New Google Chip Streamlines Video Transcoding on YouTube

Google’s YouTube infrastructure team has built a Video (trans)Coding Unit (VCU) to streamline the transcoding of videos into the many versions required for a bandwidth-efficient and profitable website. Google software engineer Jeff Calow said the Argos chip offers “up to 20-33x improvements in compute efficiency compared to our previous optimized system, which was running software on traditional servers.” The VCU package is a full-length PCI-E card with two Argos ASIC chips under an aluminum heat sink.

Ars Technica reports that, “the cards are specifically designed to slot into Google’s warehouse-scale computing system.” According to Google, “each encoder core can encode 2160p in real-time, up to 60 fps using three reference frames.”

In YouTube’s system, each compute cluster “will house a section of dedicated ‘VCU machines’ loaded with the new cards, saving Google from having to crack open every server and load it with a new card.” With the cards, video workloads like 4K video “can be available to watch in hours instead of the days it previously took.” Even factoring in the R&D to create VCU, Google said it will save money.

It notes that, “because YouTube is the world’s biggest video site, keeping it running was originally seen as an impossible task until Google bought the company in 2006 … [which has since] aggressively fought to keep the site’s cost down.” Now, the company’s “primary infrastructure problem … is providing video that works just right for your device and bandwidth while maintaining quality,” which boils down to the right codec and resolution.

Due to the huge number of devices, YouTube has to transcode a single video into dozens of versions; it currently offers nine resolutions and “multiple codecs.”

Using the most advanced codecs saves on bandwidth (and costs), which “means Google only gets to use the best codecs on new devices, and it needs to keep copies of the video around in older codecs for older devices.”

Current devices “usually get the efficient VP9 codec, while the more compatible H.264 is kept around for devices that aren’t on the cutting edge.” YouTube “generally supports devices going back almost 10 years, including low-res flip phones. At the same time, the company needs to upgrade to new codecs as soon as possible to keep bandwidth costs down.

As a result, “codecs are so important to YouTube’s success that Google actually takes a lead in developing them,” including having bought codec developer On2 Technologies in 2009, to get the VP6 codec used in Flash video. Google has since “been a major codec developer.” Since developing VP8 and VP9, it is now “moving on to its next codec,” AV1, which was created via an industry coalition.

Calow said that, “one of the key things that we’re doing in the next-generation chip is adding in AV1, a new advanced coding standard that compresses more efficiently than VP9 and has an even higher computation load to encode.” Currently, AV1 is “experimentally available on YouTube and several other video sites, but mass usage is currently held up by client support.” It is also “already being phased into Google’s server farms.”