According to reports, Facebook is developing its own server chips.
However, while Facebook hasn’t explicitly verified the report, the company hasn’t ruled it out.
Large-scale service providers such as Facebook are building their own chips to power the data centers that provide their services, joining the growing number of companies doing so.
According to reports, Google, Amazon, and Microsoft have all spent millions of dollars to develop highly-tuned bespoke silicon that is tailored to meet the needs of their respective applications’ requirements.
Apparently, Facebook has now joined the party and is creating a chip that would enable machine learning for activities such as recommending content to consumers, according to reports. The Information mentions a number of individuals who appear to be familiar with the initiative.
The business, according to a Facebook spokesman, is continually looking for methods to improve compute performance and power efficiency, both with its silicon partners and via internal efforts. The spokesperson did not explicitly address whether the company is building proprietary silicon or not.
Standard practice in the industry
According to unnamed individuals who spoke to The Information, Facebook is building a suite of bespoke chips to support its many services rather than a single unique chip.
One of the specialized processors, for example, is allegedly being developed to support machine learning (ML) activities, such as the algorithm that handles Facebook’s content suggestions algorithm. Another is being modified to aid in the transcoding of videos in order to increase the quality of live-streamed video material.
As Engadget reports on the development, this isn’t the first time Facebook has experimented with the notion of bespoke silicon. The publication cites Facebook’s ambitions to build an application-specific integrated circuit (ASIC) for video transcoding and inference work in 2019.
With the growth of Ultra HD video, it should come as no surprise that platforms are looking at specialized processors that may aid in the acceleration of the process of processing and optimizing the content without cluttering the transmission pipeline.
YouTube material is processed and delivered more efficiently thanks to a special chip that Google just installed in its data centers.