Instructions per second (IPS) is a measure of a computer's processor speed. Many reported IPS values have represented "peak" execution rates on artificial instruction sequences with few branches, whereas realistic workloads consist of a mix of instructions and applications, some of which take longer to execute than others. The performance of the memory hierarchy also greatly affects processor performance, an issue barely considered in MIPS calculations. Because of these problems, researchers created standardized tests such as SPECint to (maybe) measure the real effective performance in commonly used applications, and raw IPS has fallen into disuse.

The term is commonly used in association with a numeric value such as thousand instructions per second (kIPS), million instructions per second (MIPS), or Million Operations per Second (MOPS).

Thousand instructions per second

Before standard benchmarks were available, average speed rating of computers was based on calculations for a mix of instructions with the results given in kilo Instructions Per Second (kIPS). The most famous was the Gibson Mix, produced by J Gibson of IBM for scientific applications. Other ratings were also produced for commercial applications. Computer Speeds From Instruction Mixes pre-1960 to 1971 has results for around 175 computers, providing scientific and commercial ratings. For IBM, the earliest Gibson Mix calculations shown are the 1954 IBM 650 at 0.06 kIPS and 1956 IBM 705 at 0.5 kIPS. The results are mainly for IBM and others known as the BUNCH - Burroughs, Univac, NCR, CDC and Honeywell.

A thousand instructions per second (kIPS) is rarely used, as most current microprocessors can execute several million instructions per second. The thousand means 1000 not 1024.

kIPS is also a common joke name for 16 bit microprocessor designs developed in undergraduate computer engineering courses that use the text Computer Organization and Design by Patterson and Hennessy (ISBN 1-55860-428-6), which explains computer architecture concepts in terms of the MIPS architecture. Such architectures tend to be scaled down versions of the MIPS R2000 architecture.

Million instructions per second

MIPS are not comparable between CPU architectures. This and other limitations of the unit lead many computer engineers to define MIPS as "Meaningless Indicator of Processor Speed."

The floating-point arithmetic equivalent of MIPS is FLOPS, to which the same cautions apply.

In the late 1970s, minicomputer performance was compared using VAX MIPS, where computers were measured on a task and their performance rated against the VAX 11/780 that was marketed as a "1 MIPS" machine. (The measure was also known as the "VAX Unit of Performance" or VUP. Though orthographically incorrect, the "S" in "VUPs" is sometimes written in upper case.) This was chosen because the 11/780 was roughly equivalent in performance to an IBM System/370 model 158-3, which was commonly accepted in the computing industry as running at 1 MIPS.

Many of the minicomputer performance claims were based on the Fortran version of the Whetstone benchmark. This produces an artificial speed rating in Millions of Whetstone Instructions Per Second (MWIPS). Whetstone Benchmark History and Results provides some 700 results for minicomputers, mainframes, supercomputers and PCs. The VAX 11/780 with FPA (1977) is shown as having a rating of 1.02 MWIPS.

Effective MIPS speeds are highly dependent on the programming language used. The Whetstone Report has a table showing MWIPS speeds of PCs via early interpreters and compilers up to modern languages. The first compiler was for BASIC (1982) when a 4.8 MHz 8088/87 CPU obtained 0.01 MWIPS. Results on a 2.4 GHz Core 2 Duo (1 CPU 2007) vary from 9.7 MWIPS using BASIC Interpreter, 59 MWIPS via BASIC Compiler, 347 MWIPS using 1987 Fortran, 1534 MWIPS through HTML/Java to 2403 MWIPS using a modern C/C++ compiler. Source code, pre-compiled versions and results on PCs, for these and other benchmarks that measure MIPS, are available from Roy Longbottom’s PC Benchmark Collection (Free).

Most 8-bit and early 16-bit microprocessors have a performance measured in kIPS (thousand instructions per second), which equals 0.001 MIPS. The first general purpose microprocessor, the Intel i8080, ran at 640 kIPS. The Intel i8086 microprocessor, the first 16-bit microprocessor in the line of processors made by Intel and used in IBM PCs, ran at 800 kIPS. Early 32-bit PCs (386) ran at about 3 MIPS.

zMIPS refers to the MIPS measure used internally by IBM to rate its mainframe servers (zSeries, IBM System z9, and IBM System z10).

Timeline of instructions per second

>

Processor	IPS	IPS/MHz	Year	Source
Pencil and paper (for comparison)			1892
Intel 4004			1971
IBM System/370 model 158-3			1972
Intel 8080			1974
VAX 11/780			1977
Motorola 68000			1979
Intel 286			1982
Motorola 68020			1984
ARM2			1986
Motorola 68030			1987
Intel 386DX			1988
Motorola 68040			1990
Intel 486DX			1992
PowerPC 600s (G2)			1994
Motorola 68060			1994
Intel Pentium Pro			1996
ARM 7500FE			1996
PowerPC G3			1997
Zilog eZ80			1999
Intel Pentium III			1999
Freescale MPC8272			2000	Integrated Communications Processors
AMD Athlon			2000
AMD Athlon XP 2400+			2002
Pentium 4 Extreme Edition			2003
ARM Cortex A8			2005
AMD Athlon FX-57			2005
AMD Athlon 64 3800+ X2 (Dual Core)			2005
Xbox360 IBM "Xenon" Triple Core			2005
PS3 Cell BE (PPE only)			2006
AMD Athlon FX-60 (Dual Core)			2006
Intel Core 2 X6800			2006
Intel Core 2 Extreme QX6700			2006
P.A. Semi PA6T-1682M			2007
Intel Core 2 Extreme QX9770			2008

Historic Data

• Computer Speeds From Instruction Mixes pre-1960 to 1971 (kIPS 175 systems)
• Computer Speed Claims 1980 to 1996 (MIPS >2000 systems)
• PC CPU Performance Comparisons %MIPS/MHz

See also

• FLOPS
• benchmark (computing)
• million service units (MSU)
• Peak MIPS
• Relative MIPS
• Dhrystone MIPS (DMIPS)

References