The Atari 2600 CPU was a nondescript MOS 6507 chip. Neither Intel nor Motorola had a CPU chip in the early 1970s that could meet Atari’s price point. MOS Technology didn’t have one either, but they asked Atari what they could afford. Then they made one.
The 6507 is so nondescript, some of them don’t even have the number “6507” anywhere on them.
Four Motorola refugees found a home at scrappy MOS Technology. Chuck Peddle, Bill Mensch, and two other colleagues had worked on Motorola’s 6800 CPU, which eventually found use in the Radio Shack TRS-80 Color Computer. Chuck Peddle thought the 6800 was too expensive. Motorola disagreed, and wanted to spend its resources on projects like the 68000, which went on to power mid-1980s computers like the first Macintosh and Amiga.
So Chuck Peddle and his team defected to MOS. There, they designed a cut-down 8-bit CPU they could sell for $25 in 1975. No one else sold a CPU for less than $179 at the time.
Too expensive for Atari
In the mid 1970s, Atari started developing a next-generation console that would use plug-in game cartridges. It had teams using Intel, Motorola, and MOS chips. Only MOS came close to hitting the price point Atari needed, but the $25 MOS 6502 was still too expensive to become the Atari 2600 CPU. Atari needed a $12 CPU. MOS, bleeding cash from losing a lawsuit to Motorola, desperately needed business.
So MOS took the 6502 CPU and removed address lines and interrupts to fit it into a smaller, cheaper 28-pin package. The resulting cut-down CPU could only address 8K of memory, but that was still a lot of memory in 1977. Brian Bagnall covered the story of Atari and MOS in his book On the Edge: The Spectacular Rise and Fall of Commodore.
Atari reserved 4K of the memory for I/O space and 4K for program space. The 6507 ran at 1.19 MHz in the Atari 2600.
But other than the reduced memory space and lack of certain interrupts, the 6507 acted just like its older brother. It still used the same instruction set, so when you disassemble Atari 2600 code, it looks a lot like machine code from 6502-based devices like the Apple II, Commodore 64, and Nintendo NES.
The 6507’s philosophy is a lot like Intel’s philosophy with its 386SX CPU from 1990.
MOS sold out to Commodore in November 1976 in another marriage of convenience. MOS was running out of money and Commodore needed a reliable chip supplier. Or maybe I should say it needed a chip supplier it could count on. There’s a subtle difference, and MOS was perhaps more of the latter. Commodore founder and CEO Jack Tramiel, of course, ended up running Atari.
As the 1980s wore on, MOS had its hands full supplying chips to Commodore. Fortunately for Atari, MOS licensed the design to companies like Synertek, Rockwell, UMC, and NTE. Many later 2600s contain 6507s from these companies. All are interchangeable. Licensing CPU designs to alternative sources was common in the 1970s and 1980s. Some manufacturers required it as a hedge against supply chain issues. Having multiple sources ensured a steady supply of chips so they could keep making systems.
When 8 bits isn’t 8 bits
Some people are surprised the Atari 2600 contains an 8-bit CPU, as it’s a much more primitive system than the 8-bit Nintendo NES. Later systems were able to address a full 64K of memory at a time, and contained much more capable (and expensive) video chips than the Atari 2600’s Stella chip. The 6507 held the Atari 2600 back, but everything about the Atari 2600 was designed to hit a price point. $199 was a lot of money in 1977 but it didn’t buy a lot of computing power then. In 1977, the Apple II cost $1298 for just the base computer, with no storage.