Qualcomm ships edge AI chip with expanded power for biometric video processing, 5G integration
Qualcomm Technologies, Inc., a subsidiary of Qualcomm Inc., is aiming to grow its slice of the edge AI chip market with news that the Qualcomm Cloud AI 100 is shipping to select worldwide customers. The AI 100 is a high-performance AI inference accelerator designed to be more power efficient than previous generations of AI accelerators; applications in biometric markets include visual identification.
The Qualcomm Cloud AI 100 is designed to be used in applications ranging from data center to edge appliances and 5G infrastructure. For edge deployments, the chip is integrated with Snapdragon modules for additional application processing and 5G radio capabilities.
For biometric applications, the Qualcomm Cloud AI 100 Edge Development Kit is now available. The company notes that developers have access to a complete edge AI system that includes 5G connectivity; the system is capable of processing up to 24 simultaneous 1080p video streams. Another factor relating to faster product development: Qualcomm provides a variety of tools and supports AI frameworks such as Tensorflow, PyTorch, Caffe, and runtimes such as GLOW and ONNX. The software suite includes an applications SDK with compiler and simulator, along with the Platform SDK with runtimes, APIs, kernel drivers and tools. This software suite is designed to enable the development, testing and deployment of Qualcomm Cloud AI 100 platforms into a multitude of production environments.
Power consumption in edge applications is a key consideration. Qualcomm said the chip has the capability to process up to 400 trillion operations per second (TOPS) in top-line configuration; the Qualcomm Cloud AI 100 supports multiple form factors with card level power profiles ranging from 15W to 75W. Among the configurations offered are; Qualcomm Cloud AI 100 for low-power and high-performance AI inference processing, Qualcomm Snapdragon 865 Modular Platform for application and video processing, and the Snapdragon X55 ModemRF System Module, which includes 5G connectivity in a pre-certified module offering global carrier support.
Market research company Omdia forecasts that global AI edge chipset revenue will grow from $7.7 billion in 2019 to $51.9 billion by 2025. Edge inference has emerged as a key workload and many companies have introduced chipset solutions to accelerate biometrics and other AI workloads.
Qualcomm’s new chip comes amid a significant amount of funding for edge AI chip startups. Industry consolidation is also going to be a factor for edge AI chips. In particular, Nvidia’s proposed acquisition of ARM for $40 billion will meld Nvidia’s AI capabilities with a wide array of ARM’s CPU and embedded processor designs that are used in everything from cloud data centers to smartphones, PCs, self-driving cars and robotics, to edge IoT systems.