Another (mid-late) data point may be the original NeXT brochure[1], which claims the NeXT to be "a mainframe on two chips". It provides a definition along the lines of a throughput oriented architecture and peripheral channels (in analogy to the NeXT's two DSP chip) and, while doing so, also marries the concepts of physical size and architecture (there's also some romanticism around uncompromised designs and "ruthless efficiency" involved):
> The NeXT Computer acknowledges that throughput is absolutely key to performance. For that reason, we chose not to use the architecture of any existing desktop computer. The desired performance could be found only in a computer of a different class: the mainframe.
> Having long shed any self-consciousness over such mundane matters as size and expense, mainframes easily dwarf desktop computers in the measure of
throughput.
> This is accomplished by a different kind of architecture. Rather than require the attention of the main processor for every task, the mainframe has a legion of separate Input/Output processors, each with a direct channel to memory It's a scheme that works with ruthless efficiency.
There's also the Intel iAPX 432 "micro-mainframe" processor (1981) on two chips. (This was supposed to be Intel's flagship processor, but it was a disaster and the 8086 took over instead. The NYT called it "one of the great disaster stories of modern computing".) I think that microprocessor manufacturers had mainframe envy in the 1980s.
Another (mid-late) data point may be the original NeXT brochure[1], which claims the NeXT to be "a mainframe on two chips". It provides a definition along the lines of a throughput oriented architecture and peripheral channels (in analogy to the NeXT's two DSP chip) and, while doing so, also marries the concepts of physical size and architecture (there's also some romanticism around uncompromised designs and "ruthless efficiency" involved):
> The NeXT Computer acknowledges that throughput is absolutely key to performance. For that reason, we chose not to use the architecture of any existing desktop computer. The desired performance could be found only in a computer of a different class: the mainframe.
> Having long shed any self-consciousness over such mundane matters as size and expense, mainframes easily dwarf desktop computers in the measure of throughput.
> This is accomplished by a different kind of architecture. Rather than require the attention of the main processor for every task, the mainframe has a legion of separate Input/Output processors, each with a direct channel to memory It's a scheme that works with ruthless efficiency.
[1] https://lastin.dti.supsi.ch/VET/sys/NeXT/N1000/NeXTcube_Broc...