Vision Q8 and vision P1 DSPs extend support for the automotive, mobile and consumer electronics markets

Shanghai, China, April 23, 2021 – cadence (nasdaq:cdns) today announced that it has expanded its popular Tensilica ® Vision DSP product series introduces two new DSP IP processors for embedded vision and AI applications. Compared with the previous generation of Tensilica vision Q7 DSP, the flagship cadence ® Tensilica vision Q8 DSP has the industry-leading 3.8 trillion operations per second (TOPS), which can provide double performance, memory bandwidth and energy efficiency for high-end visual and imaging applications in the automotive and mobile markets. While Tensilica vision P1 DSP is optimized for always on applications and smart sensor applications in the consumer market, providing a very energy-saving solution.

Customers are interested in vision Q8 and vision P1 DSPs and are currently conducting a number of assessments. The newly launched DSP products improve cadence’s mature visual and AI DSP product portfolio, and provide customers with higher design flexibility through first-class technology.

Vision Q8 and vision P1 DSPs are based on similar SIMD and VLIW architectures in existing Tensilica vision DSPs, which are characterized by n-way programming model. The model retains software compatibility and can be easily ported from the previous generation Tensilica vision DSP with different SIMD widths. Like other products of the Tensilica vision DSP family, vision Q8 and vision P1 DSPs also support the Tensilica instruction extension (TIE) language, so customers can customize the instruction set. Both DSPs also support xtensa for neural networks ® Neural network compiler (xnnc) and Android neural networks API (nnapi). In addition, the new product also supports more than 1700 opencv based visual library functions, OpenCL and halide compilers for computer vision and imaging applications. Both products have passed ASIL certification. The random hardware failure certification is asil-b and the systematic failure certification is asil-d. they are very suitable for the automotive industry.

The seventh generation Tensilica vision Q8 DSP also has the following features and functions:

·Optimized for high-end mobile applications and multi camera automotive applications

·Compared with the previous generation vision Q7 DSP, it has 3.8tops computing power and 1024 bit SIMD bit width. These features bring double visual, AI and floating-point performance, and double memory bandwidth

·Single core simplifies system design and reduces power consumption by up to 20%

·4X performance improvement in non convoluted layers for AI processing

·The extended instruction set provides high-level language support for OpenCL and halide. The simplified programming model shortens the programming time. The code written at one time can be applied to different SIMD bit widths.

Tensilica vision P1 DSP features and functions include:

·Optimized for always on applications, including smart sensors, ar/vr glasses and IOT / smart home devices

·Compared with the widely adopted vision P6 DSP in the market, the new vision P1 can provide the computing power of 400 GOPs, increase the frequency by up to 20%, and only need one third of its area. These advantages benefit from its 128 bit SIMD bit width

·The optimized architecture requires less memory and runs in low power mode

“The absolute number of sensors and the demand for higher frames per second and resolution drive the market’s demand for high-performance visual and AI DSPs that support multiple data types.” Sanjive agarwala, vice president of cadence and general manager of IP business unit, said, “At the same time, the market also needs low-power vision DSP and entry-level AI technology to support intelligent sensor applications that are always online. With the launch of Tensilica vision Q8 and vision P1 DSP, cadence can provide a comprehensive vision and AI DSP product portfolio from high-end to low-end, providing our customers with the best flexibility and significantly accelerating the time to market.”

Industry analyst introduction

Mike demler, senior analyst at Linley group, said: “In the mobile, automotive and ar/vr markets, the use of 3D sensors and other types of sensor technologies continues to grow. These markets have an increasing demand for high-performance, low-power visual and artificial intelligence processing functions for edge applications. To meet these demands, system level chip designers are looking for hardware IP with a range of performance points. Cadence has successfully delivered six generations of Tensilica vision DSP, And gained a good reputation. Now, with the introduction of these two new DSPs, cadence will still be an irresistible and ideal choice for system level chip designers. “

Pierre cambou, chief imaging analyst of yole D é velopment (yole), said: “as we have seen in the” CMOS camera module industry for consumer products and automobiles – 2020 version ” As stated in the report, in the next five years, we expect that the sensing field of the image sensor market will continue to maintain a growth rate of 20-30%, and the largest growth will come from the automotive and mobile market. The sensing application in the mobile image sensor market is now very mature in biometrics and 3D; In the automotive field, in view of the surge in the number of more complex use cases, ADAS image sensors have received more and more attention. At present, the market needs a high-performance DSP with several trillion operations per second (TOPS) to meet the upcoming consumers and ADAS level 2+ applications. At the same time, new consumer use cases drive the demand for always on-line operation. Therefore, as explained in our neuromorphological sensing and computing-2019 report, efficiency and low-power operation modes have become popular features. “

Leave a Reply

Your email address will not be published. Required fields are marked *