TeraGen Corp., a fabless semiconductor start-up, is looking to carve out a position for itself in the competitive licensable-IP market with a new embedded "thread processor" technology.
The TeraGen architecture can be used to lay multiple embedded processing engines on a single-core platform by way of parallel sequences of simple software instructions called threads, according to George Alexy, president and chief executive of the Sunnyvale, Calif.-based company.
Depending on the embedded application, the thread processors can process two or more parallel instructions.
"[TeraGen's] thread processor architecture has the potential to enable an entire new generation of embedded processors," said analyst Will Strauss of Forward Concepts Co., Tempe, Ariz. "[A single chip could contain] any and all of the functions previously performed by separate microcontrollers, DSPs, and peripheral controllers."
TeraGen said it plans to have chips based on the technique in volume production by next year. The company already has signed two licensees for its first platform, an 8-bit, MCU-based solution.
TeraGen, founded in 1997, received venture financing from InterWest Partners and Sequoia Capital Partners. According to its business model, TeraGen will license its technology to other semiconductor companies, in addition to establishing a foundry base to manufacture its own devices.
"Our objective is that within the next year, we'll have a menu of different instruction-set architectures and a menu of peripherals that a potential licensee could select from and simply configure to meet their needs," Alexy said.
Tom Starnes, an analyst at Dataquest Inc. in Austin, Texas, said the concept will allow developers more flexibility in customizing their products.
"They've got a different approach in focusing on the instruction sets and peripheral functions rather than strictly on the main processor," Starnes said.