I dunno. The 6502 has been a $2 part for a long time but needs RAM and some glue logic, for a similar price you can get an AVR-8 [1] or ESP-32 [2] and get some RAM and GPIO.

[1] faster, more registers than the IBM 360, << 64k RAM

[2] much faster, 32bit, >> 64k RAM

65C02s are $8 now. That didn't stop me buying one when I was stuck at home during COVID. And a 6809 too.

But forget AVR. Yeah, for a buck or so the ATTiny85 was my go-to small MCU five years ago, and the $5 328 for bigger tasks.

But for the last three years both can be replaced by a 48 MHz 32 bit RISC-V CH32V003 for $0.10 for the 8 pin package (like ATTiny85, and also no external components needed) and $0.20 for the 20 pin package with basically the same number of GPIOs as the 328. At 2k RAM and 16K flash it's the same RAM and a little less flash than the ATMega328 -- but not as much as you'd think as RISC-V handles 16 and 32 bit values and pointers sooo much better.

And now you have the CH32V002/4/5/6 with enhanced CPU and more RAM and/or flash -- up to 8K rAM and 62K flash on the 006 -- and still for around the $0.10-$0.20 price

https://www.lcsc.com/product-detail/C42431288.html

Hi Bruce! If you make it back to the states we'll have to drink a beer and wax poetic about the 6809. Do you know if anyone ever implement the embedded RISC-V profile in hardware? Not everything I do on small systems needs a 48MHz 32-bit. But if I could get away with a low I/O count, why not use the $0.10 part? Also pretty sure I saw 8051 based SoCs going for $2. I bet if you looked hard enough you could find something like a 6502 for about the same price.

There's probably no reason not to get some of the CH32VXXX's to play with. Every now and again I have an application that needs very low power and I'm happy to spring for an MSP430. But every time I buy an MSP430, TI EoLs the specific model I bought.

Heeey, how's the Cruz treating you? If it still is.

I don't know why you'd ever want to pay a cent more for a 6502 or 8051 or AVR than for a RISC-V or ARM (e.g. Puya PY32F002A). Especially when the CH32V002/4/6 run on anything from 2V to 5V (plus a margin) which is pretty rare, and they don't need any external components.

I don't know whether the M6809 designers were the first to ever analyse a body of real software to find instruction and addressing mode frequencies and the distribution of immediates in order to optimise the encoding of a new ISA -- in a way that the 8086 people clearly didn't [1], but I think they were the first to publish about it, and I was fascinated by their BYTE articles at the time.

MSP430 is also a fun ISA. I just wish they were cheaper, and the cheap ones has more than 512 bytes of RAM. FRAM is funky. I also loooove the instruction encoding e.g. `add.w r10,r11` is `0x5B0A` where `5` is `add`, `B` is src register, `0` means reg to reg word size, `A` is dst register. Just beautiful. Far nicer for emulating on a 6502 or z80 than Arm or RISC-V too. The R2/R3 const generation is a bit whack though.

[1] e.g. on one hand deciding it was worth squeezing a 5 bit offset from any of 4 registers into a 2-byte instruction, while also providing 8 and 16 bit offsets with 3 and 4 byte instructions. They were also confident enough to relegate the 6800's SEC/CLC/SEI/CLI/SEV/CLV to two-byte instructions (with a mask so you could do multiple at once). But not confident enough to do the same with DAA, or SEX. They kept the M6800 encoding for DAA (and for as much else as possible e.g. keeping the opcodes for indexed addressing, but expanding from one option to dozens), but SEX was new to them and they could have experimented with it.

There are uC versions like the W65C134S: https://www.westerndesigncenter.com/wdc/w65c134s-chip.php