The Datapoint 2200 was a mass-produced programmable terminal usable as a computer, designed by Computer Terminal Corporation (CTC) founders Phil Ray and Gus Roche[2] and announced by CTC in June 1970 (with units shipping in 1971). It was initially presented by CTC as a versatile and cost-efficient terminal for connecting to a wide variety of mainframes by loading various terminal emulations from tape rather than being hardwired as most contemporary terminals, including their earlier Datapoint 3300.[3]
Dave Gust, a CTC salesman, realized that the 2200 could meet Pillsbury Foods's need for a small computer in the field, after which the 2200 was marketed as a stand-alone computer.[3] Its industrial designer John "Jack" Frassanito has later claimed that Ray and Roche always intended the Datapoint 2200 to be a full-blown personal computer, but that they chose to keep quiet about this so as not to concern investors and others.[2][4]
The terminal's multi-chip CPU (processor)'s instruction set became the basis of the Intel 8008 instruction set, which inspired the Intel 8080 instruction set and the x86 instruction set used in the processors for the original IBM PC and its descendants.
The Datapoint 2200 had a built-in full-travel keyboard, a built-in 12-line, 80-column green screen monitor, and two 47 character-per-inch cassette tape drives each with 130 KB capacity. Its size, 9+5⁄8 in × 18+1⁄2 in × 19+5⁄8 in (24 cm × 47 cm × 50 cm), and shape—a box with protruding keyboard—approximated that of an IBM Selectrictypewriter.[5] Initially, a Diablo 2.5 MB 2315-type removable cartridge hard disk drive was available, along with modems, several types of serial interface, parallel interface, printers and a punched card reader. Later, an 8-inch floppy disk drive was also made available, along with other, larger hard disk drives. An industry-compatible 7/9-track (user selectable) magnetic tape drive was available by 1975. In late 1977, Datapoint introduced ARCNET local area networking. The original Type 1 2200 shipped with 2 kilobytes (KiB) of serial shift register main memory, expandable to 8 KiB. The Type 2 2200 used denser 1 kbitRAM chips, giving it a default 4 KiB of memory, expandable to 16 KiB. Its starting price was around US$5,000 (equivalent to $38,000 in 2023), and a full 16 KiB Type 2 2200 had a list price of just over $14,000.
The 8-bit processor architecture that CTC designed for the Datapoint 2200 was implemented in four distinct ways, all with nearly identical instruction sets, but very different internal microarchitectures: CTC's original design that communicated data serially, CTC's parallel design, the Texas Instruments TMC 1795, and the Intel 8008.[6]
The 2200 models were succeeded by the 5500, 1100, 6600, 3800/1800, 8800, etc.
The fact that most laptops and cloud computers today store numbers in little-endian format is carried forward from the original Datapoint 2200. Because the original Datapoint 2200 had a serial processor, it needed to start with the lowest bit of the lowest byte in order to handle carries. Microprocessors descended from the Datapoint 2200 (the 8008, Z80, and the x86 chips used in most laptops and cloud computers today) kept the little-endian format used by that original Datapoint 2200.[7][8]
Processor
The original design called for a single-chip 8-bitmicroprocessor for the CPU, rather than a processor built from discrete TTL modules as was conventional at the time. In 1969, CTC contracted two companies, Intel and Texas Instruments (TI), to make the chip. TI was unable to make a reliable part and dropped out. Intel was unable to make CTC's deadline. Intel and CTC renegotiated their contract, ending up with CTC keeping its money and Intel keeping the eventually completed processor.[2]
CTC released the Datapoint 2200 using about 100 TTL components (SSI/MSI chips) instead of a microprocessor, while Intel's single-chip design, eventually designated the Intel 8008, was finally released in April 1972.[9]
Possibly because of their speed advantages compared to MOS circuits, Datapoint continued to build processors out of TTL chips until the early 1980s.[7]
Nonetheless, the 8008 was to have a seminal importance. It was the basis of Intel's line of 8-bit CPUs, which was followed by their assembly language compatible 16-bit CPUs — the first members of the x86 family, as the instruction set was later to be known. Already successful and widely used, the x86 architecture's further rise after the success in 1981 of the original IBM Personal Computer with an Intel 8088 CPU means that most desktop, laptop, and server computers in use today[update] have a CPU instruction set directly based on the work of CTC's engineers. The instruction set of the highly successful Zilog Z80 microprocessor can also be traced back to the Datapoint 2200 as the Z80 was backwards-compatible with the Intel 8080. More immediately, the Intel 8008 was adopted by very early microcomputers including the SCELBI, Mark-8, MCM/70 and Micral N.
Instruction set
Instructions are one to three bytes long, consisting of an initial opcode byte, followed by up to two bytes of operands which can be an immediate operand or a program address. Instructions operate on 8-bits only; there are no 16-bit operations. There is only one mechanism to address data memory: indirect addressing pointed to by a concatenation of the H and L registers, referenced as M. The 2200 does, however, support 13-bit program addresses. It has automatic CALL and RETURN instructions for multi-level subroutine calls and returns which can be conditionally executed, like jumps. Direct copying may be made between any two registers or a register and memory. Eight math/logic functions are supported between the accumulator (A) and any register, memory, or an immediate value. Results are always deposited in A. Most instructions are executed in 16μs, 24μs, or a leisurely 520μs when accessing M. The 520μs represents the delay of the 2200's shift register memory to fully recirculate back to the next instruction. Branch type instructions take a variable amount of time (24μs to 520μs) depending on the distance of the branch.
Datapoint 2200 version I instruction set
Opcode
Operands
Mnemonic
Time μs
Description
7
6
5
4
3
2
1
0
b2
b3
0
0
0
0
0
0
0
X
—
—
HALT
—
Halt
0
0
0
0
0
0
1
0
—
—
SLC
16
A1-7 ← A0-6; A0 ← Cy ← A7
0
0
CC
0
1
1
—
—
Rcc (RETURN conditional)
16/†
If cc true, P ← (stack)
0
0
ALU
1
0
0
data
—
AD AC SU SB ND XR OR CP data
16
A ← A [ALU operation] data
0
0
DDD
1
1
0
data
—
Lr data (Load r with immediate data)
16
DDD ← data (except M)
0
0
0
0
0
1
1
1
—
—
RETURN
†
P ← (stack)
0
0
0
0
1
0
1
0
—
—
SRC
16
A0-6 ← A1-7; A7 ← Cy ← A0
0
1
CC
0
0
0
addlo
addhi
Jcc add (JMP conditional)
24/†
If cc true, P ← add
0
1
0
0
0
0
0
1
—
—
INPUT
16
A ← input
0
1
command
1
—
—
EX command (external command)
16
command ← A (coded 8-31 only)
0
1
CC
0
1
0
addlo
addhi
Ccc add (CALL conditional)
24/†
If cc true, (stack) ← P, P ← add
0
1
0
0
0
1
0
0
addlo
addhi
JMP add
†
P ← add
0
1
0
0
0
1
1
0
addlo
addhi
CALL add
†
(stack) ← P, P ← add
1
0
ALU
SSS
—
—
ADr ACr SUr SBr NDr XRr ORr CPr
16/520
A ← A [ALU operation] SSS
1
1
0
0
0
0
0
0
—
—
NOP
16
No operation (Actually LAA)
1
1
DDD
SSS
—
—
Lds (Load d with s)
16/520
DDD ← SSS
1
1
1
1
1
1
1
1
—
—
HALT
—
Halt
7
6
5
4
3
2
1
0
b2
b3
Mnemonic
Time μs
Description
SSS DDD
2
1
0
CC
ALU
A
0
0
0
FC, C false
ADr AD (A ← A + arg)
† Variable. Can be from 24μs to 520μs.
B
0
0
1
FZ, Z false
ACr AC (A ← A + arg + Cy)
C
0
1
0
FS, S false
SUr SU (A ← A - arg)
D
0
1
1
FP, P odd
SBr SB (A ← A - arg - Cy)
E
1
0
0
TC, C true
NDr ND (A ← A ∧ arg)
H
1
0
1
TZ, Z true
XRr XR (A ← A ⊻ arg)
L
1
1
0
TS, S true
ORr OR (A ← A ∨ arg)
M
1
1
1
TP, P even
CPr CP (A - arg)
SSS DDD
2
1
0
CC
ALU
Performance
Although the Datapoint 2200 version I is somewhat faster than an Intel 8008 on register instructions, any reference to the 2200's shift-register memory incurs a large 520µs delay. Also any JMP, CALL, or RETURN can incur a variable delay up to 520µs depending on the distance to the new address. The parallel-architecture Datapoint 2200 version II is much faster than either.[5][10]
Instruction
Description
Datapoint 2200 ver I µs
500 kHz Intel 8008 µs
Datapoint 2200 ver II µs
ADB
Add B to A
16
20
3.2
ADI nn
Add nn immediate to A
16
32
4.8
ADM
Add memory to A
520
32
4.8
JMP nnnn
Jump to nnnn
24-520
44
6.4
CALL+RET
Call and Ret combined
520
64
9.6
Rcc (false)
Conditional return not taken
16
12
3.2
Code example
The following Datapoint 2200 assembly source code is for a subroutine named MEMCPY that copies a block of data bytes from one location to another. Because the byte counter is only 8 bits, there is enough room to load all the subroutine parameters into the 2200's register file. Datapoint 2200 version I transfers 374 bytes per second using this routine. A 500 kHz Intel 8008 executes this code almost four times faster, transferring 1,479 bytes per second. Datapoint 2200 version II is much faster than either at 9,615 bytes per second.[5][10] If more than an 8-bit count is needed, a more complicated copy routine with parameters held in memory would be required.
; MEMCPY --; Copy a block of memory from one location to another;; Entry parameters in registers; HL: 13-bit address of source data block; DE: 13-bit address of target data block; C: 8-bit count of bytes to copy. (1 to 256 bytes)ORG2000Q;Code at 002000 octalMEMCPYLBM;Read source byte into BCALLXCHGI;Exchange HL<->DE and increment DELMB;Save B to target byteCALLXCHGI;Exchange HL<->DE and increment DELAC;Decrement byte counter in CSU1LCAJFZMEMCPY;Continue for all bytesRETURN;Exchange DE and HL register pairs then increment DE as 16 bitsXCHGILAL;Exchange L and ELLEAD1;and inc E, low byte of DELEALAH;Exchange H and DLHDAC0;proagate Cy into DLDARETURNEND
^Thompson Kaye, Glynnis (1984). A Revolution in Progress - A History to Date of Intel(PDF). Intel Corporation. p. 13. "The 8-bit 8008 microprocessor had been developed in tandem with the 4004 and was introduced in April 1972. It was originally intended to be a custom chip for Computer Terminals Corp. of Texas, later to be known as Datapoint." "As it developed, CTC rejected the 8008 because it was too slow for the company's purpose and required too many supporting chips."