13

The Fairchild Channel F, released in 1976, was the first modern game console, in the sense of being the first one to accept games as software, rather than just modular configuration switches.

Looking at the specifications, the thing that jumps out at me is that (in addition to 64 bytes of working RAM) it has a 2K write-only frame buffer:

https://en.wikipedia.org/wiki/Fairchild_Channel_F#Technical_specifications

The above both directly describes the frame buffer, and gives the resolution, from one which one can verify that 128x64x2/8 = 2K.

This is particularly notable in that the Atari 2600, released a year later, did not have a frame buffer.

How did Fairchild provide such a thing within the technology and cost constraints?

Looking at the ads in the back of Byte magazine for December 1976, I don't see any chips advertised for any price that could easily provide that. There are e.g. 256x4 static rams advertised, but it would take 16 of those to make 2K, which would surely blow the budget, and looking at photographs of the circuit board, I don't see any large arrays of identical chips.

So how did they do it? What chips were used?

Improve this question
1
  • 2K frame buffer? Quite ahead of its time, it seems.
    – user3840170
    Commented Aug 5, 2023 at 20:34

3 Answers 3

Reset to default
19

Channel F schematics page 2/3 : Video

ChannelF uses 4 MK4096 4k * 1bit chips (4 middle-left rectangles in the schematics).

The framebuffer is write-only, the CPU cannot directly access it through memory accesses. The CPU has no address bus anyway, just IO ports.

Framebuffer updates are performed by setting the line, column and colour of the pixel to change on IO ports (BTW, these IO ports are shared with the joysticks to reduce cost).

Improve this answer
10
  • 1
    Wait, the CPU doesn't have an address bus? How does that work? :O
    – ConnieMnemonic
    Commented Aug 7, 2023 at 14:20
  • 2
    @ConnieMnemonic Instructions are fed from a separate chip containing those.
    – pipe
    Commented Aug 7, 2023 at 17:33
  • 1
    @ConnieMnemonic It's quite special, read the Channel F CPU "F8", Fairchild 3850 manual. The instruction set is... odd, the CPU has no address bus, it doesn't have a "Program Counter" inside nor index registers. these registers are instead in the ROM memory chips.
    – Grabul
    Commented Aug 7, 2023 at 17:57
  • 1
    @ConnieMnemonic: If one imagines a system as having one RAM, one ROM, and one CPU, the cost of integrating a counter and a bit of other logic into the RAM and ROM chip would be less than the cost of the extra lead wires needed to support an address bus, especially if one considers that the CPU itself wouldn't need that circuitry. Pretty soon, people realized that having microprocessors as all-in-one units which combine address generation with everything else had a lot of advantages, but before the division wasn't so sharp. If one looks at the CDP1802 processor and thinks about the way...
    – supercat
    Commented Aug 7, 2023 at 18:42
  • 1
    ...instructions work, it feels like the accumulator (D register) lives in another universe from the 16 numbered registers R0-R15, and it turns out there's a reason for that. Early versions of the design put address generation as well as the storage for the 16 registers in a separate chip from everything else.
    – supercat
    Commented Aug 7, 2023 at 18:44
10

The Fairchild Channel F, released in 1976, was the first modern game console

Let's restrict that to first successful in large numbers.

Looking at the specifications, the thing that jumps out at me is that (in addition to 64 bytes of working RAM) it has a 2K write-only frame buffer.

Picture has to go somewhere.

This is particularly notable in that the Atari 2600, released a year later, did not have a frame buffer.

The 2600 was intended to be a cost cutter. Saving four DRAM chips goes a great length toward that goal.

I don't see any chips advertised for any price that could easily provide that.

The Channel F design uses four Mostek MK4027 (or MK4015), a 1976 introduced improved version of the all classic 1973 MK4096 DRAM, each providing 4 KiBit for a total of 2x8 KiBit or 2 KiByte.

Improve this answer
9
  • 1
    Wikipedia says the Channel F was first with a microprocessor and first with ROM cartridges, but sold less than some of its predecessors, so I don't see your first line. The Atari 2600 was more expensive at launch than the Channel F, and demonstrated that the picture did not, in fact, have to go anywhere besides direct to the screen.
    – prosfilaes
    Commented Aug 7, 2023 at 21:40
  • 1
    @prosfilaes Well, if you read history at whole, you'll note that the Channel F was based on a prior 8080 design Fairchild bought and modified for their own CPU. Next, price difference was about 10%, not exactly a lot. More important, cost cutting is quite different from price cutting. Last, what relevance has it how a picture is made?
    – Raffzahn
    Commented Aug 7, 2023 at 22:19
  • 1
    Prototypes aren't usually counted. I don't know what you mean by "a cost cutter" then, as every single commercial product is designed to reduce construction cost to increase profit. Lastly, "Picture has to go somewhere." seems to be a bit snarky, since at the same time your answer admits the Atari 2600 didn't need the DRAM.
    – prosfilaes
    Commented Aug 8, 2023 at 1:57
  • 1
    It's a bit snarky, but I hesitate to call it wrong. I mean, I guess in the CRT days you could sync on the movement of the electron gun and just sort of send the right voltage at the right time to get the result you want for each pixel. Presumably modern displays/interface protocols more-or-less require the use of a framebuffer though? 8.3 million pixels gotta go somewhere.
    – aroth
    Commented Aug 8, 2023 at 5:49
  • 2
    @aroth That's how the Atari 2600 works, so that was possible for the Channel F.
    – prosfilaes
    Commented Aug 8, 2023 at 11:49
10

There were two main kinds of memory chips in 1976: dynamic RAM and static RAM. A static RAM used two active transistors and two passive pullups to hold each bit, plus two more active transistors to support reading and writing (one is needed to write ones, and one is needed to write zerores; either or both can be used for reading). Dynamic RAM, by contrast, held each bit in a combination of a transistor and a capacitor. A dynamic RAM chip could hold about four times as many bits as a static RAM of the same physical dimension, but would generally require more circuitry to interface it, including circuitry to ensure that no row goes more than a millisecond or so without at least one bit on it being accessed.

The RIOT chip on the Atari 2600 had 1024 bits (128x8) of RAM internally, which is 1/4 as much as each of the chips used on the Fairchild. While using four DRAM chips may have been "generous", I think the 128x64 resolution in four colors was probably perceived as the minimum anyone would want in a video game system (though the CDP 1802-based RCA Studio II had a horizontal resolution of 64 pixels black an white, and was typically used with a vertical resolution of 32 pixels, using half of its 512x8 static RAM as a display buffer).

While 2K of even DRAM was a fair chunk of memory, the fact that the Fairchild could use DRAM rather than SRAM probably cut the per-bit cost by a factor of four compared with systems like the Atari 2600. Incidentally, the reason the Atari 400 was able to include such a generous amount of memory compared with e.g. the VIC-20 is that the Atari 400 used DRAM while the VIC-20 used static RAM.

Improve this answer
2
  • 3
    Indeed. The cost difference between SRAM and DRAM persists to this day, and many machines like Apple II employed smart hardware tricks in order to be able to use the cheaper DRAM chips until separate memory controller chips became common.
    – jpa
    Commented Aug 6, 2023 at 17:01
  • 1
    @jpa: The cost difference persists, but there are many kinds of DRAM and some variations in SRAM as well. In 1976, there were only two practical forms of RAM memory whose price per byte was within an order of magnitude of each other. From what I understand, a RAM built out of bipolar chips could offer better performance than a MOS-based RAM, but I think a 64-bit bipolar RAM cost as much as a 1024-bit static RAM.
    – supercat
    Commented Aug 7, 2023 at 18:37

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .