufactured 31st week of 1977; this particular unit probably came from near the end of the overall production run.

The 7601-001 plays four games. Here they are, from left to right on the console's dial.

Tennis, with a blue background. The score numbers disappear when the serve button is pressed and play begins. Notice the orange and blue paddles/bats which oddly get a bit desaturated in the main playfield. If four paddles are connected and the "option" switch is set to four, then the console switches to doubles play. The game ends once one side/player gets 21 points.

Unlike the original Atari Pong and its eight-zone angles of reflection on each paddle, the collision routine in MOS' Pong is a less-sophisticated five zones.

I'm including this picture purely for the shapes of the numbers. Remember it for later.

The second game is football (soccer to us Yanks), on a green background. Each side has offense and defense controlled by the same paddle in the two-player game; in the four-player game, they are controlled separately. This game also plays to 21 points.

The third game is squash (racquetball) on a red background. There are supposed to be three walls on the top, left and bottom, but the left wall doesn't reproduce well in this screenshot. The dashed line denotes the playfield; the bats are restricted to the right side. Uniquely, if the "option" switch is set to four players, that actually enables a single player mode (the left score tallies misses, up to 21, and the right score the number of volleys before missing). The two-player mode is shown here.

Finally, skeet shooting with a single "block" target. This is strictly single player. The left number is the number of shots (you get 15) and the right is the number of hits.

The TV Game 2000K is most likely the very first 7600-based game system to emerge, so consider screenshots like this one to be representative. Compare them with others a little later on.

While the famously stingy Tramiel no doubt didn't like the perpetual license fees Philips-Magnavox was charging, there were already buyers for the chip in sufficient numbers to make its manufacture profitable. The next one to market was probably the 1976 Allied Leisure (yes, that Allied Leisure, see above) Name Of The Game and Name Of The Game II systems. The NOTG was one of the few 7600/1 systems to use the MCS ceramic DIP form of the chip (as opposed to the more common plastic DIP we just saw), which is strong evidence of its early use, and they may well be the first system of any kind that used the NTSC 7600. Coleco's upcoming new console, however, stalled late in prototyping, apparently to expand the games available. Assuming the mask ROM concept really existed, though, this was eminently possible with the 7600: all MOS/CSG had to do was write a new program and rewire the display circuitry. None of these additional 7600 variations ever emerged in any other product, not even Commodore's, so it's likely they were originally designed specifically for Coleco.

Meanwhile, several new consoles emerged with the chip, most of them around mid-to-late 1977. One of these licensees was Radofin, makers of cheap Hong Kong electronics crap, later to go down in home computing infamy as the designer of the doomed Mattel Aquarius ("The System for the Seventies!"). Using the 7600 they developed a console called the SC8000 (or, depending, the "SC Eight Thousand") and released it through Kmart in the United States that year.

I got one in fairly thrashed condition, promising "8 TV GAMES Full Color !"

There was no manual, but it did come with a light gun and all four paddles. However, it was monstrously dirty to boot, so it's time to clean it up for presentation. Hey, we haven't had a Refurb Weekend in awhile, right?

Disassembling the main unit. Here's a good example of how redonkulous this machine's design is: you see those two interesting big sliders? You'd think they'd do something, right?

Well, here's the underside of them, and here's where they fit. Yes, that's empty space. There's nothing beneath them. They are absolutely bereft of function. But they're clicky, and have a great feel, so in the interest of being charitable we'll just say that Radofin was ahead of its time with giving you not one but two fidget, uh, sliders.

We'll get to the buttons and switches which do do something in a minute. The game selector is the big circular switch in the middle of the FCC-mandated Faraday cage.

If we pry off the lid of the can (not without some effort), we find a MPS 7600-001 3177. This is practically the doppelgänger of our MPS 7601-001 we saw in the 2000K, but there's an important difference we'll see when we get to the actual games.

Before and after with the paddles. Much better!

