New GPU chipsets are always being introduced, and the After Effects team does not qualify or recommend individual GPU chipsets, however, here are some guidelines you can follow to get the best GPU for your workflow.
- Individual GPU technologies are less important than overall GPU performance. After Effects supports OpenGL, OpenCL, CUDA, and Metal to varying degrees. Choose a high-performance card that meets your individual budget and system needs.
- Premiere Pro utilizes the GPU more broadly than After Effects currently does, and its technology is shared with After Effects. The list of recommended GPUs for Premiere Pro (see Premiere Pro system requirements) is a good place to start.
- Other applications in your workflow may have a GPU requirement that is higher than After Effects. Take all of them into consideration.
- Check if you have Multiple GPUs in the same machine.
- Check if you have unsupported GPUs on your Mac machine.
After upgrading to After Effects version 17.x or later, there may be driver issues, and you may need to upgrade your driver.
Some of the driver issues that you could face are:
- System incompatibilities that are known to cause instability and crashes that lead to data loss.
- The current version of your network device software may cause issues with your Adobe application
- Intermittent crash while editing.
- You can get error messages such as, "This version of your operating system is incompatible with your Adobe application."
- No previews, garbled previews, frame drops, performance issues including slow playback or frame glitches.
This article explains what is needed for using CUDA graphics with the 2019 versions of After Effects (17.0 and higher).
Adobe strongly recommends updating to NVIDIA driver 451.77 or later when using After Effects.
NVIDIA CUDA graphics acceleration requires CUDA 10.1 drivers.
CUDA is not a requirement for running the Adobe video apps, but if you prefer CUDA graphics acceleration, you must have CUDA 10.1 drivers from NVIDIA installed on your system before upgrading to After Effects versions 17.0 and later.
These drivers are updated regularly so check the NVIDA website to be sure you have the most current version for your GPU.
You can find the latest GPU drivers here:
- Requires macOS 10.13.6 (most recent version of High Sierra).
- A current NVIDIA GPU with at least 4 GB of memory.
- NVIDIA display driver version 3188.8.131.52.40.128.
The current version of your NVIDIA drivers for macOS 10.13.6 do not support CUDA 10.1 and cause issues with your Adobe application. Adobe does not recommend upgrading beyond macOS 10.13.6 as Mac0S 10.14 (Mojave) does not currently support CUDA.
Driver version - 184.108.40.20676.
For more information, see How to manually install your Intel® Graphics Driver in Windows® 10.
There is a host of After Effects features that use GPU to accelerate rendering. To view these effects, select Project Settings > Video and Effects Rendering. For a list of GPU-accelerated effects and features, see GPU-accelerated effects.
Some third-party effects, like Element 3D by Video Copilot uses the GPU independently of After Effects. Refer to the documentation from the publisher for guidance on what GPUs and technology are supported. Effects such as Magic Bullet Looks, hook into the Mercury GPU Acceleration pipeline (such effects are also GPU-accelerated in Premiere Pro).
Mercury GPU Acceleration allows After Effects to render supported effects using the GPU, which can significantly improve render time.
You may recognize the Mercury name from Premiere Pro. After Effects uses the same technology that is used by Premiere Pro Mercury Playback Engine for rendering. (The playback engine in After Effects is otherwise different from Premiere Pro, so After Effects only uses the rendering component of that technology.)
Mercury GPU Acceleration is a project setting. To enable it, select File > Project Settings, click the Video Rendering and Effects tab, and set the Use option to Mercury GPU Acceleration. Depending on your computer and GPU, you may see multiple such options. After Effects supports the following GPU technologies:
- OpenCL (macOS and Windows)
- CUDA (Windows only, with a Nvidia GPU)
- Metal (macOS only, 10.12 and later)
NVIDIA CUDA is not supported in MacOS 10.14 and later. If you are using an Apple-authorized NVIDIA GPU, you can continue to use the Metal Mercury Playback Engine.
A couple of technical points worth noting about the above list:
- All of the VR effects such as VR Blur only work on the GPU. Unlike the other effects, they do not currently have a CPU fallback. We recommend a GPU with high VRAM, 4GB or better, to use these effects. Their advantage for VR over other effects is that they are seamless, and they wrap the ends of the VR image together. Also, some of them are useful on non-VR footage because they are wholly new to After Effects, like VR Chromatic Aberrations.
- Layer transforms and layer quality require layer motion blur to be enabled because by themselves, they do not render significantly faster on the GPU. But rendering motion blur on the GPU requires it to be aware of the transforms and quality, so these calculations are done on the GPU when motion blur requires it. This is an opportunity to point out that in a mixed CPU and GPU rendering environment, there is a performance cost to moving frames between CPU and GPU memory. If an effect is not faster to render on the GPU than on the CPU, you lose performance time while copying the frames back and forth.
- The Hardware Accelerate Composition, Layer, and Footage Panels option in Preferences > Display is enabled by default, and uses OpenGL to prepare the rendered frames for screen display during previews/playback. Once After Effects renders the frame, it next prepares that frame for display, taking into account the screen resolution, scaling, overlays like guides and layer handles, and color management. Specifically, View > Use Display Color Management, when a working space color profile has been enabled for the project. When the Hardware Accelerate option is disabled, After Effects processes all of that on the CPU, but the GPU can accelerate this process, especially color management. The GPU requirement for this is very low, and any modern video card with a small amount of VRAM is adequate.