36bit

36bit refers to a data width of 36 bits, equivalent to 4.5 bytes. It has been used as a natural word size in several early computer systems, particularly in the 1960s and 1970s.

Notable examples include the DEC PDP-6 and PDP-10 family and the IBM 704/709 family, among others. In such systems, program instructions, addresses, and data could be stored within a single 36-bit word, with the word often partitioned into multiple fields for opcodes, registers, and addresses. Character encoding varied by machine, with some designs packing several characters into one word.

As memory and storage moved toward 8-bit bytes and wider word sizes, 36-bit architectures became rare. Today they survive mainly in retrocomputing projects and certain emulators. Some legacy software and data designed for 36-bit architectures may require conversion to modern widths.

In arithmetic terms, an unsigned 36-bit integer can represent values from 0 to 2^36 − 1 (0 to 68,719,476,735). A signed 36-bit value using two's complement ranges from −2^35 to 2^35 − 1 (−34,359,738,368 to 34,359,738,367). The non-power-of-two width has implications for data packing and alignment, often necessitating special handling when interfacing with byte-addressable, 8-bit-centric modern systems.

Overall, 36-bit architectures occupy a historical niche in computing, notable for their packed word formats and distinctive memory layouts, and they remain of interest to historians, researchers, and emulation projects.