Microsoft launched the Brainwave project to provide faster chips to attract AI developers.

category:Internet
 Microsoft launched the Brainwave project to provide faster chips to attract AI developers.


Microsoft said that these chips will speed up the image processing of the AI model compared with the Amazon cloud computing services AWS or Google cloud computing services. Microsoft, Amazon, Google, and other similar (Chinas Alibaba) companies offer a variety of chips, from the GPU to the Googles independent TPU. Microsoft says the use of a commercial FPGA made by a long-term partner, Intel (Intel), is another advantage in that customers will be able to set similar technologies locally, not on the cloud computing platform. This is part of Microsofts strategy to provide customers with more flexible services. We dont want to use FPGA to sacrifice performance or competitive advantage by using FPGA, said ProjectBrainwaves director, DougBurger. I think we hold our destiny in our own hands. These chips will initially be used in Microsofts Azure machine learning service at a cost of $0.42 per hour. Berg said that as time goes on, Microsoft will allow developers to directly access the chip. For years, Microsoft has been attracted by the FPGA. In 2014, the company said it would use them to speed up Bing search queries. The following year, Microsoft installed FPGA on all its servers purchased by Azure, initially to provide better network performance. Intel entered the FPGA field in 2015 and spent $16 billion 700 million to acquire Altera. Diane Bryant, a former Intel executive and now working in Google cloud, said that Microsofts support was an important reason for Intels acquisition of Altera. DianeBryant Based on early results, the AI system supported by Brainwave can reduce the false positive by 75%, that is, there is no problem but is sent to a rechecked circuit board. ProjectBrainwave seems to be a good choice, Litwack said, because it maintains a lower price while maintaining a high level of performance compared with GPU. For Jabil, its natural for Jabil to use Azure services, because the company has long been a Microsoft customer, from Windows to Office. Although Microsoft is using FPGA for public promotion, Berg does not reveal whether the company is following Google, creating its own chip to speed up the AI workload in the data center. He said: if you want to make an economic and reasonable decision, you certainly need to consider various options. (small) source: NetEase science and technology report editor: Hou Wei Cheng _NT4124 Based on early results, the AI system supported by Brainwave can reduce the false positive by 75%, that is, there is no problem but is sent to a rechecked circuit board. ProjectBrainwave seems to be a good choice, Litwack said, because it maintains a lower price while maintaining a high level of performance compared with GPU. For Jabil, its natural for Jabil to use Azure services, because the company has long been a Microsoft customer, from Windows to Office. Although Microsoft is using FPGA for public promotion, Berg does not reveal whether the company is following Google, creating its own chip to speed up the AI workload in the data center. He said: if you want to make an economic and reasonable decision, you certainly need to consider various options. (small)