Forgot your customer ID? Click here
The warning "num samples per thread reduced to 32768, rendering might be slower" is not an emergency, but it's a useful signal from your render engine. It tells you that memory constraints are forcing a more conservative work distribution. By understanding its cause—usually GPU VRAM limits or driver caps—you can take targeted actions: reduce tile size, lower samples, upgrade hardware, or simply accept the slight slowdown.
When Blender reduces the num_samples_per_thread , it changes how data is packed and sent to your processor.
Because the GPU computes fewer samples simultaneously, it must communicate with the host system far more frequently. This introduces latency, causing your GPU usage to fluctuate and your render times to stretch significantly. Render with vray memory error - Extensions - SketchUp Forum
| Warning / Error | Meaning | |----------------|---------| | Out of GPU memory, falling back to CPU | Severe VRAM shortage. | | Render tile too large, splitting | Same root cause as our warning. | | Kernel failed to launch: invalid resource size | Driver rejecting per‑thread buffer size. | | CUDA error: launch timeout | Different issue, but often related to large work loads. |
It starts like a tiny whisper from the engine: a single line of text in a console, a fleeting warning on startup. You glance at it, half-curious, half-annoyed. “num samples per thread reduced to 32768 — rendering might be slower.” Technical, terse, and strangely human: an engine telling you it’s doing its best with less to work with.
If your scene has high max_ray_depth (e.g., 32 bounces), each sample generates many ray states. Multiplying samples per thread by ray depth can quickly overflow memory.
You cannot always eliminate the warning entirely, but you can reduce its performance impact or adjust settings to avoid triggering it.
To avoid encountering this warning in the future, follow these best practices:
A huge number of lights, complex shaders, or ray-traced lighting increases the need for higher samples per thread OctaneRender Forums.
To prevent a total crash or an "Out of Memory" error, V-Ray automatically scales back the amount of work (samples) it assigns to each thread to fit the scene into your remaining VRAM. While the scene will likely still render, it will be significantly slower because the hardware is not operating at full efficiency. How to resolve it
If your code or the library requests a larger batch size (e.g., for better vectorization), the driver silently caps it.
std::cout << "Active sample count: " << engine.getActiveSamples() << std::endl;