Cadence launches comprehensive terminal-side Tensilica AI platform to accelerate intelligent system-level chip development
New Tensilica AI engine improves performance, AI accelerator provides one-stop solution for consumer, mobile, automotive and industrial AI system-on-chip design
Executive summary
● Field-specific, scalable and configurable artificial intelligence platform, based on the mature and mass-produced proven Tensilica architecture
● Provide industry-leading performance and energy efficiency for terminal-side AI applications
● Comprehensive and versatile artificial intelligence software to meet the needs of all target markets
● Low-end, mid-range and high-end AI product lines, suitable for all PPA goals and cost budgets
● Currently, it can be expanded from 8 GOPS to 32 TOPS, and can be expanded to hundreds of TOPS to meet future artificial intelligence needs.
Shanghai, China, September 14, 2021 - Cadence Electronics (Cadence, Inc., NASDAQ: CDNS) today released the Tensilica® AI platform for accelerating the development of artificial intelligence system-level chips, including for different data requirements and terminal side ( on-device) three supporting product families optimized for AI requirements.
The comprehensive Cadence® Tensilica AI platform covers the low-end, mid-range and high-end markets, providing scalable, energy-efficient device-to-edge AI processing capabilities that are key to today's increasingly common AI system-on-chip designs. Compared to the industry-leading standalone Tensilica DSP, the new accompanying AI Neural Network Engine (NNE) consumes 80% less energy per inference and delivers more than 4x TOPS/W performance, while the Neural Network Accelerator (NNA) One-stop solution delivers flagship-level AI performance and energy efficiency.
Targeting smart sensor, Internet of Things (IoT) audio, mobile vision/voice AI, IoT vision and Advanced Driver Assistance System (ADAS) applications, the Tensilica AI platform delivers the best power, performance and area (PPA) in a common software platform ) and scalability. The Tensilica AI Platform product family is based on the highly successful Tensilica DSP, which is already in volume production in leading AI system-on-chips for specific applications in the consumer, mobile, automotive and industrial markets, including:
All NNE and NNA products include stochastic sparse computation to improve performance, runtime tensor compression to reduce memory bandwidth, and pruning and clustering capabilities to reduce model size.
This comprehensive, general-purpose artificial intelligence software targets all target applications, simplifying product development and enabling flexible and easy migration as design requirements change. The software includes Tensilica Neural Network Compiler, which supports the following industry-standard frameworks: TensorFlow, ONNX, PyTorch, Caffe2, TensorFlowLite, and MXNet for automatic end-to-end code generation; Android Neural Network Compiler; TFLite Delegates for real-time execution ; and TensorFlowLiteMicro, for microcontroller-class devices.
client feedback
Pete Warden, technical lead at Google TensorFlow Lite Micro
“Expanding low-power on-device AI capabilities requires extremely efficient multi-sensor computing. The Cadence and TensorFlow Lite for Microcontrollers (TFLM) teams have been working together for years to develop solutions that enable energy efficiency at the forefront of AI Applications. The trend of using LSTM-based neural network operators for optimal performance and efficiency in real-time audio networks is a key example. Working closely with Cadence, we are integrating a highly optimized LSTM operator on the Cadence Tensilica HiFi DSP. "
Albert Liu, founder and CEO of Kneron
“Deploying on-device AI on our KLT720, an AI system with 1.4TOPS performance, is key to our customers’ success and our mission to make AI everywhere and for everyone. Grade chip, designed for automotive, smart home, smart security, industrial control applications, medical and AI Internet of Things (AIoT) Cadence's Tensilica Vision DSP has high performance, low power consumption, powerful computing power, and AI ISA extensions and required. AI software that can handle the latest AI challenges.”
Cristiano Castello, Senior Director of Microcontroller Product Innovation at NXP Semiconductors
"Integrating the Cadence Tensilica HiFi 4 DSP into NXP's i.MX RT600 crossover MCU not only provides high-performance DSP capabilities for a wide range of audio and speech processing applications, but also improves inference performance, even at ultra-low power, Artificial intelligence technology can also be implemented in battery-powered products. The HiFi Neural Network Library enables NXP to take full advantage of the AI performance of the HiFi 4 DSP and integrate it into the NXP eIQ machine learning software development environment that supports TensorFlow Lite Micro and Glow ML inference engines. middle."
Mike Demler, Senior Analyst, The Linley Group
“As AI applications rapidly expand from the cloud to the edge, integrating terminal-side AI accelerators has become a necessity to meet the low-latency requirements of ADAS, mobile, smart sensors, and IoT. AI system-on-chip requires the use of mature accelerator IP, To meet the different needs of each market and include a comprehensive software solution to address changing performance and power requirements, Cadence provides a clear migration path to the endpoint with Tensilica AI Base, AI Boost and AI Max technologies. A widely recognized IP provider in the comprehensive AI IP solution market.”
Availability
The NNE 110 AI engine and NNA 1xx AI accelerator series support Cadence's Intelligent System Design™ strategy, which aims to provide ubiquitous intelligence support for the superior design of system-level chips and is expected to be fully operational in the fourth quarter of 2021 Listed. For more information about the Tensilica AI platform and new AI IP, please visit: www.cadence.com/go/TensilicaAI (copy the URL and open it in your browser, or click at the end of the article to read the original article).
About Cadence
With more than 30 years of expertise in computing software, Cadence is a key leader in the electronic design industry. Based on the company's intelligent system design strategy, Cadence is committed to providing software, hardware and IP products to help electronic design concepts become reality. Cadence's customers are the most innovative companies around the world, delivering everything from chips and circuit boards to the most dynamic application markets such as consumer electronics, hyperscale computing, 5G communications, automotive, mobile, aerospace, industrial and medical. Systematic excellence in electronics. Cadence has been ranked among Fortune magazine's 100 Best Companies to Work For for seven consecutive years. For more information, please visit the company's website at cadence.com.
© 2021 Cadence Design Systems, Inc. All rights reserved. All rights reserved worldwide. Cadence, the Cadence logo and other Cadence marks listed at www.cadence.com/go/trademarks are trademarks or registered trademarks of Cadence Design Systems, Inc. All other marks are the property of their respective owners.
Recommended reading
Click "Read the original text" to learn more~
京公网安备 11010802033920号