The refurbished unit. Virtually everything on the 2000K has an analogue here: we see the various game option switches, buttons for power, serve and reset, and the nice thumping-huge circular game selector switch. But this console has clicky do-nothing sliders as well! Take that, Tramiel, you cheapskate!

That said, it's actually a very smart looking box, but I honestly don't understand the sliders. They may be a holdover from the almost identical Radofin SC4000, another Kmart Pong (though not sure with what internals) that had the slide paddles built into the console — though on separate sides, so it wasn't precisely the same design, and they weren't "clicky." Perhaps this was a later concept which was never realized and it was cheaper to leave them on, which would be totally on brand for Radofin. [UPDATE: On lobste.rs, someone suggested these were for keeping track of matches won. That seems plausible.]

Unlike the six AA batteries for the 2000K, the SC8000 wants 6 big old D cells. Notice the different patent listing on the back for the Euro-patents instead of the American and Japanese ones, probably because Radofin was using the same molds internationally for all of their localized products, but the sticker has the American Kmart logo and was "Manufactured in Hong Kong for S. S. Kresge Company [then US Kmart's parent], Troy, Michigan."

Additionally, this machine carries the FCC Part 15 labelling on a sticker and is an NTSC unit, so it really does seem to have been sold State-side. The wired-in paddles connect to the rear. The two 5-pin DIN ports can either take both external paddles, or one of them is used for the target gun. There is a 9V power jack as well.

Let's power it up. These are the same games, but they are in different order on the selector, and despite being manufactured the exact same week as our 2000K's chip, the chip in this console has some unusual differences.

Tennis, also on a blue background, but a darker (almost richer) shade and with rather large scoring numerals. The video quality is surprisingly good for a unit this obviously cheap and has much less bleed than the 2000K's visuals.

Potentially to appeal to the Great White North audience, though never known to be actually played by the McKenzies (hosers), the soccer game was unexpectedly renamed to hockey. The incongruous brown background certainly could suggest blood mixed with ice, so who knows. Gameplay also differs: in the two-player mode shown here, each side effectively only has a goalie.

Racquetball-squash stayed mostly the same except for the absence of the play line, here shown in single player mode.

Finally, skeet shooting, identical except for the score numbers, of course.

File these differences away while we look at Commodore's other 7600-series console specific to Europe, the TV Game 3000H (don't ask me to explain this naming scheme), which appears to have come out in 1977.

Like the 2000K, the 3000H is bilingual in both German and English (mine came with two manuals, one in each language, and the other side of the box is printed in German too). This particular unit came with the optional AC wall-wart which does me little good in the land of 110 volts. The gun was "extra" and not in the box, but it does have all four paddles.

Even if it didn't it would be impossible to lose all four, because player 1's paddle is integrated into the machine (I guess someone at Commodore noticed Radofin's slider idea? however, Radofin was hardly the only one). This is a rather obvious cost-cutting move compared to the better-appointed 2000K and makes it obnoxious to play with because you'd have to "grab" the whole console to move the slider in a typical fashion. The other switches are the same.

For some reason the patents were not reproduced on the 3000H's bottom case. It also uses six AA batteries.

Like the SC8000, there are two 5-pin DIN ports on the back, both occupied by the paddles for the four player games, or one with the light gun. It seems quite possible that the Radofin's pistol might work with the 3000H, but I'm not really game to try on these relatively rare machines.

The cost reduction in the 3000H was more than just skin deep: internally it used a cut-down 7601 that MOS labeled as the 5601, and which is only ever known to exist in this console. Unlike the 7601 the 5601 actually required a heatsink, which doesn't speak highly for the design. The 5601s I've seen, including the two of my own, seem to lack date codes entirely and are simply labeled "M5601" or "M5601-XXX" with variant sequences.

The 5601's depredations included fewer overall colours and no colour backgrounds at all, and it was solely produced in the Pong-and-target variant. On the other hand, it also has rather cleaner if less interesting video, and would have played a lot better on a black-and-white television. This is tennis. Modulo the colours, all the games are otherwise the same.

