[kiwidrew]

Most of the "modern" embedded stuff comes as a complete system-on-chip these days, i.e. all the peripherals and memory are integrated into the CPU on a take-it-or-leave-it basis. It's also rare for chips to offer an external parallel bus interface (the classic A0-A15 and D0-D7 pins), which means that any extra peripheral devices need to hang off an I2C/SPI port where the CPU core can't access them directly.

And the available chips (AVR, ESP32, LPC2xxx, etc.) are all proprietary designs with a single manufacturer. Even the ARM chips, which generally use some variant of a Cortex-M core, have wildly different peripherals. So migrating between families is difficult/impossible.

In contrast, "classic" chips like 8051, 6502, x86 and 680xx all have external parallel bus interfaces and are (or were) produced by multiple manufacturers (often as part of "second source" agreements).

So when building a system using these chips, the designer has a large degree of flexibility and freedom to design the system architecture. Whereas building something using modern embedded chips is mostly an exercise in parametric search trying to find an existing chip which offers exactly the right set of peripherals for the intended design. It reduces the system designer to a mere consumer of off-the-shelf SoCs instead of being a true builder/architect.

[rcar]

That's helpful; thanks. I guess coming from much more of a software background, having the ability to write C code (which feels like an acceptable veneer over the hardware to me and works well across chips) and having programmatic access to the pins feels pretty empowering and all-encompassing. However, it does make sense to me that someone coming from a hardware-first view of the world would feel those barriers to direct hardware access much more acutely and recognize the limitations that I don't.

I appreciate the thoughtful and detailed response.

[noone_youknow]

This. Well said.