I think the on-chip stack was a historical artifact. The Datapoint 2200 is the TTL-based computer whose architecture was copied for the 8008 and TMC 1795. The 2200 used serial shift-register memory [1], so writing the PC to a stack in memory would be very slow and inconvenient. Instead it used 8 words of (expensive) RAM as the stack for system calls. The 8008 and TMX 1795 copied the Datapoint 2200 architecture, including the 8-address stack. Intel moved to a "normal" stack in memory with the 8080, which is in many ways a cleaned-up 8008.
Another interesting historical artifact from the Datapoint 2200's shift-register memory and serial processor is that when you're operating on one bit at a time, you want to start with the lowest bit, so you end up with a little-endian processor. For compatibility the 8008 was little-endian, and thus x86, and thus the processor you're probably using now.
Another interesting historical artifact from the Datapoint 2200's shift-register memory and serial processor is that when you're operating on one bit at a time, you want to start with the lowest bit, so you end up with a little-endian processor. For compatibility the 8008 was little-endian, and thus x86, and thus the processor you're probably using now.
[1] http://www.righto.com/2014/12/inside-intel-1405-die-photos-o...