Football-soccer.

Squash-racquetball.

Skeet-shooting.

It does seem that the 3000H was produced in greater numbers than the 2000K, and apparently sold relatively well due to its lower cost, but the bottom was already starting to fall out of the Pong market and the 3000H became Commodore's last Pong machine.

Eventually Coleco's own 7600 system finally arrived tardy in 1978, named the Coleco Telstar Arcade.

The Telstar Arcade is without a doubt one of the strangest console form factors ever made. Instead of multiple external controllers, the console incorporates them into the legs of its unusual triangular case: dial paddles and the light gun, but also "slam" buttons and an actual steering wheel and gear shift (!). This unit is not entirely intact and lacks the proper holster and the "Coleco" hub logo in the steering wheel, but it otherwise functions. As we expect by now, it also has a power switch, a skill switch, a rotary game selector and a reset button. Uniquely there is no battery option and it came with a 9V wall-wart.

The gun is the only controller in the standard configuration that is detachable. If you pull out its plug, it's got a 5-pin DIN connector like on the 3000H and SC8000, but you can't connect an extra paddle there. Instead, a "remote receptacle" with an extra pin is used to connect an extra pair of paddles (the manual says it's "used with other cartridges" but it looks like the only one that did is #2). There is no four-player switch; presumably the extra pins tell the console extra players are extra present.

I'm not quite sure I'd like to shoot a real gun like this, though. Notice anything weird about the barrel?

Continuing the form factor's theme, the cartridges themselves are triangular too, with a Battlestar Galactica-y textured sparkly finish. To make sure the connection mates well, a plastic tab holds the cartridge firmly in position and the flat silver lip near my hand goes under a ridge. You slide the lip under the ridge first and rotate the cartridge down at an angle, engaging the tab which clicks it flat into place. To remove it, you have to pull the tab free (away from the ridge) to let the cartridge pop up. The contacts on the cartridge themselves are also interesting: several are shorted together.

If we take the bottom off (the Sanders-Magnavox patents don't appear on this unit either; perhaps they stopped caring about it by then as long as they got paid), we see the inevitable internal speaker and a few circuit boards for the switches and the cartridge connector. Like the other units, there are only a handful of small internal chips.

On the circuit board between the steering wheel and the gun, we see just discrete logic, such as a Motorola MC14528B dual monostable multivibrator (still manufactured!) and an MC14015B dual 4-bit static shift register (also still manufactured). These chips all bear 1977 date codes, and most are 20th week.

For the other big circuit board with the control panel, rather than demount it and risk damage, we'll just get the camera probe back out and give the unit a colonoscopy. Why, it's just like the flex sigs I did as a resident, except with rather less gas and Go-Lytely.

That doesn't look like bowel wall.

It's almost as old as I am!

Some actual chips in view, more discrete logic. This is a Texas Instruments CD4069UB CMOS hex inverter. It has a date code of 50th week 1976.

This one is harder to make out, but appears to be a TI 4050 hex buffer.

None of these chips have the form factor we now recognize as the MOS 7600, and none of them are enough on their own to yield even one game of sufficient sophistication as described in the manual, let alone several of them.

That's because the 7600 is in the cartridges themselves. Coleco eventually manufactured four and all Telstar Arcades shipped with #1, making it the most common cartridge, but the only other one I have is #3. Let's start with #1.

The cartridge opens by removing a screw roughly at each vertex to reveal a small circuit board (with apparently a yuuuuuge ground plane) and a single socketed chip, here an MPS 7600-002 manufactured 34th week 1977. The springs on those posts ensure that the board and the contacts beneath it stay firmly pressed down against the mating contacts on the console.

In fact, I should issue a note of caution to the curious at this point. When I cracked this cartridge open and put it back together, the console went haywire when I tried to play it again. After some panicked internal inspection, it turned out to be because I had not properly aligned the contacts in their little aperture, meaning it didn't mate with the console's contacts when the cartridge was installed. There is only the smallest of a notch to guide you and the springs don't make it easy to reassemble. Don't open these if you don't have to (there is literally nothing to see other than the chip).

