V-Ray GPU Minimal System Requirements.
- NVIDIA GPUs of Maxwell generation or later when using CUDA or RTX modes.
- All GPUs recognized by the system are usable, although only 4 are officially supported.
- All compute cores are utilized in CUDA or RTX mode, with RT cores only being utilized when in RTX mode. While only RTX-class cards can use RTX mode, these cards can also fully use CUDA mode (and in doing so, utilize the system’s CPU(s) as well).
- Intel 64, AMD64, or compatible processor with SSE4.22 support.
- All sockets and cores are utilized when running CUDA x86 mode (with or without GPUs). No GPU is required in the system when running solely in CUDA x86 mode.
- 8 GB of system memory minimum. The actual amount depends on a scene's complexity.
- The current Recommended Windows Driver is 546.01 Studio driver for GeForce cards and 535 U2(537.70) for Quadro & Tesla cards.
V-Ray GPU Recommended System Requirements
- V-Ray GPU performance scales nearly linearly across CUDA cores and core clock speed for a given GPU generation when running in either CUDA or RTX modes.
- The scaling between GPU generations Pascal to Ada Lovelace is shown in this article.
- V-Ray GPU performance scales nearly linearly across multiple GPUs as shown in this article.
- While GPUs in a multi-GPU system can vary in their type, speed, and generation, the highest efficiency is usually found when combining GPUs of the same generation and of similar performance.
- The entire scene being rendered (geometry, textures, buffers, etc.) must fully fit into GPU memory when using either CUDA or RTX mode, see our guide on managing GPU memory.
- When running in just CUDA x86 mode, the system’s RAM is used and paging is supported. If your scene exceeds your GPU’s memory, you can often still render the scene using CUDA x86 mode.
- V-Ray GPU can pool memory across pairs of GPUs that support NVLink and have a physical NVlink bridge installed. For example, two cards having 24GB each can support scenes requiring 48GB when using NVLink. NVLink support can be found on the larger GPUs previous to the Ada generation. More about NVLink setup here.
- When using multiple GPUs with different memory for rendering, the scene should fit into the GPU with the least amount of memory.
- Intel 64, AMD64 or compatible processor with SSE4.22 support, PCIe 4.0(or newer) capable CPU and motherboard are recommended. Apple M1 & M2 processors are supported when running CUDA x86 on MacOS.
- CUDA x86 mode performance scales nearly linearly across CPU cores and clock speed, including M1 & M2 processors from Apple.
- When using both CUDA and CUDAx86 modes (aka hybrid GPU+CPU or XPU) the contribution of the CPU is similar to that of adding another (typically smaller) GPU to a multi-GPU configuration. CPUs with very high core counts (e.g., an AMD Threadripper) can often rival the speed of a GPU as shown in this graph.
- For multi-GPU machines, it is recommended to have 6 physical cores per GPU. For example, a machine with 4 GPUs should have at least 24 physical cores to reach the optimal performance.
- 32 GB of system memory or more, the system memory should be equal or greater than 2x the GPU memory.
- Microsoft Windows 10(version 2004 or newer), Windows 11 64-bit.
- Linux 64-bit distributions with glibc 2.17 or later, V-Ray GPU performance is slightly better on Linux than on Windows.
- For CUDA x86 mode, MacOS 11 Big Sur(or newer) is also supported.
- V-Ray GPU is found in all Chaos V-Ray integrations as well as V-Ray Standalone and in Chaos Cloud Rendering.
- To use V-Ray GPU in V-Ray Standalone, use -rtEngine=5 for CUDA mode or -rtEngine=7 for RTX mode.
- To use V-Ray GPU in Chaos Cloud, make sure the renderer is set to V-Ray GPU in your host DCC.