Google also found that over time, it was able to get its hardware encoders to beat software encoding. Google also focuses on building clusters of machines, but it is fairly clear that Google can put a large number of VCUs to work. The net impact is that the VCU is more efficient than a dual socket 2017-era Intel Xeon syhstem for h.264 and five servers for VP9. We covered this in the previous article on the VCU, but here is the architecture from Google’s whitepaper.
HC33 Google VCU VCU FirmwareĪt the system level, these are deployed with 20 VCUs per system. The VCU has its own firmware that runs the ASIC and allows userspace choices of codecs for example. Here is the VCU network on chip topology: HC33 Google VCU NoC Topology Here is the drill-down into the ASIC: HC33 Google VCU ASIC There are ten encoder cores adn three decoder cores along with LPDDR interfaces. The VCU design goals included maximizing the utilization. Google has many applications for video such as YouTube, Google Drive, and more.
Overall, Google seems to be very focused on using VCU ASICs in the future. HC33 Google VCU Design By High Value Problems It also kept the limited team working on high-value features and problems.
This meant that Google could design the VCU using a higher level language (C++) making the development much faster. It used a high-level synthesis design flow. Google has teams that design hardware in addition to software. This has the software programmable registers. This has the interfaces to read frames and decompress/ compress the frame buffer: HC33 Google VCU Video Encoder Core 6 Here is the reconstruction and entropy coding: HC33 Google VCU Video Encoder Core 5 Here is the motion search and rate-distortion otpimization engine: HC33 Google VCU Video Encoder Core 4 Here is the temporal denoiser/ filter: HC33 Google VCU Video Encoder Core 3 Here is the pre-processing: HC33 Google VCU Video Encoder Core 2 Since we are doing these live, we are just going to show the slides for the video encoder cores. HC33 Google VCU Cut Down YouTube Compute Cycles The impact of deploying the VCU ASICs was a massive decrease in CPU use. HC33 Google VCU Why Develop Own Video Chips Near Parity To SW Quality It also wanted to get close to software encoding, but with a lower power and faster ASIC. HC33 Google VCU Why Develop Own Video Chips Wants NA HC33 Google VCU Why Develop Own Video Chips NeedsĪs a result, Google wanted a number of capabilities not available from commercial products. Google realized it needed to create its own chips to handle higher bitrate source video. HC33 Google VCU Video Is Getting Harder To Compress Times Increase Ultimately, saving 30-40% on bandwidth is a good goal, but it requires a lot of compute on a growing problem to make that a reality. Also, the higher levels of compression generally require more compute. Instead it is to work with content that continues to get higher resolution and higher framerate. However, the challenge is not to encode the same content. HC33 Google VCU Video Is Getting Harder To Compress Each codec gets more efficient at compressing video leading to smaller file sizes and smaller streams. There are a number of different compression and encoding codecs. HC33 Google VCU Video Is A Majority Of Internet Traffic It now says that video is more than 60% of overall video traffic and as video gets higher resolution and framerates, this increases bandwidth needs. Specifically, it is one of the firms directly impacted by the overall types and mix of Internet traffic. Google VCU Video Coding Unit at Hot Chips 33 We are doing this one live, so please excuse typos. At Hot Chips 33, the company gave more insight into the solution than it did in the original paper. We previously covered the Google VCU or Video Coding Unit in Google YouTube VCU for Warehouse-scale Video Acceleration.
0 kommentar(er)
