Stream Processors Delivers Industry's Highest-Performing Digital Signal Processor

With 112 GMACs, Storm-1 SP16HP offers a cost-effective, easy-to-design, C-programmable alternative to FPGA and multi-DSP designs

SAN JOSE, Calif. – May 23, 2007 – Stream Processors Inc. (SPI), a fabless semiconductor company focused on making parallel processing simple, today announced the industry’s highest-performing DSP, the Storm-1™ SP16HP. As the newest member of the Storm-1 family, the SP16HP provides a cost-effective, easy-to-design, C-programmable alternative to FPGA and multi-DSP designs in digital imaging, video and wireless infrastructure applications.Introduced today at the In-Stat Microprocessor Forum 2007, the SP16HP sets a new standard for software-programmable processors by offering 112 GMACs (16-bit) or 448 GOPS (8-bit) of compute performance in a single chip with a power-efficiency of less than 0.1 mW per MMAC.

Developing applications in C for SP16HP shortens design times from months to weeks compared to the hardware-oriented methodology used in FPGA designs. According to a recent DSP DesignLine article from industry analyst firm BDTi, “It's not uncommon for an FPGA implementation to require five times as long as the equivalent DSP processor implementation. This huge difference translates into higher cost and slower time-to-market, which can be deal-breakers for many DSP-oriented products.”

FIGURE 1 – Comparison of High-Performance DSP and FPGA Solutions

DSP Industry Comparisons^*
Supplier	SPI	Xilinx	Altera	TI
Device Type	Storm-1 SP16HP	Virtex-5 5VSX35T	Stratix III EP3SL70	TMS320 C6454
Architecture	DSP	FPGA	FPGA	DSP
Clock Frequency	700 MHz	550 MHz	300 MHz	1 GHz
GMACS/chip	112	106	86	8
Devices for ~100 GMACS	1	1	1	13
Design Methodology	Software programmable	Reconfigurable hardware	Reconfigurable hardware	Software programmable

* Source: Company website and other officially published material

Based on SPI’s award-winning Stream Processor Architecture, SP16HP provides a single-core programming model, eliminating the pitfalls of having to partition, synchronize and load-balance threads across multiple cores. The RapiDev™ compiler directs data movement by allowing predictable and optimal application performance and removes the need for assembly coding and manual management of caches and DMA. Will Strauss, president of market research firm Forward Concepts added, “High-end applications are addressed largely by complex multi-chip DSP and FPGA-based systems. SPI’s Storm-1 device offers the performance to tackle these applications in a single chip. More importantly, the company believes that its practical C-programming environment could cut months of development time and negates the drawbacks of multi-core and FPGA design methodologies. Pricing also positions the SP16HP competitively in a market that’s attracting an increased level of design activity.”

“Many developers today have resorted to PCs to run their high-end signal processing applications because they are unfamiliar with multi-DSP or FPGA coding methodologies,” said Bill Dally, Stanford Professor and SPI co-founder. “SPI allows these users to move to an embedded platform using familiar C-programming tools and accelerate their application performance with lower power consumption."

Pricing and Availability
Priced at $149 in quantities of 10,000 units, the SP16HP-G220 device is currently sampling with full production in the third quarter of 2007. The chip is housed in a 31x31mm flip chip ball grid array (FCBGA) package and is implemented in 130nm standard CMOS process.

About Stream Processors, Inc.
Stream Processors, Inc. (SPI) is a privately held fabless semiconductor company delivering an innovative stream processing technology that helps consumer and industrial companies accelerate product development cycles and dramatically reduce system development costs. SPI was founded in 2004 to address the new era of compute-intensive applications requiring radically increased levels of processor performance and power efficiency. The company's technology and products improve application productivity by making parallel processing easier to program and use. Additional information can be found at http://www.streamprocessors.com/.