Monitoring V-Ray GPU performance and memory usage on Windows

Overview

V-Ray GPU fully uses your available GPU compute resources during rendering, but Windows Task Manager may not display this accurately. Use a dedicated monitoring tool to verify real GPU usage.

Why Windows Task Manager may show low or 0% GPU usage

By default, V-Ray GPU takes full advantage of the hardware's capabilities and efficiently utilizes all of the available processing power during rendering. 

Windows Task Manager’s Performance graphs can misreport compute workloads and may default to showing 3D engine activity instead of compute usage. 

As a result, it can appear that additional GPUs are idle (for example, showing 0% on GPU 2 and GPU 3) even while V-Ray GPU is actively rendering on them: 

Monitor GPU usage accurately with GPU-Z or MSI Afterburner

We recommend using an external tool like GPU-Z or MSI AfterBurner for monitoring GPU resources especially GPU usage.

1. Install a monitoring tool.
2. Start a V-Ray GPU render in your host application.
3. Open the monitoring tool and view per-GPU sensors/graphs (e.g., GPU Load, Memory Controller, and Power) to see real compute utilization.
4. If needed, adjust the tool’s settings (sampling rate, sensors displayed, and on-screen display) so you can track each GPU individually.
5. Use the monitoring tool’s readings as the source of truth for GPU usage; do not rely on Windows Task Manager for compute workloads.

Understanding GPU metrics

Windows Task Manager

• GPU performance graph: Shows overall GPU usage. This is not accurate for reporting compute (CUDA) performance.
• GPU memory usage graph: Shows total GPU memory usage.
• Dedicated GPU memory: Reports the amount of physical GPU memory (VRAM) in use.
• Shared GPU memory: Refers to system memory the GPU can borrow if needed; it is slower than dedicated GPU memory. It is helpful when using out-of-core rendering (“Use System Memory for Textures”).
• GPU temperature: Shows the temperature of the GPU chip. It is also visible under each device name on the left side of the Task Manager window.
   

Stats tab in the VFB

Use the Stats tab in the V-Ray Frame Buffer (VFB) to monitor GPU engine, memory, and performance while rendering Interactively (IPR) or in Production mode in V-Ray 5 or later.

Note: Some stats like Load % is only available when the Sampler is set to Progressive.

Engine: Displays the current GPU engine (CUDA or RTX).

Available memory: The VRAM V-Ray GPU can use on this device. This is lower than the card’s total VRAM because:

• Windows WDDM reserves part of the GPU memory.
• DCC (e.g. 3ds Max, SketchUp, etc.) take a chunk of the GPU memory for the GPU connected to monitors.
• The GPU driving monitors and your DCC (for example, 3ds Max) reserves additional memory.
• Other open applications (e.g. Photoshop, After Effects, Nuke) also take a chunk of the available GPU memory for a GPU device.

Used memory: The amount of VRAM V-Ray GPU is using for the current scene. The RTX engine typically uses more VRAM than the CUDA engine.

Load %: Current GPU utilization; available only when the Sampler is set to Progressive.

Performance: Useful metrics related to progressive rendering performance.

Timing: Render time after scene compilation.

Memory tracking: Detailed breakdown of GPU memory usage by scene elements (textures, geometry, etc.). Note: when using "Use System Memory for Textures" Checkbox the stats tab might show incorrect data for GPU memory. This will be improved in the future.