First announced in 2018 at Huawei Connect with planned specs, the Ascend 910 however surpassed expectations and managed to lower power consumption further than originally planned – down to 310W from the expected 350W..
“For half-precision floating point (FP16) operations, Ascend 910 delivers 256 TeraFLOPS. For integer precision calculations (INT8), it delivers 512 TeraOPS,” says Huawei.
The company also introduced MindSpore, an AI computing framework that aims to dramatically reduce training time and costs for making AI applications in various scenarios, whether for devices or the cloud.
“In a typical neural network for natural language processing (NLP), MindSpore has 20% fewer lines of core code than leading frameworks on the market, and it helps developers raise their efficiency by at least 50%,” explained Huawei.
Used for AI model training, the Ascend 910 paired with MindSpore is, according to Huawei, two times faster at training AI models than other mainstream training cards that use TensorFlow.
“Ascend 910 performs much better than we expected. Without a doubt, it has more computing power than any other AI processor in the world,” said Eric Xu, Huawei’s Rotating Chairman.
He also added that Huawei has more AI products to unveil soon.
That will happen between September 18 and 20 in Shanghai, at Huawei Connect 2019.