Way back in 2002, Jim Turley mentioned that about 14% of all CPUs sold were 4-bit CPUs, while about 8% of all CPUs sold were 32-bit CPUs. (Most people I know were surprised that any 4-bit CPUs were still being made, much less that they were doing so well).

It's 2011 now -- Do 4-bit CPUs still outsell 32-bit CPUs and 64-bit CPUs combined in unit volume? Where could I go to look up the latest numbers for sales by unit volume? What are the top websites and magazines to learn more about modern 4-bit CPUs and their development tools?

  • 1
    \$\begingroup\$ what the heck is a 4-bit cpu!? \$\endgroup\$ – vicatcu Apr 13 '11 at 2:51
  • \$\begingroup\$ Dose a 4-bit CPU truly have only 16 possible commands? \$\endgroup\$ – Faken Apr 13 '11 at 4:39
  • 2
    \$\begingroup\$ @vicatcu: The Intel 4004 was a well known 4-bitter. \$\endgroup\$ – mgkrebbs Apr 13 '11 at 5:19
  • \$\begingroup\$ @Faken The bit size is related more strongly with data and register size than the instruction length. The PIC mcus are an example of this. Further, many processors employ variable instruction length. In the case of a 4 bit processor, you could have 12 of the instructions be the very common instructions, and use 4 instructions to indicate that the next instruction is a special instruction. In this way you would have 12 four bit instructions, then 4 pages of 16 more instructions, giving you a total of 60 instructions. There'd be 12 one nibble instructions, and 48 two nibble instructions. \$\endgroup\$ – Adam Davis Apr 13 '11 at 12:42
  • \$\begingroup\$ Though in typical usage, 4 bit processors really do have very, very few instructions because they are meant to require very little silicon real estate, and so you might find a four bit processor that only has 16 instructions. Technically you only need a handful of instructions to implement a Turing machine. \$\endgroup\$ – Adam Davis Apr 13 '11 at 12:44

The reason that 4 bit CPUs beat out the high end processors at the time was primarily due to their use in watches and clocks. A 4 bit CPU, running in BCD mode (ie, each 4 bits was one 0-9 digit), is very optimal for clock and timing applications, and in volume they are nearly as cheap as transistors. They are used in many other very low cost, very high volume applications.

I don't know whether they still beat 32 bit CPUs, however the trend has been going up steeply for 32 bit processors due to cell phones and portable computing devices. I can't imagine that 4 bit CPUs are going up at the same rate.

64 bit CPUs are still largely related to computers, and are probably still low volume enough that they have not yet eclipsed 4 bit CPU volume.

  • 3
    \$\begingroup\$ They are also very popular in the super cheap IR remotes you get with almost every piece of electronics. \$\endgroup\$ – Mark Apr 14 '11 at 5:21
  • 1
    \$\begingroup\$ Also: 4-bit CPUs running in BCD mode are often used in calculators -- such as the HP Saturn (microprocessor) used in some HP graphing calculators. \$\endgroup\$ – davidcary Apr 14 '11 at 16:41
  • \$\begingroup\$ "very optimal" haha \$\endgroup\$ – vicatcu Apr 15 '11 at 23:55

I doubt it. 32-bit has gained enormously in market share the last 5 years. More and more designs that in the past would have been done in 8-bit are now being done by 32-bitters, mainly ARM. Due to technological progression a 32-bit RISC controller doesn't need more real estate than an 8-bit CISC. That translates in lower cost.
Same for 4-bitters. Most 4-bit controllers are older designs being produced in an older process (larger feature size).

  • 2
    \$\begingroup\$ I still have a love for building < $1 PICs into projects. 8 bits will never die and likely still beats 32 bit installations by far I suspect/guess. \$\endgroup\$ – kenny Apr 13 '11 at 11:58
  • 1
    \$\begingroup\$ @kenny - IIRC NXP has an ARM controller for < $1 too. That's just the point: 32-bit is affordable (more so than 16-bit like 68HC12 ever has been, for instance). But I agree that 8-bitters will be around for quite a while still, especially in smaller packages. I'm not sure if you can fit an ARM in a SOT-23 package. \$\endgroup\$ – stevenvh Apr 13 '11 at 13:55
  • \$\begingroup\$ Thanks, I've been looking at the NXP Arms though I guess mostly the high end ones, since I have seen any that cheap. My fav language is C# and tinyclr.com looks interesting.... Hopefully someone will port that to a lower end NXP. \$\endgroup\$ – kenny Apr 13 '11 at 17:23
  • \$\begingroup\$ @self - not a SOT-23 package, but comparable in size: LPC1102. 16 pins (balls) on a 2.17 x 2.32 mm WLCSP16! \$\endgroup\$ – stevenvh Apr 27 '12 at 14:03

That 2002 statistic might have been a bit surprising, but it wasn't some two bit factoid from USAToday. The author was once an editor for the Microprocessor Report.

If you think back to the technology and prices of the mid-1990's (when many design decisions affecting volumes in the early 2000's were made), a lot of embedded applications were simple and well-suited to 4-bit and 8-bit processors. (Heck, it's still true today for most of the "invisible" processors in use - like the LCD thermostat, or the microwave, or the smart dimming dome light in your car.)

The problem with 16-bit and 32-bit processors back then was that it was needlessly more expensive to provide memory for them. RAM was not cheap back then. And wider RAM was much more expensive for the same capacity. (In fact, the early PC's were quasi-16-bit machines. They had an 8-bit external memory bus.)

Fast forward to this decade, and one key change is that the newer embedded processors have plenty of embedded RAM on-board; thanks in large part to improved semiconductor processes.

Without the external RAM penalty, it's just as easy to grab a 32-bit processors for a new design. And the volumes are there that you don't pay much more for the 32-bit. In fact, for bang-per-buck, the older 8-bit processors are awful for new designs. And, I can't even imagine anyone today would even bother with datasheets for a 4-bit processor.

So, have 4-bit processors died? Given that even soft core processors are 8-bits, I'd say yes. The fun question today is what the split is between 8-bits, 32-bits, and 64-bits.

Circling back to the original question - I used to see "processor yearbooks" that detailed processor offerings from different manufacturers - and they were broken into groups by processor bit size, and whether they were MCU or CPU's. I haven't seen one of those things lately -- I think there's far more players in the market today, many of them from Asia-centric companies that have little or no sales presence in the U.S. Additionally, "processors" may be hidden inside FPGA's so that it would be hard to count them.


The size of the chip has less to do with its word-length than it has to do with its architecture. For example, an 8-bit CISC-like machine would probably suck up more real-estate than a 32-bit RISC-like machine.

As others have commented, 4-bit CPUs are still being sold and still find application in all kinds of things from air-conditioning units through to microwave ovens and maybe some toasters. However, they will be slowly phased out as the cost of 32-bit machines are now as low as the smaller machines (due to economies of scale).

So, if you are working on a new design, it would be a good idea to build it on a newer chip. The slight difference in cost today, may well disappear in a few years time.


Your Answer

By clicking "Post Your Answer", you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.