Google builds its own chips to turbo-power AIs
Intel might worry, but Google says it won't commercialise its AI processors
Google is creating its own chips to power its AI machines, rivalling traditional silicon manufacturers like ARM and Intel.
The tech giant decided not to source its chips from other providers, instead building its own customised processors, dubbed Tensor Processing Units (TPUs), for its machine learning software. Google believes the TPU chip will lead to increased efficiency in AI processing.
"We've been running TPUs inside our data centers for more than a year, and have found them to deliver an order of magnitude better-optimized performance per watt for machine learning," Google said in a blog post. "This is roughly equivalent to fast-forwarding technology about seven years into the future (three generations of Moore's Law)."
Since AI has only recently found commercial use cases, companies that process AI functions are still solely using graphics processing units (GPUs) to do the job. While this is not a problem, GPUs were originally designed to facilitate the needs of gaming and graphic-based applications, therefore Google felt the need for a specialised chip with AI in mind.
"TPU is tailored to machine learning applications, allowing the chip to be more tolerant of reduced computational precision, which means it requires fewer transistors per operation," Google said.
As a result the TPUs complete tasks designated for AI with ease compared to their GPU counterparts. According to Google, TPUs and GPUs work alongside each other to facilitate the neural networks, which are required for AI-based machine learning.
However, Google's TPUs will likely entirely replace GPUs once they are further tailored to Google's computation needs.
"TPUs already power many applications at Google, including RankBrain, used to improve the relevancy of search results and Street View, to improve the accuracy and quality of our maps and navigation," Google said.
Google does not plan to sell its AI-based chips to other companies, reports Wired, a move that would likely threaten the business of chip firms like Intel, nVidia and ARM.
Top 5 challenges of migrating applications to the cloud
Explore how VMware Cloud on AWS helps to address common cloud migration challengesDownload now
3 reasons why now is the time to rethink your network
Changing requirements call for new solutionsDownload now
All-flash buyer’s guide
Tips for evaluating Solid-State ArraysDownload now
Enabling enterprise machine and deep learning with intelligent storage
The power of AI can only be realised through efficient and performant delivery of dataDownload now