BusinessWire
October 09, 2006 06:00 AM Pacific Time

Ambric Announces World's First TeraOPS-Class Chip Using GALS Architecture That Enables Massively-Parallel Programming For Complex Embedded Applications

Initial Markets for the Am2000 Chips are Complex, High-performance Video and Image Processing; Broader DSP and FPGA Markets to Follow

In-Stat Fall Microprocessor Forum 2006

SAN JOSE, Calif.--(BUSINESS WIRE)--Ambric®, Inc. introduced the Am2000™ family of scalable integrated circuits (ICs) and will present a paper on the chips in a session here tomorrow. The Am2000 ICs are the world's first teraOPS-class chips using globally asynchronous, locally synchronous (GALS) architecture that enables massively-parallel solutions for complex embedded applications. Ambric's new chips feature teraOPS-level performance and an innovative, structural object-programming model that dramatically accelerates embedded application development and debug. Initial target markets for the chips are complex, high-performance video and image processing applications, with other applications for the broader, high-end digital signal processing (DSP) and field-programmable gate array (FPGA) markets to follow. Ambric is a privately-held fabless semiconductor company headquartered in Beaverton, Ore., with funding from ComVentures, OVP Venture Partners, and Northwest Technology Ventures.

“The structural object programming model of Ambric's Am2000 chips dramatically speeds development compared with VLIW assembly coding for DSPs or RTL for FPGAs,” said Howard Bubb, CEO and chairman of Ambric. “And, the GALS architecture enables linear scalability of hardware and software to keep pace with Moore's Law at 65-nanometer processes and beyond.”

“I joined Ambric's technical advisory board because reconfigurable computing is the right way to use extremely-large-scale IC technology,” said independent consultant Robert Colwell, chief architect of the Intel IA32 Pentium processor and author of the book, Pentium Chronicles: The People, Passion, and Politics Behind Intel's Landmark Chips. “It's reconfigurable fabrics done right: programmability first—followed by processing granularity, memory partitioning, and then distributed interconnect paths. Ambric has banished the von Neumann memory bottleneck, long the bane of programmers trying to manually juggle multiple control threads. Power dissipation, interconnect bandwidth, complex flow control, and clock gating have been solved. They did it by simultaneously getting several great ideas right. Ambric is not just one generation ahead of other chip companies—Ambric is in another league entirely."

Highly Programmable Architecture Distinguishes Ambric's Am2000 ICs

Practical, efficient programmability is a major differentiator of Ambric's chips. The crucial importance of development ease was highlighted in the recent article, “Ambric Discloses Massively Parallel Architecture”, by Berkeley Design Technology, Inc. (BDTI), (InsideDSP.com, August 23, 2006). From the article: "While massively parallel processors offer the possibility of dramatic performance gains over traditional architectures, these gains will only be realized if the programming model is user friendly and efficient."

The Am2000 ICs are programmable, globally asynchronous, massively parallel, multiple-instruction, multiple-data (MIMD) arrays of 32-bit reduced-instruction-set computing (RISC) processors and memories in a fabric of asynchronous messaging channels. In developing the Am2000 chip family, Ambric first created its structural object-programming model for massively-parallel computing, and then architected the silicon to create a context-free object programming environment in hardware. The result is chips that deliver extreme performance and are straightforward to program.

The Major Benefits of the Am2000 Chips Are Three-fold:

1. Ease of Development: The new structural object programming model builds on existing object programming techniques, resulting in an intuitive parallel programming environment, even for applications with data and control complexity. Ambric's hardware encapsulation of software objects is a fundamental enabler for a practical, highly-usable parallel programming model. This unique model eliminates the complexity and synchronization problems that undermine the scalability of conventional multi-core chip architectures and multi-threaded programming. And, to speed the debug of real-time systems, an independent on-chip network provides debug visibility and transparent, runtime debugger control of each processor and memory.

Ambric's development tool suite employs behavioral simulation driven from the structural system model and the individual, single-threaded programs for each processor. Programs are written in a high-level language or a straightforward RISC assembly language. Unlike FPGAs, Ambric chips do not require developers to design with a hardware description language (HDL) or at the register-transfer-level (RTL). And, unlike DSPs, there is no need for complex, very-long-instruction-word (VLIW) assembly language.

Ambric's Eclipse-based integrated development environment (IDE) includes system-level graphical design tools, object-level compilers, simulation, mapper-router, on-chip debug, and software libraries.

2. High Performance: Extremely high performance and power efficiency are hallmarks of massively-parallel architectures. The Ambric Am2045 delivers up to 1.08 teraOPS with power scaling from 3 watts to 14 watts, depending on the number of processors used and the frequency to which each processor is set. The Am2045 can also fully sustain a throughput of 60 GMACS (16-bit with 32-bit result). Compared with published benchmarks from the leading DSP vendor, the throughput of an Am2045 chip is 10 to 25 times higher than the throughput of the vendor's 1 GHz 90-nanometer DSP.

