PicoChip Designs has received first silicon based on its design ofmassively parallel array of processors. The company, based in Bath,England, has developed a complete baseband platform for 3Ginfrastructure that it believes solves three critical problemsconfronting manufacturers.
The design should reduce the cost of basestations while providingthe flexibility to allow software upgrades to new revisions of thestandard and speed up development.
Based on a massively parallel array of processors, PicoChip saysits device delivers 10 to 100 times the processing power and speed ofthe best available solution based on legacy DSP architectures.
The picoArray is supported by an integrated toolchain, a set ofsoftware libraries for 3G and design support. It is a massivelyparallel array of heterogeneous processors, with more than 400custom-designed devices linked by a patented interconnect mechanism.Each of these processors is programmable, allowing manufacturers tointegrate their own algorithms and intellectual property. PicoChip'slibraries can also be used to implement a complete 3G Node B.
The first implementation of a picoArray &endash; the PC101&endash; contains a parallel array of over 400 individual processors.Several varieties of processor were designed and optimized to suitdifferent roles, each with roughly the general processing power of anARM9 processor, and dramatically more in appropriate cases. Thecomputational power of the picoArray means that one device canreplace a large number of expensiveFPGA or legacy DSP chips.
An advantage of the picoArray is that it is fully deterministic:since both processing behavior and interconnect are fully defined atcompilation, there is no run-time scheduling or arbitration andperformance is completely predictable. The single developmentenvironment eliminates many of the co-design and systems integrationissues of ASIC/FPGA/DSP approaches.
Although the company's initial focus is in UMTS, the picoArray isa platform solution could be competitive in a range of applications,including cdma2000, broadband wireless access VDSL headends, WLANconcentrators, and other demanding communications systems. In thesecases, itcan replace high-end DSPs, in-house ASICs and large FPGAs toaddress a potential total market which the company says in excess$3billion.
Early next year the company expects to ship a demonstration board based on eight chips which will provide 64 channel UMTS design.