The manual for cartridge #1 is part of the manual for the console. You choose the game with the rotary dial. The first mode is, incredibly, a racing game.

You are the yellow car. The left number is your miles driven and the right is the time remaining. There is no actual finish line; the only goal is to drive as far as you can. The car only starts in low gear (think Pole Position), but you'll need high gear for faster speeds and to accrue points more quickly.

Each crash costs you a variable amount of time and you have to shift back into low gear to get the car started again. The chip is capable of registering a score of up to 199, after which it wraps back around to 100 (!), suggesting that the +100 flag is a separate bit.

Tennis, on, I guess, a clay court. The need to use the dial paddles right on the console is a little inconvenient but Coleco embroiders the situation with "slam" buttons that put a little "English on the ball." In practice the effect looks largely random but it certainly does add a little zest to the game. Scoring goes to 21.

Finally, quick draw (the skeet shooting mode). It plays to 21 instead of 15, and uses an "outlaw" sprite instead of just a plain block.

There was no fourth position for this cartridge; it's just the same as the skeet shooting in the third position. In fact, when this machine was originally produced, there wasn't even a fourth cartridge (there was eventually). But we do have #3, so let's try that.

This chip is a MPS 7600-004 from 46th week 1977. Although the chip is later than our cartridge #1, the board is earlier (Rev A compared to Rev C). Coleco made a standalone console with this chip called the Telstar Gemini. It has two pinball modes and two skeet modes. Unfortunately I don't have the manual for this one but gameplay is pretty easy to figure out.

The flyer and box for this cartridge imply that "Bonus Pinball" was supposed to be on a pink background, but this looks more teal. It's a weird combination of Pong and pinball where the olive coloured elements can move around using the paddles (both are active). As you hit things you get points.

It's very easy to get a lot of points.

Very easy (it rolls over at 99, more credence to the theory that the +100 was an independent bit and not a carry).

"Deluxe Pinball" is a minor variation with different targets, on a blue background. This screenshot demonstrates the "authentic flipper action" (my actual Stern and Williams pinball machines beg to differ), except that both flippers move with the left slam button (the right slam is the ball launch), and if you position the bats just right you can pretty much rack up infinite points in either game without ever touching the flippers at all.

"Shooting Gallery" and "Shoot the Bear," which is important, because a bear this disfigured should indeed be put out of its misery. The same misshapen sprite is used in both games.

Cartridge #2 is hockey, tennis, handball and "target," and sounds exactly like what's in the SC8000, so it's undoubtedly got an MOS 7600-001 (but I suspect it uses the score generator circuitry of the other games, not the oversize numbers in the Radofin). The manual advertises it comes with the "remote controls" and supports up to four players.

That means cartridge #4 must have one last variant, 7600-003, containing Naval Battle, Speed Ball and Blast Away. These games had no corresponding analogue in any other system. Naval Battle subdivided each score into single digits and gave two players each a submarine to defend their home base from an oncoming destroyer (and each other). The score tracked torpedoes left to fire and the number of base hits (win at 9). Speed Ball was an enhanced Pong using a central block off which the ball could carom, and played to 21; the slam buttons here allowed a player to catch the ball on their paddle when properly timed. Finally, Blast Away was a two-player target mode allowing players to take turns shooting a spaceship, and the only known 2-player shooting mode on any 7600/1 system. Like Naval Battle, each score was split into single digits, here for number of hits and number of shots. Each player got only five shots at the ship.

It's worth comparing the other 7600 game variants with competing ideas from TI, GI and even Atari (prior to the VCS 2600, that is). GI introduced game chips also in 1977 for racing, a sub hunt game (where you hunt the subs instead of controlling them), a Breakout-Pong combination, a two-player tank battler that compared favourably to the VCS' Combat title and a motorcycle racer. These CPUs-in-a-cartridge made up the game options for GI's "Gimini Economy" (Gimini 8600) concept. A few of these chips made their way into Coleco and Atari systems but mostly served clone systems in Europe where the Atari VCS was initially less common. GI's interest in further such special purpose chips waned when the Gimini 8900 concept (based on the CP1610 general purpose CPU) got a big design win as the Mattel Intellivision, and the other Gimini concepts faded away. As a transitional product Atari had their own Breakout-Pong combination called Video Pinball; the apparent presence of RAM in these systems suggests they used a primitive microcontroller of their own.

Texas Instruments, typical of their modus operandi in that era, had the most over-engineered solution. Instead of all-in-one chips, TI offered separate ICs for scoring, various types of graphics (walls, balls, stick figures, cars, rockets, etc.), game logic chips and position generators. You combined them together to make a complete game, which was fairly flexible for circuit designers but caused an expensive parts count for manufacturing, and the chip line ended up being a flop. The chip with the most lasting impact from this series was the SN76477, the programmable complex sound generator, which evolved into the later and much more popular SN76489 digital complex sound generator.

After all that, are we any closer to determining whether the 7600 was actually programmable? Unfortunately, no. The NTSC 7600-001 in the SC8000 and the PAL 7601-001 in the TVG 2000K, despite being rather different chips manufactured at almost the same time, nevertheless have their sizeable differences exist entirely in the video generation circuitry which was already widely reported as hardwired. Even though the SC8000 "hockey" is missing the other two bats, its most notable gameplay difference, this could simply be done by having the video circuit never draw them. Likewise, the 5601, as deprived as it is, differs only from the 7601 in how the screen is drawn and not how the games are played. If the mask ROM were present, it appears to be the same in all of these systems, which negates the advantage of having it.

As for the variants in the Coleco Telstar Arcade and Gemini, these games seem no more complex than those in the General Instrument family we just discussed. Gimini Economy 8600 consoles have a 32-pin connector between (in the console) a colour encoder such as the AY-3-8615, a clock crystal, an RF modulator and the game controls, and (in the cartridge) one of the compatible chips or even the AY-3-8610 which had multiple games in a single IC. If this sounds very much like the Telstar Arcade, you get a sparkly triangular star. The data sheets and example schematics for the Gimini 8600 design and games do survive and none of them mention mask ROM or indeed any internal processing elements in the game chips of any kind. This doesn't prove they're not there, but it sounds like something that would have been called attention to at the time. Furthermore, there is no RAM or ROM on the console side either which would suggest the presence of an embedded processor element, and it wouldn't make sense for RAM or ROM to be duplicated in the individual cartridges themselves due to the resulting higher part count (none of the documented cartridge designs shows it anyway). After all, this was supposed to be their economy design!

I think one can thus only conclude that the reputed programmability of the MOS 7600/1 is completely unsubstantiated by any presently (un)available documentation. That doesn't mean it isn't there, but Occam's razor would say that with other systems of similar capability from the era using hardcoded gameplay logic, the 7600/1 series probably does so as well. I've updated the entry at the Secret Weapons of Commodore accordingly to reference this, and I think the only way we're going to find out what's actually inside is to get a duff one and send it to Ken Shirriff for die photography. Ken, are you interested in solving the mystery? Or does someone actually have proof of how they work?

As postscript, the 7600 was used in a few other systems, including the Allied Leisure NOTG series (as mentioned), the APF TV FUN Sportsarama, the Dayya Marume 2000, the Granada ColorSport VIII, the Hanimex 888/888G TV Scoreboard (also sold by Tandy Radio Shack), the Roberts Rally X and the Venture Electronics Video Sports. However, the Pong clone market rapidly became unprofitable with new second-generation game systems like the 2600 and Intellivision offering greater capabilities, and Commodore quit producing the chip around 1978. It no longer appears in my 1982 Commodore-MOS catalogue.