MasPar MP-2 PE
32-element SIMD processor chip for MasPar supercomputers
The MasPar MP-2 PE is a custom CMOS chip designed by MasPar Computer Corporation. It integrates 32 processing elements for single instruction multiple data computing. Released in 1992, it powered the MP-2 minisupercomputer with high parallel efficiency.
- Contains 32 processor elements on one chip
- Uses 1.0-micrometre CMOS process technology
- Operates at 12.5 MHz clock speed
- Each element has 64 Kbyte local memory
- Dissipates only 0.8 watts of power
What it is
The MasPar MP-2 PE is a specialized integrated circuit built for parallel computing. MasPar Computer Corporation designed this chip to serve as the core of their MP-2 minisupercomputer. It contains thirty-two identical processing units on a single piece of silicon. These units operate in lockstep under a central controller. This design allows the system to perform the same operation on many data points simultaneously. The chip was fabricated using a one micrometre process and shipped in 1992.
How it came about
Jeff Kalb founded MasPar in 1987 after leaving Digital Equipment Corporation. Researchers at DEC had developed a parallel architecture based on the Goodyear MPP. They improved it by using four-bit processors and adding global connections. When DEC declined to sell the technology, Kalb formed his own company. The first product, the MP-1, arrived in 1990. The enhanced MP-2 followed in 1992. MasPar sold over two hundred systems before exiting the hardware market in 1996.
Inside the chip
Each chip holds thirty-two processor elements arranged in two groups. Every element is a full thirty-two bit arithmetic logic unit with floating point support. They share a common memory interface called an M-machine. Each unit has its own sixty-four kilobyte of local dynamic random access memory. There is no cache memory on the chip. The elements connect to eight nearest neighbours via serial links. A global router handles input and output tasks. The architecture uses a Harvard style with separate instruction and data paths.
How it was made
The company designed the chip in-house using custom complementary metal oxide semiconductor technology. The manufacturing process used a one micrometre node with two levels of metal. External vendors such as Hewlett Packard or Texas Instruments fabricated the wafers. Each chip dissipates only zero point eight watts of power. It operates at a clock speed of twelve point five megahertz. The final component is housed in a two hundred and eight pin plastic quad flat package.
Where it went
These chips powered the MasPar MP-2 minisupercomputer systems. Customers used these machines for high performance computing tasks that required massive parallelism. The systems were often front-ended by VAX host computers. Programmers accessed the parallel power through extensions to Fortran and C languages. The National Aeronautics and Space Administration Goddard Space Flight Center was one notable user. Samples from their installation are now preserved at the Computer History Museum.
Versions and family
The MP-2 PE was binary compatible with the earlier MP-1 processor element. The main difference was the integration of thirty-two units per chip in the MP-2. The MP-1 had fewer elements per chip. MasPar did not release an MP-3 hardware version. The company stopped making hardware in June 1996. It transformed into NeoVista Software to focus on data mining applications. Accrue Software later acquired this new entity.
Why it matters
MasPar challenged larger competitors like Intel and Thinking Machines in the supercomputer market. The company criticized government funding decisions that favoured rival architectures. Their unique single instruction multiple data approach offered low latency communication. The design balanced processing power with memory and network bandwidth efficiently. Although the hardware business ended, the technology demonstrated the viability of fine-grained parallel processing for specific scientific workloads.
Problems and reception
The company faced criticism from rivals regarding government support policies. MasPar publicly debated these issues in industry publications like Datamation. Despite the technical innovation, the firm could not sustain its hardware operations long term. It shifted focus to software before being acquired. The physical chips became collectibles, with the company once laminating them onto business cards as a promotional item.

Questions people ask about MasPar MP-2 PE
How many processors are on one MasPar MP-2 PE chip?
The chip contains thirty-two identical processor elements. Each element functions as a full thirty-two bit arithmetic logic unit with floating point capabilities. They operate together under a central control unit.
What manufacturing process was used for the MasPar MP-2 PE?
The chip was made using a one micrometre complementary metal oxide semiconductor process. It featured two levels of metal interconnects. External foundries like HP or TI handled the actual fabrication.
Does the MasPar MP-2 PE have cache memory?
No, the chip does not include any cache memory for the arithmetic logic units. The design relies on local dynamic random access memory for each processor element instead.
What was the clock speed of the MasPar MP-2 PE?
The processor operated at a clock rate of twelve point five megahertz. This relatively low speed was sufficient because the memory interface matched the data access needs of the arithmetic units.
Good reading and sources
The facts on this page come from the Wikipedia article "MasPar", 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.