Microsoft offers new AI chips to its cloud computing platform Want to recreate the success of "Wintel"?
In the 1980s and 1990s, Microsoft became the industry leader with the success of the Windows operating system running on Intel processors. Now, Microsoft hopes that another combination of hardware and software will help it regain its success and catch up with Amazon and Google in the competition to provide cutting-edge artificial intelligence through the cloud.
According to arsTECHNICA, Microsoft hopes to expand the popularity of its Azure cloud platform with a new computer chip designed for the era of artificial intelligence. From now on, Microsoft will provide Azure users with chips from the British startup Graphcore.
Founded in the UK in 2016, Graphcore has attracted a lot of attention from AI researchers and has received hundreds of millions of dollars in financing. The company said its chips will accelerate the computational speed required for artificial intelligence. But so far, the company has not disclosed the chips, nor has it shown the results of its early testing.
In December last year, Graphcore received $200 million in financing, including Microsoft. Microsoft is in desperate need of hardware to make its cloud services more attractive to customers of artificial intelligence applications. Unlike most chips used in artificial intelligence, Graphcore's chips are designed from scratch to support calculations that help machines recognize faces, understand speech, parse language, drive cars, and train robots.
Benchmarks released by Microsoft and Graphcore show that the performance of the chip is better than that of NVIDIA and Google's top-of-the-line artificial intelligence chips. Using code written specifically for Graphcore hardware may make it more efficient. They claim that some image processing tasks are many times faster on the Graphcore chip than their competitors. They also said that they can train a popular artificial intelligence language processing model, BERT, which is comparable in speed to any existing hardware.
BERT is very important for artificial intelligence applications involving language processing. Google recently said it is using BERT to power its core search business. And Microsoft is currently using Graphcore's chips for internal artificial intelligence research projects involving natural language processing.
According to Karl Freund of Moor Insights, which tracks the artificial intelligence chip market, the results show that the chip is flexible, yet advanced. Highly specialized chips may perform better than NVIDIA or Google chips, but their programmability is not enough for engineers to develop new applications. Graphcore does a good job of programming.
Freund added that the agreement with Microsoft is critical to Graphcore's business and provides customers with an entry point to try new hardware. But re-developing AI code for the new platform requires a lot of work, with a few exceptions, the chip's benchmarking is not enough to attract companies and researchers to abandon the hardware and software they are used to.
In recent years, the rapid development of artificial intelligence has deeply affected the computer chip market. With hundreds of simple processing core graphics chips, parallel digital computing can be completed more effectively. Nvidia made a fortune with the help of artificial intelligence wave. Google announced its development of its chip tensor processing unit (TPU) in 2017. Facebook is also developing its own AI chip. Amazon revealed that it is also entering the chip manufacturing field a year ago.
The wave of artificial intelligence has sparked a wave of hardware startups developing professional chips, some of which are optimized for specific applications, such as autopilot or surveillance cameras. But it is still difficult to open a piece of heaven and earth for itself.
According to arsTECHNICA, Microsoft hopes to expand the popularity of its Azure cloud platform with a new computer chip designed for the era of artificial intelligence. From now on, Microsoft will provide Azure users with chips from the British startup Graphcore.
Founded in the UK in 2016, Graphcore has attracted a lot of attention from AI researchers and has received hundreds of millions of dollars in financing. The company said its chips will accelerate the computational speed required for artificial intelligence. But so far, the company has not disclosed the chips, nor has it shown the results of its early testing.
In December last year, Graphcore received $200 million in financing, including Microsoft. Microsoft is in desperate need of hardware to make its cloud services more attractive to customers of artificial intelligence applications. Unlike most chips used in artificial intelligence, Graphcore's chips are designed from scratch to support calculations that help machines recognize faces, understand speech, parse language, drive cars, and train robots.
Benchmarks released by Microsoft and Graphcore show that the performance of the chip is better than that of NVIDIA and Google's top-of-the-line artificial intelligence chips. Using code written specifically for Graphcore hardware may make it more efficient. They claim that some image processing tasks are many times faster on the Graphcore chip than their competitors. They also said that they can train a popular artificial intelligence language processing model, BERT, which is comparable in speed to any existing hardware.
BERT is very important for artificial intelligence applications involving language processing. Google recently said it is using BERT to power its core search business. And Microsoft is currently using Graphcore's chips for internal artificial intelligence research projects involving natural language processing.
According to Karl Freund of Moor Insights, which tracks the artificial intelligence chip market, the results show that the chip is flexible, yet advanced. Highly specialized chips may perform better than NVIDIA or Google chips, but their programmability is not enough for engineers to develop new applications. Graphcore does a good job of programming.
Freund added that the agreement with Microsoft is critical to Graphcore's business and provides customers with an entry point to try new hardware. But re-developing AI code for the new platform requires a lot of work, with a few exceptions, the chip's benchmarking is not enough to attract companies and researchers to abandon the hardware and software they are used to.
In recent years, the rapid development of artificial intelligence has deeply affected the computer chip market. With hundreds of simple processing core graphics chips, parallel digital computing can be completed more effectively. Nvidia made a fortune with the help of artificial intelligence wave. Google announced its development of its chip tensor processing unit (TPU) in 2017. Facebook is also developing its own AI chip. Amazon revealed that it is also entering the chip manufacturing field a year ago.
The wave of artificial intelligence has sparked a wave of hardware startups developing professional chips, some of which are optimized for specific applications, such as autopilot or surveillance cameras. But it is still difficult to open a piece of heaven and earth for itself.