Arnold GPU supports a set number of Arnold features, including arbitrary shading networks, SSS, hair, atmospherics, instancing, and procedurals. See here for a detailed list of Arnold GPU features and known limitations.
|Arnold GPU works on NVIDIA GPUs of the Turing, Volta, Pascal, and Maxwell architectures. Multiple GPUs will improve performance, and NVLink can be used to connect multiple GPUs of the same architecture to share memory (On Windows, we recommend enabling SLI as well).|
See the full list of supported GPUs.
Recommended NVIDIA drivers:
The very first time you render with the GPU, the GPU renderer has to create a cache for the types of shaders and objects used. This can delay the time to the first pixel for your first render. To avoid these one-time delays, we recommend doing some of this in advance by pre-populating the cache before you do any renders as this will fill in the cache with a subset of the most common shader/object combinations. The Arnold plugins (like MtoA) have Pre-Populate GPU Cache menu command, and kick has a -gpu_warm flag. Note that pre-populating the cache can take up to 15 minutes. The cache only needs to be re-populated after installing a new Arnold version, updating to a new NVIDIA driver, or changing the hardware configuration of GPUs on the system.
You can easily switch between CPU and GPU with a single click. In the Arnold plugins, you'll find a new Render Device setting in the Render Settings > System section.
Matching noise can take a little experimentation because Arnold GPU uses Camera (AA) sampling only. We recommend you also use Adaptive sampling. Here are some guidelines:
All textures must fit in memory. We recommend you use tiled and mip-mapped TX textures. If you're running out of memory, you can set a maximum resolution for textures in the Render Settings.