Aimed at developers working on AI hardware and software solutions, Linaro has announced a new 96Boards.ai website and associated 96Boards development platforms at Linaro Connect Hong Kong 2018 (HKG18). These include HiKey 970, Ultra96, ROCK960 PRO & Enterprise Edition. 96Boards is Linaro’s initiative to build a single software and hardware community across cost-effective development boards based on Arm technology.
As per the announcement, “Linaro members HiSilicon, Socialtext, and Xilinx, and new 96Boards steering committee member Rockchip have all released 96Boards development products with AI hardware capabilities. Additionally, Qualcomm Technologies plans to join 96Boards.ai with the DragonBoard 820c development board, featuring the Qualcomm Snapdragon 820E embedded platform.”
“Together, these platforms offer an unrivaled choice of heterogeneous compute solutions for AI development. These products offer different combinations of CPU, GPU, NPU, DSP and FPGA with both on-die and independent acceleration supporting hard-coded, optimized performance as well as programmable flexibility”, the company noted in the announcement.
The DragonBoard 820c development board supports a wide range of interfaces and is ideal for prototyping complex embedded computing applications including for VR and AR, machine vision and artificial intelligence for commercial drones and robots. The board features the Snapdragon 820E embedded platform with a 64-bit Armv8 -compliant quad-core Qualcomm Kryo CPU, Qualcomm Adreno 530 GPU and Qualcomm Hexagon 680 DSP.
Qualcomm said that it aims to provide support for Linux on its Snapdragon Neural Processing Engine SDK in the second half of 2018 for use on DragonBoard 820c, so developers can take advantage of the Snapdragon 820E heterogeneous hardware architecture to accelerate on-device AI applications.
Yang Zhang, 96Boards Director and Chief Scientist of the Chinese Academy of Sciences AIRIA, explained:
“96Boards.ai brings together the latest SoCs with AI acceleration capability from multiple world-leading silicon vendors to provide the most advanced development and prototyping platforms for intelligent products. This new open AI platform will empower algorithm developers, researchers, product designers and SoC vendors to focus on their own, unique value-added differentiation.”
Tomas Evensen, Chief Technology Officer of Embedded Software Xilinx, said:
“As an inaugural member of the 96Boards.ai community we are excited to deliver to customers performance that is orders of magnitude higher than conventional microprocessors. As the world’s leading supplier of field programmable gate arrays (FPGAs), we continue to make great strides in all facets of AI acceleration. Xilinx’s unique heterogeneous combination of programmable elements, including; applications processors, real-time processors, digital signal processing elements and of course the FPGA logic fabric, gives us a unique advantage in bringing an unprecedented amount of performance and flexibility to the 96Boards.ai community. The new Ultra96 board, available from Avnet Electronics, is the flagship low-cost vehicle to demonstrate those benefits.”
Speaking on the launch, Eric Zhou, product marketing director of Huawei wireless terminal chipset BU, explained:
“HiKey970 is the third generation of HiKey series in 96boards. The HiKey970 integrates Huawei’s HiAI framework and other mainstream neural network frameworks, supporting both CPU and GPU AI calculations and NPU-based neural network computing hardware acceleration, which can greatly help on-device AI development.”
Pengcheng Zou, CTO of ThunderSoft, commented:
“As a world-leading smart system technology, service and solution provider, ThunderSoft is always the believer in open standards and collaborative innovation. That’s why we are committed to build AI kit and solutions around 96Boards Open AI platforms. With our years of expertise and experience in smart system development, Thundersoft will enable the next generation of embedded AI innovation based on 96Boards.ai platforms.”
More information: https://www.linaro.org/blog/high-performance-computing-and-linaro/