Field-programmable object array
Programmable logic device bridging FPGA and ASIC gaps
The field-programmable object array is a programmable logic device class designed to merge FPGA flexibility with ASIC performance. Developed by MathStar in the mid-2000s, it used silicon objects like ALUs for high-speed processing before production ceased in 2008.
- Designed to bridge gap between FPGA and ASIC
- Uses grid of programmable silicon objects
- First chips fabricated in 2006 by MathStar
- Production stopped in May 2008 due to losses
- Patents sold to OLK Grun GmbH in 2011
What it is
A field-programmable object array is a type of programmable logic device that can be configured after manufacturing. It aims to combine the reprogrammability of field-programmable gate arrays with the high performance of application-specific integrated circuits. The architecture uses a grid of programmable silicon objects rather than individual logic gates. This approach allows for deterministic timing and gigahertz-level operating speeds, addressing the performance limitations often found in traditional FPGAs while retaining flexibility.
How it came about
Douglas Pihl conceived the idea while working on a project funded by DARPA. He established MathStar in 1997 to develop and manufacture these devices. The concept was patented in 2004. Prototype units were produced in 2005, leading to the first fabrication batch in 2006. Despite the technical innovation, MathStar never achieved profitability. The company halted production in May 2008. It later merged into Sajan Inc in 2010, which acquired the patents. Sajan sold several of these patents to OLK Grun GmbH in November 2011.
Inside the chip
The core consists of a grid of silicon objects connected by synchronous interconnects. These objects include arithmetic logic units, register files, and multiply-and-accumulate units. Each object has supporting structures for clock synchronization and built-in self-test. Peripheral circuitry handles memory and input-output functions. Communication occurs via nearest neighbor links for adjacent objects and party-line connections for remote ones. Nearest links allow one hop per clock cycle, while party lines support four hops per cycle. Programming operates at the object level using System C, which simplifies development compared to gate-level design.
How it was made
MathStar fabricated the first batch of chips in 2006. The Arrix family was the primary product line released that year. Specific manufacturing process nodes are not detailed in the available records. The devices were packaged with peripheral circuitry for memory and I/O interfaces. Production ended in 2008 due to financial difficulties. The intellectual property, including patents on the FPOA technology, changed hands through corporate mergers and sales in the following years.
Where it went
These devices suit hardware acceleration tasks where FPGAs are typically used. Applications include digital signal processing, medical imaging, and computer vision. They are also employed in speech recognition, cryptography, and bioinformatics. The optimized silicon objects provide benefits in flat field error correction and fast Fourier transform computations. Other uses involve video encoding and decoding, machine vision, and artificial intelligence acceleration. The technology supports aerospace applications and computer hardware emulation as well.
Versions and family
The Arrix range was the notable family of FPOAs. It contained three specific types of silicon objects: arithmetic logic units, register files, and multiply-and-accumulate units. The exact number and arrangement of these objects varied by family. No other specific variants or successor families are recorded before production ceased. The technology did not evolve into a broader commercial product line beyond the initial offerings from MathStar.
Why it matters
The FPOA attempted to solve the trade-off between FPGA flexibility and ASIC performance. It offered a unique architecture with deterministic timing and high clock speeds. However, it did not achieve commercial success or widespread adoption. MathStar failed to post a profit, leading to the shutdown of production. The patents were eventually sold off, indicating the technology did not sustain a standalone market presence. Its legacy remains in the conceptual approach to object-based programmable logic.
Problems and reception
MathStar struggled financially and never posted a profit. This lack of commercial viability led to the decision to shut down production in May 2008. The company was subsequently merged into Sajan Inc. There are no records of specific technical bugs or recalls, but the business failure suggests limited market reception or adoption challenges. The patents were sold in 2011, marking the end of active development by the original creator.
Questions people ask about Field-programmable object array
What is a field-programmable object array?
It is a programmable logic device that uses a grid of silicon objects like ALUs instead of individual gates. It aims to combine FPGA flexibility with ASIC performance and speed.
Who invented the FPOA technology?
Douglas Pihl developed the concept while working on a DARPA-funded project. He founded MathStar in 1997 to manufacture the devices, and the idea was patented in 2004.
Why did MathStar stop producing FPOAs?
The company never posted a profit despite releasing chips in 2006. Due to financial struggles, MathStar decided to shut down production in May 2008 and later merged with Sajan Inc.
How are FPOAs programmed?
Programming is done at the silicon object level using System C. This higher-level approach eases the learning curve compared to gate-level FPGA design and speeds up development.
Good reading and sources
The facts on this page come from the Wikipedia article "Field-programmable object array", 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 Field-programmable object array
All FPGAs and programmable logic
🇺🇸 United StatesVirtex
High-performance FPGA family by AMD and XilinxFPGA · Other · 1999