CPUs (And SOCs)

A lot of the great dedicated chess computers (and non-chess computers) used the 6502 CPU: IMO, the Dialog DA14531 (link) is better in many ways:

Cheaper: it costs a little over £1 (£0.50 if you order a thousand of them)

Lighter: less than a third of a gram (you'd barely be able to feel it on your hand)

More capable: the 6502 is just a CPU. The SOC has a lot more functionality, including memory, Bluetooth etc

Power consumption: the SOC runs on milliamps: tiny batteries last a long time

Speed: 16 MHz (the 6502 was originally 1-3 MHz)

Size: See below:





For a little more money, Dialog will sell you much more powerful SOCs than that, but for me, this diddy little SOC is a much better device than the 1975 6502 CPU. However, it wasn't around at that time!

Z80 is 1/2 the speed of a 6502.


From todays point of view it’s uninteresting which cpu.
But it’s important how many different engines you get with one cpu.

And 6502 is a good hotspot.

Z80 v 6502: there's a good discussion here - link.

Summary: 6502 was faster, but a lot of developers preferred the Z80 because it had a much richer instruction set and more registers, making it a lot easier to write and maintain complex software.

Z80 v 6502 can be seen as being similar to a RISC v CISC debate - link. The 6502 is not exactly a RISC processor, but it does have a much smaller instruction set than the Z80. The 8088, which became a big winner when IBM selected it for the PC in 1981, seems to have been a compromise between the two.

Quick note: some of the Z80 advantages over the 6502 don't relate to the instruction set - e.g. extra registers, which, without doubt, reduces the number of memory fetches needed to get a result.

RISC has made a big comeback in recent years in the form of ARM - a world-beating RISC architecture widely used in portable devices due to its low power requirements (CISC computers generally draw their power from wall sockets).

Also, Seymour Cray, a brand name supercomputer engineer, and other supercomputer suppliers, tended to favour RISC: the reasons are complex, but seem to come down to this: supercomputers are complex machines that use a lot of energy, so best make them out of simple components that don't use much energy. Cray also realised that most of the instructions in a CISC are not used very often in numerical programming.

To try to simplify a complex issue, here's a short list of the strengths of each architecture (bearing in mind that I'm not an expert). Looking at this list, I can see why Clive Sinclair chose the Z80 for his ZX80 device, which was big hit in the UK - the first home computer for under £100:


CISC

* smaller programs (especially valuable in the early days of home computers which had very little RAM)

* programs easier to build and maintain (this would only apply to developers working in assembly - but some early chess programmers actually did work in assembly, and at the time, this would have given them a big speed advantage IMO)

* speed: having complex instructions will mean that fewer instructions will need to be executed to get the same result

* memory access is better (some CISC instructions can directly access memory!)

* easier to design and build the rest of the computer (link)

* a compiler will be less complicated (obviously this advantage only applies to people who write compilers)


RISC

* higher raw speed

* lower power consumption (e.g. the SOC featured in the OP which runs on milliamps)

* most instructions completed in a single clock cycle (meaning that the speed of code execution is all down to the skill of the author of the compiler: these days, there are many compilers that are able to produce efficient code)

* probably better when building a system that uses multiple CPUs (e.g. Cray supercomputers). This is a guess on my part - I don't really know - but it is a fact that it's what Seymour Cray used to do