Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

As noted by Ken, not only the "special registers" are weird in that dictionary definition, but also the mention about "main storage".

The term CPU has been introduced by IBM in 1954, in the manual of operation of IBM 704, and it was defined thus: “The central processing unit accomplishes all arithmetic and control functions”.

This definition was clearly meant to say that the CPU is the aggregate of 2 parts of the 5 computer parts listed by John von Neumann, i.e. the CPU is the "central arithmetical part" + "the central control part" from von Neumann's list. The other 3 parts were the main memory, the input peripherals and the output peripherals.

The IBM definition has remained valid until today, even if the package that contains the CPU also contains cache memories and possibly even a part or all of the DRAM memory, like in Apple CPUs or Intel Lunar Lake.

Because the CPU is the most important part of a CPU package, the whole package is referred as the CPU ("pars prō tōtō"), even when the package also contains some part of the memory.

I assume that the same way of using the "CPU" term has caused the confusion from the dictionary entry.

In IBM 704, the CPU filled an entire cabinet and the main memory was in a separate cabinet. In later computers, by the time of Honeywell 800, it became possible to locate both the CPU and the main memory in a single cabinet, which was referred to as the CPU cabinet, as the CPU was its most important part.

Whoever has copied that CPU definition, presumably from a manual of Honeywell 800, has confused the description of the content of the "CPU cabinet" of that particular computer with the definition of the term "CPU" as the name of a block used in the description of computer architectures.



Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: