ASIC-Based 8K Encoding can Save Power and Money
FIGURE 2. NETINT’S QUADRA VPU.

ASIC-Based 8K Encoding can Save Power and Money

Back before “politically correct” was a phrase or even a concept, it wasn’t unusual to hear someone say, “there’s more than one way to skin a cat.” Speaking totally metaphorically, of course, and with apologies in advance to dour cat lovers without a sense of humor, that still is the case.

Here we look at two ways to produce an 8K60 contribution stream.

The first alternative was presented by video guru Colleen Henry in her Demuxed talk, Building an 8K Encoder + Live Streaming Platform, which you can find here. As Colleen describes in the presentation, “the Ryzen Threadripper 3990x is the chip that we went with. It has 64 cores and 128 logical threads. And I’m sitting next to it right now, and I’m going to pull out my thermal camera and just show how hot this thing is (she was joking). It’s hilarious. But you have to have a very multithreaded job for it. But, you know, the SVT line from Intel and x264 does a good job of it.”

You can correctly assume from Colleen’s comments that encoding 8K video is a very challenging operation for any computer and that it would require a substantial portion of the 128 cores available on that processor to get it done, plus the two NVIDIA graphics cards specified in the video.

Table 1 includes the major components incorporated into Colleen’s 8K system, with cost and power requirements. Colleen did mention in her talk that less expensive technology can be used, but the main point is straightforward; you will need a very hefty and power-hungry system to produce a single 8K60 contribution stream. And, if you need ten contribution streams, you’ll have to order up to ten separate units.

Colleen’s system was for H.264, VP9, and AV1; she doesn’t mention HEVC. One system capable of live HEVC encoding is the Intel system assembled for the 2020 Olympics, documented here. Considering only the four Intel Xeon 8380H Scalable CPUs deployed on the system, power consumption would have been 1,000 watts, and the CPUs would have cost over $48,000. While producing a single stream for a one-off event like the Olympics may be justifiable, it’s not for 24/7 operation for multiple channels.

The Quadro Alternative

What’s the alternative? NETINT’s ASIC-based Quadra Video Processing Unit (VPU) can output a single 8K60 stream in H.264, HEVC, and AV1 formats and draws only 20 watts of power. The ASIC is available in both a PCI form factor and a U.2 form factor compatible with inexpensive expansion racks typically used for SSD storage.

ASIC-Based 8K Encoding can Save Power and Money
FIGURE 2. NETINT’S QUADRA VPU.

Table 2 shows what the math looks like for a single and ten units. Because Quadra’s operation is self-contained, you can run it on most low-end computers. CPU utilization is generally very low, which keeps the main CPU’s power usage down during operation.

TABLE 2. COST AND POWER CONSUMPTION FOR AN ASIC-DRIVEN SYSTEM.

For ten 8K streams in H.264, HEVC, or AV1 formats, you can buy a workstation with a U.2 expansion rack and install ten Quadras. Again, the expansion rack is the expensive part; you won’t need a hefty CPU, and the CPU will be mostly idle during operation. You can gauge the CAPEX and power savings yourself.

Summing Up

No criticism or disrespect for Colleen; she’s an incredible industry resource who makes code and other resources available on her GitHub site. Her company, Meta, has been at the forefront of the push toward green computing and recently endorsed and adapted the LCEVC codec as part of this focus. Her article is a teaching experiment designed to help readers and watchers learn to create their own streaming encoder and infrastructure; not a recommended approach for volume 8K contribution encoding. Ditto for the Intel system was clearly a technology demonstration, not a production system.

Still, these systems help make the point that while software-based 8K encoding may be possible, it’s not practical or environmentally conscious and certainly isn’t scalable. That’s where ASIC-based solutions start to shine.

About Jan Ozer

Avatar photo
I help companies train new technical hires in streaming media-related positions; I also help companies optimize their codec selections and encoding stacks and evaluate new encoders and codecs. I am a contributing editor to Streaming Media Magazine, writing about codecs and encoding tools. I have written multiple authoritative books on video encoding, including Video Encoding by the Numbers: Eliminate the Guesswork from your Streaming Video (https://amzn.to/3kV6R1j) and Learn to Produce Video with FFmpeg: In Thirty Minutes or Less (https://amzn.to/3ZJih7e). I have multiple courses relating to streaming media production, all available at https://bit.ly/slc_courses. I currently work as www.netint.com as a Senior Director in Marketing.

Check Also

Announcing Free Course on Controlling the AMD MA35D with FFmpeg

I’m pleased to announce a new free course, MA35D & FFmpeg Quick Start: Essential Skills …

Choosing the Best Preset for Live Transcoding

When choosing a preset for VOD transcoding, it almost always makes sense to use the …

There are no codec comparisons. There are only codec implementation comparisons.

I was reminded of this recently as I prepared for a talk on AV1 readiness …

Leave a Reply

Your email address will not be published. Required fields are marked *