This means that users of V-Ray 3.6 will be able to split renders across both GPUs and CPUs simultaneously. So if you’re using a high-powered GPU workstation you can share the load with a CPU renderfarm to get a dramatically reduced render time.
What makes this news particularly exciting is addition of NVLink to the list of V-Ray’s supported hardware features. NVLink is a technology that shares GPU memory across multiple cards, creating a combined pool of RAM. At Creative Rooms and Autodesk University this month we were showing off two GP100s connected with this new technology to render a 32,000,000 polygon scene in 30 seconds flat at HD.
At the moment, the GP100 is the only card that supports NVLink but NVIDIA have suggested that more will be announced in the near future.
Now that hybrid rendering has been introduced as well it gives artists more flexibility in their approach to renders. The idea is that V-Ray will support both CPU and GPU, or a combination of both. It’s this last fact that has us talking.
Imagine being able to split the heavy lifting between your CUDA cores - sharing combined GPU memory via NVLink - and your CPUs. Or using a CPU renderfarm to do the heavy lifting while you’re using GPUs to see images on your local machine and vice versa. The days of being restricted by hardware are drawing to a close and the era of software-defined rendering has started. And it’s an exciting place to be.
Keep up to date with the latest news in VFX by signing up to our newsletter below.