Huawei has announced the launch of Ascend 910, its AI processor for data training as well as its AI computing framework MindSpore. Ascend 910, which was first unveiled in 2018, has been improved to reduce the power consumption to 310W. The chipset and MinSpore can work in tandem to train AI models two times faster than training cards that use TensorFlow while in a standard training session based on ResNet-50.
Huawei is touting Ascend 910 as the world’s most powerful AI computing chipset. The processor is said to deliver 256 TeraFLOPS for half-precision floating points (FP16) and 512 TeraFLOPS for integer precision calculations (INT8) respectively.
Ascend series of chipsets will not be offered standalone in the market, Huawei has said. The chipsets will be embedded in the company’s acceleration cars, cloud services, etc. “We have been making steady progress since we announced our AI strategy in October last year. Moving forward we will continue to invest in AI chipsets. Our plan is to launch Ascend 320 in 2021 and Ascend 920 moving forward so that we have a complete AI computing infrastructure,” said Huawei’s rotating chairman Eric Xu.
Meanwhile, the company says that Huawei’s MindSpore AI framework can help developers raise their efficiency by at least 50 per cent, thanks to fewer lines of code than leading frameworks on the market. The framework is said to be secure and it ensures privacy as well given it does not process data. Rather it works on processed data.
Another highlight is it has been designed to support different kinds of deployment needs whether big or simple. It supports other processors as well. MindSpore will go open-source in the first quarter of 2020. The idea is to enable engineers to develop AI applications, which Huawei says will contribute hugely to its ecosystem. MindSpore also supports other processors as well as CPUs and GPUs in addition to Ascend.
“The development of ecosystem is very important. The biggest challenge for MindSpore is whether we can sell it as compelling ecosystem as other mainstream frameworks currently available in the market. But there are advantages with Huawei when it comes to ecosystem development because we have smart devices in our portfolio, we have high-AI engine in smart devices where developers of mobile applications can access Huawei’s AI computing power to develop different types of applications and services leveraging our AI capabilities.
According to Xu, AI is still at a very early stage and he believes that Huawei has comprehensive enterprise business. On asked what will the company’s strategy be in India, especially since the uncertainty regarding Huawei’s 5G rollout in the country, Xu said, “Our AI strategy is not necessarily tied to 5G. As for our AI strategy in India, it is no different from our overall strategy.”
Huawei will develop Atlas as well as MDC products based on Ascend processors, which can be provided to universities and other partners in India as they develop applications to address industry-specific challenges. “Soon, we will have training and inference class services based on Ascend processors as well which can also be made available in the Indian market.”
Disclaimer: The author is in Shenzhen, China attending Huawei AI Chip and New Open-Source Computing Framework launch conference on the invite of Huawei India