Partha Ranganathan and other Google engineers examined the big picture and discovered that transcoding (for YouTube) was consuming a significant portion of compute cycles in the company’s data centers. Google’s off-the-shelf chips weren’t particularly good at specialized tasks like transcoding.
Transcoding is used by YouTube’s infrastructure to compress video to the smallest possible size for your device while presenting it at the highest possible quality.
They required an application-specific integrated circuit, or ASIC, which is a chip designed to perform a single task as effectively and efficiently as possible. Bitcoin miners, for example, use ASIC hardware that is designed specifically for that purpose.
“The thing that we really want to be able to do is take all of the videos that get uploaded to YouTube and transcode them into every format possible and get the best possible experience,” said Scott Silver, VP of engineering at YouTube.
It didn’t take long to persuade upper management of the benefits of ASICs. The company’s first video chip project was approved after a 10-minute meeting with YouTube CEO Susan Wojcicki.
The company’s first video chip project was approved after a 10-minute meeting with YouTube CEO Susan Wojcicki.
Google began deploying its Argos Video Coding Units (VCUs) in 2018, but the project was not publicly announced until 2021. Google claimed at the time that the Argos VCUs outperformed traditional server hardware running well-tuned transcoding software by a factor of 20 to 33.
Google has since activated thousands of second-generation Argos chips in servers around the world, and at least two follow-ups are already in the works.
The obvious reason for designing your own chip for a specific purpose is to save money, but this isn’t always the case. In many cases, large technology companies use custom chips to gain a strategic advantage.
Consolidation in the chip industry also plays a role, as there are now only a few custom chipmakers to choose from in a given category, making general-purpose processors that aren’t particularly good at specialized tasks.
According to Jonathan Goldberg, principal at D2D Advisory, what is really at stake is control of the semiconductor companies’ product roadmap. “And so, they build their own, they control the road maps and they get the strategic advantage that way,” Goldberg added.
Google’s Argos custom chip isn’t the only one. In 2016, the company unveiled the Tensor Processing Unit (TPU), a custom ASIC designed to power artificial intelligence applications.
Google has since released more than four generations of TPU chips, giving it a competitive advantage in the field of AI. Google also used a custom-built Tensor SoC in its Pixel 6 series of smartphones, bringing hardware and software under the same roof for its mobile line.
To read our blog on “YouTube adds watermark to short videos to depress cross-posting,” click here
