AWS for M&E Blog
Optimize GPU costs for video editing and CG/VFX with Amazon AppStream 2.0
Having the correct number of GPU workstations that are secure, cost efficient, and high-performance for video editing and film production can be a challenge. With Amazon AppStream 2.0, using the multi-session fleet feature, it doesn’t have to be.
The challenges of securing GPUs for video editing and CG/VFX production
In professional video editing and computer generated/visual effects (CG/VFX) production environments, many artists use high-performance GPU workstations for their daily tasks. For example, high-performance GPU workstations are essential for several processes during a film production including 3D modeling, lighting, compositing, video editing, and color grading.
As production environments can change daily, a project may require additional production lines to be added in order to meet the customer demands and aggressive delivery schedules. This creates a challenge for artists, as many GPU workstations impose limits on the number of additional workstations that can be purchased or rented to support rapid increases or decreases in production demand.
Another challenge in CG/VFX/video editing relates to resources not being used to their full potential. This occurs when resources are secured for high-load tasks, such as rendering, but aren’t needed after those tasks are complete—leaving them to sit idle.
One solution is to use scalable cloud GPU instances with pay-as-you-go pricing. The advantage of GPU instances in the cloud is that the number of instances can flexibly increase or decrease. However, if multiple artists want to use a high-performance GPU exclusively for a long period of time, this may introduce higher operating costs.
Solution
To address these challenges, Amazon Web Services (AWS) offers a way for artists to optimize costs and share resources in a secure environment via AppStream 2.0.
AppStream 2.0 is a fully managed AWS End User Computing (EUC) service designed to stream software-as-a-service (SaaS) applications and convert desktop applications to SaaS without rewriting code or refactoring the application. Artists can use AppStream 2.0 to access cloud-based machines with video editing and CG/VFX software installed. The service also supports a multi-session fleet feature (multi-session), which allows multiple users to share a single GPU instance.
AppStream 2.0 multi-session users share compute, memory, storage, and system software associated with a given instance, while continuing to enjoy the ability to auto-scale resources based on actual usage. By having multiple artists share resources, such as CPUs and GPUs, it is possible to use the resources more efficiently and lower costs. Multi-session is particularly beneficial if high-performance workstations cannot be provided to each artist due to budget or other reasons.
AppStream 2.0 offers more than 10 GPU instance types (G5 and G4) to choose from, and artists can control the number of users accessing a single machine resource. Therefore, they can build an appropriate cloud editing environment based on the team size, workload, and project budget.
Testing Blender and Unreal Engine in AppStream 2.0
To illustrate and validate using Appstream 2.0 for video editing and CG/VFX use cases, the team at AWS did a test. Here are the results of a multi-session, simultaneous test of Blender 4.1 and Unreal Engine 5.4.2. For the test AWS used a stream.graphics.g5.xlarge GPU instance.
The following image (Figure 1) shows the screen of Artist A using Unreal Engine on the left, and the screen of Artist B using Blender on the right.
By checking the resource usage in the Task Manager, users can confirm that both artists are using the same resource. The Task Managers indicate that the users are sharing CPU and GPU resources. When rendering using Unreal Engine, the GPU – 3D usage rate increased in the task manager (Figure 1, red box on left side). At the same time, when rendering using Blender Cycles, the GPU – CUDA usage rate increased (Figure 1, green box on right side). Both artists were able to render while sharing resources.
The test proved that despite using a single instance resource, multiple artists were able to complete their work and cost efficiency improved. This is a clear example of efficient use of GPU workstations.
Considerations
It is important to recognize that some software may be prohibited from running on the cloud. So, artists need to check the license of the software they want to run on AWS. Also, artists should be aware that some software only allows one process to run on a single machine. When such software starts a second time, it may alert the user with an error message. In this case, it cannot be used with multi-session. When setting up an AppStream 2.0 multi-session environment, please verify the licensing terms and operational requirements of the software you plan to use in advance.
Conclusion
There are many resource challenges in professional video editing and CG/VFX production environments. Using AppStream 2.0 multi-session functionality can help optimize resource utilization, which can support the GPU needs of multiple artists and potentially have significant benefits to cost savings.
If you are interested in cost optimization and resource sharing for video editing and CG/VFX workloads, please consider using AppStream 2.0.
Contact an AWS Representative to know how we can help accelerate your business.