How to assign transcoding to Alveo™ U30 accelerator card
In this post, we’ll explore how to maximize AMD Alveo™ U30 cards using Callaba for video transcoding.
On AWS VT1 instances or in on-premises setups, Callaba helps distribute workloads efficiently across each Xilinx transcode device for smooth video processing and streaming.
Understanding AMD Alveo™ U30 cards
Each AMD Alveo™ U30 card includes two transcode devices.
Think of these as two separate processors on each card that divide the workload.
So if you have U30 cards, you have four transcode devices, and so on.
Performance Specifications
The performance of AMD U30 cards varies based on the resolution and frame rate of your video streams.
Here’s a quick overview:
- 2 streams at 4K, 60fps
- 8 streams at 1080p, 60fps
- 16 streams at 1080p, 30fps
- 32 streams at 720p, 30fps
So if you’re working with 1080p at 30fps, each card can handle 16 streams total.
Since each card has two devices, that means 8 streams per device.
By balancing your load across devices, you can use each card to its full capacity without risking overload.
Distrubute streams to devices in Callaba
Now let’s get into the specifics of setting up and assigning devices for efficient transcoding.
Let’s say we’re using an AWS vt1.6xlarge instance.
It has four transcode devices.
Each device can handle up to eight 1080p 30fps streams.
So, knowing all this, to distribute the workload effectively, we'll assign the streams across the devices as follows:
- Streams 1-8: Use device 0 (the default device).
- Streams 9-16: Assign to device 1.
- Streams 17-24: Assign to device 2.
- Streams 25-32: Assign to device 3.
This distribution ensures each transcode device remains within its capacity limit and helps prevent overload on the default device (device 0).
Callaba makes this easy by allowing you to specify which device each stream should use during the transcoding setup.
Example setup in Callaba
Let’s go through an example. Here I have set up eight Web Players with media-accelerated transcoding.
Now, let's add the ninth Web Player (you can do the same thing with re-streams)
Click "Add New"
Name your Player
Select your stream source
Open Video Settings and set video transcoding to Transcode via media accelerator.
In the Device field, assign device 1 for the ninth transcoding process.
With this setup, you can control the workload distribution, ensuring each device handles an optimal amount of work, enhancing your transcoding efficiency and reliability.
What is AMD Alveo™ U30 media accelerator?
AMD Alveo™ U30 media accelerator is a specialized card built for high-efficiency video transcoding.
The U30 can manage high channel density—many streams at once—at low cost per channel and low power usage (only 18-25 watts), which is ideal for companies that need to stream a lot of video.
How accelerator works
The U30 takes over heavy video processing tasks from the CPU.
It encodes, decodes, and scales video using built-in hardware for formats like H.264 (AVC) and H.265 (HEVC), which are common in streaming and media. The U30 card supports adaptive bitrate streaming, adjusting video quality based on network conditions.
Benefits of the U30 over CPU transcoding
CPUs can process video, but handling many streams at once requires a lot of power, which makes it expensive for high-volume tasks. A CPU-heavy setup also needs extra hardware to manage multiple streams, while the U30 does this on its own.
By moving these tasks to a dedicated card, companies can reduce CPU load, lower power costs, and let the CPU focus on other tasks, making everything run faster and more efficiently.
Advantages over GPU-based transcoding
While GPUs can handle video tasks, the U30 has unique advantages. Unlike general-purpose GPUs, the U30 is purpose-built for video encoding and decoding. This specialization makes it more efficient for large-scale video tasks.
It also uses less power than GPUs, which typically require much more energy to process multiple streams.
The U30 uses only about 18-25 watts, compared to most GPUs that can use up to 300 watts or more, saving on both power and cooling costs.
Where to buy the Alveo U30 accelerator
For on-premises usage, you can purchase the AMD Alveo U30 media accelerator through AMD’s official website and various authorized distributors.
In cloud environments, the Alveo U30 is accessible via Amazon EC2 VT1 type instances, offering a ready-to-use solution for scalable video transcoding without the need for dedicated hardware on-site.
Why Callaba?
Callaba complements the AMD Alveo U30 by providing a streamlined way to allocate and manage transcoding tasks across devices.
With Callaba, you can specify which streams each transcode device on the U30 card should handle, optimizing resource usage and preventing any one device from becoming overloaded.
This setup enables users to get the most out of their media accelerator, whether deployed on-premises or in the cloud.
For AWS VT1 Instances
On AWS, Callaba software can be launched using VT1 instances equipped with AMD Alveo™ U30 cards.
These instances are specifically optimized for video transcoding, providing a scalable, cloud-based solution.
For full instructions, visit How to launch Callaba on AWS.
For Self-Hosted (On-Premises) Setup
For self-hosted setups, you can install Callaba on a server equipped with an AMD Alveo™ U30 card.
More details can be found in the self-hosted installation guide.
Both setups enable high-efficiency video processing, allowing you to fully use AMD’s media acceleration capabilities with Callaba.
Learn more
- How to launch Callaba on Amazon Web Services
- How to create a playlist of videos with Pay Per View monetization
- Multilingual Web Player
- Set up SRT backup stream in case of main stream interruption
Follow us on social media
LinkedIn • Reddit • Instagram • Facebook • Youtube • X (Twitter) • Discord
