This question already has an answer here:

CUDA Error: Out of Memory in cuArrayCreate(&handle, &desc)

I get this error when rendering with Cycles on my GPU. I'm not entirely sure what's wrong with this... My scene DOES have some very high-res textures (up to 10K) and most of the textures also have a map for Gloss, Reflection, Normals, and displacement. My specs are as follows:

CPU: Intel i5-4460

GPU: Geforce GTX 1060 6GB VRAM



marked as duplicate by Duarte Farrajota Ramos, iKlsR Jan 13 '17 at 3:06

This question has been asked before and already has an answer. If those answers do not fully address your question, please ask a new question.

  • 1
    $\begingroup$ Cuda is out of memory, means your scene is too complex for the available memory $\endgroup$ – Duarte Farrajota Ramos Jan 6 '17 at 16:39
  • $\begingroup$ @DuarteFarrajotaRamos would reducing the resolution of most of the textures used append this? Or is it because of the geometry? I've handled much more complex geometry in scenes before... $\endgroup$ – Josh Silveous Jan 6 '17 at 17:11
  • 2
    $\begingroup$ Basically your scene exceeds the resources of your GPU. Texture size should not be larger than that of the size they will be in the final render. For example if you have an object in the distance that has a texture of 4k but will only be 100 pixels on the screen, you are just wasting resources. Only the textures that will be seeing in great detail and occupying a large portion of the screen in the final render should require large size textures. 3D rendering is all about optimizing resources to get acceptable results, use every trick that you can think of to avoid overtaxing your system. $\endgroup$ – cegaton Jan 6 '17 at 17:32
  • 1
    $\begingroup$ Think also that if your GPU is the only graphics card on your system, the resources are shared with the monitor(s). So not all of the vRAM is available for rendering $\endgroup$ – cegaton Jan 6 '17 at 17:36
  • $\begingroup$ @cegaton I was not aware that when GPU rendering your textures are flushed to VRAM, Not RAM. And I was wasting resources, like you said. Could you put your response in answer format so I can accept it? $\endgroup$ – Josh Silveous Jan 6 '17 at 18:07

Browse other questions tagged or ask your own question.