Use of video streaming encoder playing cards within the information heart is on the rise, and AMD is the newest to deal with the calls for of high-volume streaming.
Even earlier than the pandemic pressured everybody to earn a living from home, videoconferencing utilization was climbing. As soon as colleges and companies turned depending on Zoom calls, video streams began clogging information facilities and community pipes throughout the nation. Reliance on video amongst customers additionally took off as TikTok, Twitch, and Fb turned broadcast platforms.
With customers demanding broadcast-quality video – nobody desires blurry, blocky, poor decision – enterprises are left to take care of elevated pressure on server CPUs.
Cue video streaming playing cards. Intel not too long ago got here out with one known as Arctic Sound, primarily based on its Xe GPU structure. Now AMD is releasing its personal accelerator known as the Alveo MA35D video processing unit (VPU).
The MA35D offloads video processing features from the CPU to the VPU, minimizing information motion between the CPU and VPU, lowering general latency, and maximizing channel density with as much as 32x 1080p60, 8x 4Kp60, or 4x 8Kp30 streams per card
The platform delivers as much as a 52% discount in bitrate for bandwidth for AV1 transcoding financial savings versus a comparable software program implementation. That’s as a result of on the core of the processor are two customized AV1 encoder blocks constructed utilizing a 5nm course of know-how node. The result’s the flexibility to transcode as much as 32 1080P/60 bitstreams at solely 1W per channel.
The cardboard helps a wide range of well-liked video streaming codecs, together with H.264/AVC, H.265/HEVC, VP9, and AV1, the final of which is changing into a well-liked streaming protocol for its excessive compression, low latency, and royalty-free utilization.
The MA35D replaces AMD’s Alveo U30, which was FPGA-based. The MA35D makes use of a customized ASIC with two devoted AV1 processors on the chip. In comparison with the U30, the Alveo MA35D delivers as much as 4 occasions increased channel density, 4 occasions max decrease latency in 4K decision, and 1.8x larger compression effectivity.
Kevin Krewell, principal analyst with TIRIAS Analysis, stated that with the explosion of video streaming for conferences and interplay, a video accelerator is required. Video compression in and of itself is just not notably taxing to the CPU, however if you multiply it by tons of or 1000’s of customers, that provides up.
“These playing cards are designed for information facilities, the place you are coping with a large quantity of video streaming out and in. So the effectivity of a {hardware} encoder is way larger than any CPU or attempting to do it in software program,” he stated.
AMD switched from an FPGA to an ASIC design after getting suggestions from cloud suppliers doing video, he stated. “Pretty much as good because the FPGA was at performing encoding of video streams, an ASIC primarily based on that very same know-how will at all times be extra environment friendly. Plus the actual fact they went to a 5 nanometer course of node made it extra environment friendly,” he stated.
The AV1 protocol is rising in reputation as a result of it’s an open commonplace and no royalties must be paid to make use of it, making it a extra interesting compression protocol than H.264, and that advantages the Alveo card, Krewell added.
The Alveo MA35D card is sampling now and will probably be accessible in manufacturing in Q3 of this 12 months.
Copyright © 2023 IDG Communications, Inc.
