Huawei launched the AI processor Ascend 910, and an AI computing framework called MindSpore. Huawei had announced both of these AI in October last year. Eric Xu, Huawei’s Rotating Chairman commenting on how perfectly well-rounded both of these projects came out, said, “Ascend 910 performs much better than we expected. Without a doubt, it has more computing power than any other AI processor in the world.”
If both the Ascend and MindSpore are combined, it will be two times faster at training AI models than any other training cards based on ResNet-50. The official blog of Huawei explains, “For half-precision floating-point (FP16) operations, Ascend 910 delivers 256 TeraFLOPS. For integer precision calculations (INT8), it delivers 512 TeraOPS. Despite its unrivaled performance, Ascend 910's max power consumption is only 310W, much lower than its planned specs (350W).”
AI computing frameworks are important steps towards making sure a tightknit privacy protection, making AI applications more accessible, and it helps in developing AI applications faster. MindSpore works keeping all of these on the forefront. MindSpore does not process the data itself and ensures privacy as it deals with gradient and model information only.
Privacy protection has become a major issue, and MindSpore will surely be keeping models secure and trustworthy. Developers will have a smooth experience while developing advanced AI applications thanks to MindSpore’s “AI Algorithm As Code.” Other than the Ascend processors, MindSpore also supports GPUs, CPUs, and other types of processors.
Huawei’s AI strategy is to help boost the development of AI technology from every corner. The Huawei website lists ten areas in which Huawei wants to bring in AI and an eventual change: providing stronger computing power to increase the speed of complex model training from days and months to minutes, providing more affordable and abundant computing power, offering an all-scenario AI portfolio, meeting the different needs of businesses while ensuring that user privacy is well protected, investing in basic AI algorithms.
Algorithms of the future should be data-efficient, using MindSpore and ModelArts to help automate AI development, reducing reliance on human effort, continuing to improve model algorithms to produce industrial-grade AI that performs well in the real world, developing a real-time, closed-loop system for model updates, making sure that enterprise AI applications continue to operate in their most optimal state, maximizing the value of AI by driving synergy with other technologies like cloud, IoT, edge computing, blockchain, big data, and databases, making AI a basic skills for all application developers, and investing more in an open AI ecosystem and build the next generation of AI talent to meet the growing demand for people with AI capabilities.
Ascend 310 comes with a 16-channel FHD video decoder. Ascend 310 has already been used in many products such as shuttle buses, new energy vehicles, and autonomous driving. Huawei’s cloud services such as image analysis, optical character recognition, and intelligent video analysis are done by Ascend 310.
The Huawei website mentioned that more is to come at the Huawei Connect 2019 scheduled to be held between September 18 and 20 in Shanghai.