OptiX and NVLink FAQ – Chaos Help Center

This article covers information on setting up OptiX and NVLink for usage with V-Ray. It examines their purpose and answer questions such as:

What are OptiX and NVLink and what are they used for?
How to set OptiX and NVLink?
How does V-Ray benefit from OptiX and NVLink?
Which Nvidia driver is best for using V-Ray GPU?
How to determine if a GPU device is compatible with V-Ray GPU?

Overview

GPU rendering has gained popularity and new methods are introduced for boosting the performance, and for reducing some limitations.

This article explains the OptiX and NVLink technology, the setup required for proper functioning with V-Ray and their effect on the rendering process.

What is OptiX / RTX?

OptiX is a new hardware-accelerated ray-tracing technology introduced with NVIDIA's new Turing graphics card architecture. It offers hardware-accelerated ray-casting, which speeds up the overall rendering performance. Turing hardware devices are also known as RTX GPUs (RTX stands for Real-Time Ray tracing).

V-Ray GPU supports OptiX rendering on RTX cards for increased performance. The speed-up depends on the geometry and material complexity in the rendered scene with some projects rendering up to 40% faster. The more rendering effort is usually spent on geometry, rather than materials, the greater the speed boost.

To learn more about RTX acceleration in V-Ray, have a look at Philip Miller’s article for CG Labs.

What is NVLink?

NVLink is a wire-based hardware connection between GPU devices that allows multiple graphics cards to share their available memory. This is significant for GPU rendering, since it allows more memory to be used for large scenes, that otherwise would not fit on a single device.

There may be a slight decrease in speed performance with NVLink enabled. This is expected and comes from the overhead of accessing memory over longer distances.

To avoid performance hits, some of the frequently used data is kept on all devices. This means that the memory is not fully split between the graphic cards and some data is kept on all GPUs.

There are different types of NVLink connectors depending on the physical space between the GPUs and whether GeForce or Quadro cards are used. NVLink can only be used to bridge identical graphics cards.

Setting up NVLink for GeForce Devices

Requirements:

Identical GeForce devices with NVLink support
Enabled SLI-mode from Nvidia Control Panel
SLI Ready Motherboard

GeForce devices support only a single NVLink connection per workstation.

Enabling NVLink on GeForce

Open the NVIDIA Control Panel.
Go to "Configure SLI, Surround, and PhysX".
Under "SLI configuration", select "Maximize 3D performance" and click Apply.

After you have enabled NVLink, V-Ray will be able to use it. To verify that the set up is complete, run a test render and check the V-Ray Messages Log for the following message:

NVLink Detected!
Device[0] (Quadro GV100) is using NVLink with ID: 0
Device[1] (Quadro GV100) is using NVLink with ID: 0
Number of CUDA devices: 2

Setting up NVLink for Quadro Devices

Requirements:

Identical Quadro devices with NVLink support
All Quadro GPUs should run in a special TCC driver mode
Monitors should not be attached to a device that is running in TCC mode
A dedicated graphics card should be installed for a monitor instead
SLI Ready Motherboard

Switching Driver Mode to TCC

Switch the driver mode to TCC through the nvidia-smi command-line utility that is installed along with the NVidia drivers. Its default location is:

C:\Program Files\NVIDIA Corporation\NVSMI

To use nvidia-smi you need to start the Command prompt. Press Start and type cmd.

Use the following command to enter the directory of the nvidia-smi tool.

 cd C:\Program Files\NVIDIA Corporation\NVSMI

Use the following command to change the TCC mode:

 nvidia-smi -g {GPU_ID} -dm {0|1}

Where the value for dm is either:

0 = WDDM
1 = TCC

And {GPU_ID} is the unique ID number that nvidia-smi associates with each GPU.

To view the GPU_IDs run this command:

 nvidia-smi -L

The result is printed in the command prompt:

GPU 1: GeForce RTX 2080 (UUID: GPU-21600cd2-5f9f-fb16-e1dd-e45e99e85a01)

In this example, the {GPU_ID} is 1.

Sample Command:

nvidia-smi -g 1 -dm 1

To verify that NVLink is enabled, run a test render and check the V-Ray Messages Log for the following message:

NVLink Detected!
Device[0] (Quadro GV100) is using NVLink with ID: 0
Device[1] (Quadro GV100) is using NVLink with ID: 0
Number of CUDA devices: 2

V-Ray GPU System Requirements

Make sure your PC configuration meets the minimum system requirements to use V-Ray GPU.

CUDA and RTX requirements for V-Ray GPU:

NVidia GPU architecture requirements. Make sure that the GPU architecture meets the system requirements listed in the technical documentation.
Compute capability requirements. All NVidia GPUs have a compute capability value defined by NVidia. The value is also printed in the V-Ray Messages:

Make sure that the GPU Compute capability meets the required compute capability.

The RTX engine will work with NVidia GPU(s) without RT cores that meet the system
requirements. However, it is strongly recommended to use GPUs with RT cores to 
benefit from the performance boost.

Recommended NVIDIA Drivers

V-Ray GPU performs a graphics driver check when starting the render. The V-Ray Messages Log prints a warning message if the current NVIDIA driver is different from the recommended one.

The current recommended driver can also be found on the V-Ray GPU page on our website.

Using the recommended driver provides the best V-Ray GPU experience.