3. Superior Hardware and Software Scalability: The GALS architecture of the Am2000 chip family eliminates long wires that inhibit performance as the technology scales to 65-nanometers and beyond. Ambric processing ‘brics' may be added as geometries shrink, linearly increasing performance with each generation. Power and performance continue to scale, as there are no long busses or pipelines to slow the part. As brics are added, the software also scales linearly. New software objects may be mapped into additional brics without affecting the existing software objects. Debug of N objects remains an order-N problem.

Ambric Am2000 Chips are Superior to High-performance DSPs and FPGAs

Ambric's Am2000 chips are superior in many respects compared with today's top-of-the-line DSPs and FPGAs. Ambric chips:

Provide more than 10X the throughput of high-end DSPs

Replace a board using several DSPs and/or FPGAs with a single,
easy-to-program part

Lower the cost of software development

Lower product cost, use less power, and have a smaller footprint than do multiple DSPs and FPGAs

Reduce time-to-market

Enable software scalability and object re-use

Are flexibly field-reconfigurable within milliseconds

“Today's advanced video processing is leaving uni-processors in the dust,” said Jon Peddie, president, Jon Peddie Research, Inc. “H.264 encoding for high-definition resolution is more difficult by an order-of-magnitude than is MPEG-2 at standard definition. The only way that the processing can be done in real time is to go massively parallel, and Ambric has nailed it. The chip has the brute horsepower and the ease-of-coding to handle the most difficult video processing.”

Ambric Am2000 Chips Overcome the Challenges of Complex Embedded Computing

Ambric's technology targets the top challenges in embedded computing: ever-increasing complexity, the lack of scalability, and inadequate programming techniques and tools, especially as standards and system requirements evolve.

“The beauty of Ambric's new chip is its programming model,” said Will Strauss, president of Forward Concepts. “Ambric has solved the chronic problem of having no practical way to program massively-parallel chips. The new Ambric chips can replace DSPs and FPGAs and deliver higher performance per dollar than either.”

As David Patterson, a professor of Computer Science at the University of California, Berkeley, stated, "Processors are the transistors of tomorrow." Professor Patterson led the design and implementation of RISC I, likely the first very-large-scale integration (VLSI) reduced instruction set computer and one of the inspirations for Ambric's efficient processor design.

Exemplifying this principle, the architecture of the Ambric chips tiles to thousands of processors, and the software library elements are scalable. In addition, the library elements may be re-used without customization or negative effects. This lets the designer add performance through parallel scaling rather than through mere clock scaling.

"Ambric's massively-parallel architecture is well suited to the needs of network security," said Linley Gwennap, principal analyst of The Linley Group. "To adapt to constantly changing security threats at gigabit speeds, deep-packet inspection must be flexible and fast. Security will be an important application for Ambric's architecture."

Software Libraries: High-performance Video and Image Processing

The Am2000 chips are general-purpose embedded processors. To assist early-adopter customers in its initial markets, Ambric is developing a set of software libraries for high-performance video and image processing applications. These markets, estimated to reach a combined total of $750 million by 2007, include applications such as broadcast TV infrastructure, video contribution and post-production, and video security. In 2007, Ambric plans to target the broader DSP and FPGA markets.

“Video infrastructure is a hot market for high-performance reconfigurable chips,” said Gerry Kaufhold, principal analyst, In-Stat, Inc. “Ambric's new chip is ideal for advanced applications such as video pre-processing, transcoding, rate-shaping, and other demanding algorithms. From what I've learned about the chip, the image quality, compression, and throughput it produces should be world-class.”

Pricing and Availability

The top-end Am2045 chip features 360 32-bit processors and on-chip, distributed RAM totaling 4.6 Mbits. Am2000 ICs are also available with 280, 192, or 96 processors. Pricing for Ambric Am2000 chips starts at $99 U.S. for the 96-processor version in quantities of 1,000 units.

Engineering samples of the Am2000 chips plus developer boards and software development kits will be available in January 2007 to select customers. For more information on the Am2000 family of ICs, contact Ambric at sales@ambric.com.

About Ambric, Inc.

Ambric is a fabless semiconductor company that has developed the world's first chip that makes massively parallel software programming practical for complex embedded electronics systems. The company's scalable, teraOPS-class chips deliver performance that is more than an order of magnitude higher than high-end DSPs and rivals the performance of FPGAs on many applications. Ambric products will help electronics companies accelerate time-to-market for their products while slashing their system development costs.

Established in 2003 and headquartered in Beaverton, Ore., Ambric has received funding from ComVentures, OVP Venture Partners, Northwest Technology Ventures, and private investors. Visit http://www.ambric.com for the latest news and information on the company.