Amazon to do an Apple, announces shift of Alexa work from Nvidia to Inferentia chips
After Apple, Amazon is all set to switch to in-house chips and has announced that a part of its Alexa work will shift from Nvidia chips to Inferentia ones. This is going to make work faster and cheaper.
Amazon has announced that it is going to move a portion of its computing work for smart voice assistant Alexa to its own in-house chips, according to a report in Financial Express. The move will see Amazon using their own Inferentia chips instead of the Nvidia chips that it has been using thus far.
This move is obviously bad news for Nvidia but for Amazon, it is going to make their work cheaper and quicker.
Also Read: Amazon launches Echo, Echo Dot, Echo Dot with Clock in India: Check out price, specs here
As it works right now, when an Echo user asks Alexa a question, the query is received by one of the data centres set up by Amazon where it is processed over several steps. Once the computer responds to the query the response is translated into audible speech so as the Echo can respond to the user.
This processing was being done by Amazon using Nvidia chips. Now, most of this computing work is going to be rerouted to Amazon's own custom-made chip called Inferentia. Inferentia was first announced in 2018 and the chip has been specially designed to speed up machine learning tasks like recognition of images or text to speech translations at large volumes.
Also Read: Amazon's Alexa app will work as an in-car display for Echo Auto soon
Big tech companies like Microsoft, Google, Amazon etc who are in the business of cloud computing are the biggest computing chips buyers which is a major benefit for companies like Intel and Nvidia.
However, now, several companies are moving away from third-party chip makers to their own in-house solutions
Just before Amazon announced its move, Apple introduced its own Apple Silicon M1 chip and announced three new Macs which will come with the M1 instead of the Intel chips it has been using so far.
Amazon has “justified” their own move by stating that they found a 25% improvement in latency and 30% reduction in cost by moving some of the work to Inferentia.
Follow HT Tech for the latest tech news and reviews , also keep up with us on Twitter, Facebook, Google News, and Instagram. For our latest videos, subscribe to our YouTube channel.