MathStar is a fabless semiconductor company that designs, manufactures and markets a new class of logic platform chips called "Field Programmable Object Arrays," (FPOAs). FPOAs are high-performance, reprogrammable integrated circuits that process logic functions at a clock rate up to 1 GHz. MathStar's FPOA offers up to four times the performance of commercially available FPGAs.
Today's modern electronic devices and systems offer better functionality, better performance, smaller size, lower power and lower price than ever before. These advantages are largely derived from the increasing capabilities of silicon integrated circuits (chips). It is the availability of these integrated circuit chips that fuels improvements in a wide range of system level solutions, which in turn creates increased demand for even higher performing chips. MathStar believes that system level solutions in many modern applications, such as image and media processing, have evolved to the point of requiring hundreds of billions of computations per second (GOPS) of digital processing performance to compete in today's markets. To achieve these computational rates, many of these applications currently use special-purpose Application Specific Integrated Circuits (ASICs) tailored to one specific product application. Although the use of ASICs can result in optimized chip performance, due to their increasingly higher design costs and long development schedules, they are generally viewed as cost effective only for large-volume, fixed-function applications. FPOAs are an attractive alternative to ASICs in high performance applications because they combine the life-cycle cost, time-to-market and time-in-market advantages of existing programmable logic devices while delivering the performance of ASICs.
An FPOA consists of silicon objects arranged in an internal grid pattern, memory and input/outputs. The object grid is overlaid with a patented, high-speed interconnect system creating the silicon object array. This array of objects and interconnect matrix is programmed to execute unique customer applications. FPOAs provide a new way of implementing high-performance algorithmic functions in silicon integrated circuits. They are not programmed at the gate level but instead are programmed at the silicon object level, which reduces their development time and cost as compared to other methods of chip design. MathStar believes that FPOAs are ideally suited to implement high-performance algorithmic functions in many image processing, communications, computing, industrial and military/aerospace applications.
MathStar is developing a library of different silicon object types that can be mixed and matched in the array in any pattern. Presently, our object types include arithmetic logic unit, truth table, register file and a multiply/accumulate unit. Each of these objects can be programmed to perform a number of different functions, thus enabling this relatively small number of object types to satisfy a range of applications. Each of the object types shares a common interface to the interconnect system. This interface allows us to quickly and easily add object types into the array and rearrange the objects within the array, to create new chips optimized for particular applications and algorithms. MathStar plans to enhance the existing set of objects and develop new object types to continuously expand served markets.
Founded by communications industry veteran Douglas M. Pihl, MathStar is a publicly held company traded on the NASDAQ Global Market (MATH). The company is headquartered in Portland, Oregon and has 60 employees.
More Info | Previous Page | Company Search
If you found this page useful, bookmark and share it on: