PDP-8, PDP-11...
Ah - memories. Booting OS/8 off the toggle switches and paper tape, then an actual hard drive! This was my first taste of assembler programming. I wrote a real time process scheduler, and integrated it with an FFT routine for brain wave analysis in real time. Then we graduated to 16 bits and RT-11. Our local DEC salesman would slip us the source code tape for RT-11 to help with writing device drivers for lab equipment - what a gem! Then bringing up TSX - a multi-user RT-11 emulator and running lots of Teco terminals and somehow we got a version of nroff/troff running under RT for creating and printing documents.
Then RSX-11M, which came with the source code. I remember reading through it and thinking - here are the tracks of a mighty elephant. Indeed, David Cutler went on to create VMS and then NT. But I think RSX-11M was the high point. On a smallish PDP-11 we ran a dozen video terminals (ADM-3a's, I think) for document processing, as well as millisecond data collection in the lab on dozens of A/D channels, all without a hiccough. A friend who was running an early Unix on an 11/70 (a big machine) was losing time because it fell behind on the 60Hz clock ticks - it turned off interrupts every time it had to think about memory mapping, which was often and long.. But Unix was a lot better for getting a program up and running - it had this strange language called "C". And yeah - these were megaherz machines.
There were a few PDP-10s around, but I never did much with them. 36-bit words?