
Intel 8231/8232
Early single-chip floating-point coprocessors for microcomputers.
The Intel 8231 and 8232 were early floating-point units licensed from AMD. Released in the late 1970s, they accelerated math for systems like the Apple II and i8080 by handling complex calculations independently of the main CPU.
- Licensed from AMD Am9511 and Am9512 designs.
- Supported asynchronous operation independent of CPU speed.
- Used in Apple II accelerator cards for faster math.
- Achieved up to 71,429 FLOPS at 4 MHz clock speed.
What it is
The Intel 8231 and 8232 are integrated circuits that perform floating-point arithmetic. Intel marketed these chips as companions to its i8080 central processing unit. They were not original Intel designs but licensed versions of the Am9511 and Am9512 created by AMD. These components represented some of the first single-chip solutions for hardware-based math acceleration. The 8231 acted as an Arithmetic Processing Unit, while the 8232 served as a Floating-point Processor Unit. Their primary role was to offload complex numerical tasks from the main processor, significantly speeding up scientific and engineering applications on early microcomputers.
How it came about
AMD introduced the Am9511 in 1977 and the Am9512 in 1979, claiming them as the world's first single-chip floating-point units. Intel licensed these designs to offer similar capabilities for its own ecosystem. The resulting Intel parts, the 8231 and 8232, entered the market shortly after their AMD counterparts. This arrangement allowed Intel to provide hardware acceleration before developing its own in-house solutions. The collaboration ended when Intel launched the 8087, a proprietary design tightly coupled with the x86 architecture. The licensed chips thus filled a critical gap in the late 1970s computing landscape.
Inside the chip
These chips utilized an 8-bit data bus, matching the standard of contemporary processors like the i8080. The 8231 supported 32-bit floating-point and fixed-point math using Chebyshev polynomials for trigonometric functions. The 8232 adhered to the draft IEEE-754 standard for 32-bit and 64-bit precision but limited operations to addition, subtraction, multiplication, and division. A key feature was asynchronous operation, allowing the coprocessor to run at a different speed than the host CPU. This enabled parallel processing where the main CPU could continue working while the FPU calculated, communicating only via interrupts upon completion.
How it was made
Intel manufactured these devices, though the underlying architecture originated with AMD. The chips were available in two speed grades: 2 MHz and 4 MHz. The 4 MHz versions offered improved performance over the original AMD Am9512, which capped at 3 MHz. Pricing in bulk quantities was set at US$149.00 for the slower model and US$235.00 for the faster one. The design allowed for flexible integration via direct memory access or interrupt-driven methods, requiring minimal additional logic circuitry to interface with various microprocessor systems.
Where it went
Although intended for the i8080, the flexible interface allowed use in diverse systems. Documentation cited compatibility with the Z80, MC6800, i8085, and Z8000. Notably, the 8231A powered accelerator cards for the 6502-based Apple II, such as the Applied Analytics MicroSPEED II. This allowed an approximately 1 MHz system to benefit from 4 MHz numeric processing without slowing the host. Before the 8087 arrived, Intel even suggested these chips for the new 8086. They provided a massive speedup over software-only math, which struggled to reach 1000 FLOPS on 8-bit hardware.
Versions and family
The 8231 was initially released, followed by the revised 8231A. The 8231A and 8232 both supported asynchronous operation and speeds up to 4 MHz. The 8231 focused on a broader set of arithmetic and trigonometric functions with proprietary precision standards. The 8232 prioritized IEEE-754 compliance for basic arithmetic operations. Both served as predecessors to the Intel 8087, which later became the standard for x86 floating-point processing. The AMD originals, Am9511 and Am9512, remained distinct but functionally similar alternatives in the market.
Why it matters
These chips broadened the market for hardware acceleration by decoupling the FPU from specific CPU families. Their ability to operate asynchronously meant older or slower systems could gain significant computational power. Performance ranged from roughly 332 FLOPS for complex power calculations to over 70,000 FLOPS for simple subtractions at 4 MHz. This was radically faster than software emulation on 8-bit processors. The design philosophy influenced how peripheral processors interacted with main CPUs, demonstrating the value of parallel task execution in early microcomputer architectures.
Questions people ask about Intel 8231/8232
What is the difference between the 8231 and 8232?
The 8231 handles 14 arithmetic and trigonometric functions using a proprietary standard. The 8232 supports only four basic operations but follows the IEEE-754 standard for true single and double precision floating-point numbers.
Can these chips work with processors other than Intel?
Yes. Their flexible interface allows connection to Z80, MC6800, and 6502 systems. They were notably used in accelerators for the Apple II, proving their compatibility beyond the Intel i8080 ecosystem.
How fast were these chips compared to software math?
They were radically faster. While an 8 MHz 8086 might achieve 1000 FLOPS in software, the 8232 could reach over 70,000 FLOPS. This offloaded heavy calculations, freeing the main CPU for other tasks.
Good reading and sources
The facts on this page come from the Wikipedia article "Intel 8231/8232", written by its contributors and used under the CC BY-SA 4.0 licence, and from its Wikidata record (CC0). EveryChip rewrote the text in its own words, with AI assistance, and compares every page with the article to make sure no passage is copied. It may be reused under the same CC BY-SA 4.0 licence with a link back.
It may contain mistakes, and figures change over time. Nothing here is investment or buying advice. If something is wrong, tell us. Last refreshed 2026-11-14.
Like Intel 8231/8232
All analog, power and mixed-signalMore from Intel
All of them
🇺🇸 United States and 2 moreEEPROM
Byte-erasable non-volatile memory for small data storageMemory · 1978