Please read the Set up GPU Rendering page prior to the FAQ.
The code base for V-Ray GPU differs from the code base for the CPU engine. The GPU code base is optimized for GPU hardware and supports a subset of the features of CPU rendering. For a complete list of supported features, please see the V-Ray GPU Supported Features page.
Yes. There are instructions how to do that in the GPU Setup section.
Check the list of devices in the GPU Devices menu in the Export tab of the V-Ray Renderer node. If you don't see your device on the list there (the list is empty), try to deselect the In process checkbox. This will make V-Ray run as a separate process, working around the limitation.
|
V-Ray will display a warning if the GPU used for monitoring is participating in the rendering process.
If you use many GPUs in DR nodes you may find that the network can become a bottleneck at some point, since the GPUs are producing data very fast. Increasing Bundle Size / Rays Per Pixel increases the size of the workload chunks that are being given to the DR nodes and helps reduce the communication between the client and the servers (you can try something like 192/32 or 256/64).
You can use free 3rd party tools like MSI Afterburner and EVGA Precision to monitor both GPU memory usage and utilization. RT GPU also reports how much memory it uses for textures/geometry/light cache/etc in the V-Ray log.
Yes, but within its limitations.
You can set V-Ray GPU as production renderer by the V-Ray Renderer node's parameters (Export tab > Mode).
By default, the V-Ray IPR node will use the settings from the V-Ray Renderer node. The main differences are that the IPR mode is interactive and the Production Rendering Mode is not.
In most situations, using Brute Force for your primary engine and Light Cache for your secondary will give you the best ratio between quality and speed in terms of not having to tweak a lot of GI settings to get a nice result. The render will have clean light bounces from accurate GI calculations, with minimal blotchiness or flickering for animated sequences
Procedural textures are supported for GPU bump maps, but because V-Ray GPU is a very different engine, the bump map itself can look a bit different compared to V-Ray.
If you get the following error message, this means that some or all of your video cards do not have sufficient memory to load the scene.
To solve this problem, you can enable texture resizing for GPU rendering and set a texture size limit. This will cause IPR GPU to use smaller textures for rendering to conserve memory. See more information on GPU texture resizing.
The difference in render speeds depends on the video cards and the CPUs that are compared, as well as on the scenes used as benchmarks. It is normal to achieve a big speed boost with the GPUs compared to the CPU with some scenes and to have equal performance with others.
Yes, you can speed up your rendering using DR on multiple systems with CUDA enabled devices.
Keep in mind that some drivers may have limitations and may not be able to use GPUs if there is no monitor attached to the machine or you are logged in via Remote Desktop (for example). Check the GPU vendor documentation for more information.
Yes. Note that GPU supports motion blurred frames that are integer numbers only. Make sure the starting frame of your render is an integer number for motion blur to work correctly. For example, starting the render from frame 0 will produce correct motion blur, but starting from frame 1.25 will produce incorrect motion blur.
There are several ways to start rendering with V-Ray IPR. First, create both the V-Ray Renderer and V-Ray IPR nodes in your output network. Make sure you have added a camera into your scene.
Though GPU rendering with CUDA on macOS devices is no longer supported, the advanced users can bring back the old behavior by using the following environment variable: VRAY_APPSDK_UNHIDE_DEVICES=1 written in houdini.env.
If you have registered for V-Ray for Houdini, then you automatically have access to the nightlies section. You can access them here. Note that you have to provide your Chaos Group credentials.