The 8088/8086 were probably the first - and last - "breadboardable" x86 CPUs as everything after that came in square, non-DIP packages. However, there have been hobbyists making their own electronics for x86-based computers up to the 486: http://dubel.org/computer/
Starting with the Pentium, however, things got really hairy...
FWIW...The 80286 came in a 40-pin DIP, and of course lots of non-x86 processors came (still come) in DIP as well.
From a hobbyist perspective, PGA packages aren't really alot more difficult to deal with than DIP. Even LCC & PLCC packages have easy through-hole sockets.
QFP & PGA are, of course, nightmares. :-)
The real problem is it becomes increasingly difficult to breadboard a machine as processor clock goes above 10-20MHz. I've actually been puzzling how slow a PCI bus can run, since prototyping something at 33MHz is really hard.
Oops, yes. I've been considering an lpc812 16-pin device for a project but it would need a bunch of '595s for extra output bits. Instead I'm using a 64-pin stm32f030.. the overall price is the same, but the stm32f030 has more flash and RAM. Even so, NXPs development system seems more coherent (even though I'm not an eclipse fan).
The 80286 never came in a 40-pin DIP. It's impossible since the 24-bit address bus and 16-bit data bus were not multiplexed and required 40 pins. The PGA, LCC and PLCC versions all had 68 pins.
From a hobbyist perspective, PGA packages aren't really alot more difficult to deal with than DIP. Even LCC & PLCC packages have easy through-hole sockets.
QFP & PGA are, of course, nightmares. :-)
The real problem is it becomes increasingly difficult to breadboard a machine as processor clock goes above 10-20MHz. I've actually been puzzling how slow a PCI bus can run, since prototyping something at 33MHz is really hard.