Amazon relocates machine learning tasks to its custom designed chips, away from NVidia Corp

In a development that underscores cost cutting measures adopted by Inc, the e-commerce giant said it has shifted some computing power for its Alexa voice assistant to its own custom-designed chips, away from chips supplied by Nvidia Corp.

When consumers ask devices equipped with Alexa, including Amazon’s Echo line of smart speakers, a question the query is sent to Amazon’s data centers for processing. When Amazon’s computers come up with an answer, the reply, in a text format, is translated into audible speech for the voice assistant.

Previously, Amazon handled this bit of computing using Nvidia chips; now it has shifted the “majority” of this computing requirement to its own “Inferentia” chip, which it had first announced in 2018. The Inferentia chip has been custom designed to speed up machine learning tasks including translating text to speech or recognizing images.

Cloud computing tech giants, including Amazon, Microsoft Corp, and Alpahbet Inc’s  Google are some of the biggest buyers of these machine learning specialized chips, which are driving the boom in data centers.

Major tech companies are increasingly designing their own custom designed chips, with the latest case in point being Apple which on introduced its first Mac computer with its own CPU earlier this week on Tuesday. Earlier it used chips from Intel.

In a statement Amazon said, following the shift to its own Infertia chip, processing tasks by Alexa got a 25% reduced boost in latency, at a 30% lower cost.

In a separate statement, Amazon said, “Rekognition,” its cloud-based facial recognition service, has also adopted its Inferentia chips. However, Amazon did not say which chips its facial recognition service had used previously or the percentage of shift of computing power to its own chips.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s