Blender 2.78a on archlinux (FX-8320 @ 4.4GHz, GTX1060)

Rendering times for RyzenGraphic_27.blend from amd website:

CPU - 0:58
CUDA - 1:20

Live rendering times when switch "Viewport Shading" to "Rendered":

CPU - 0:22.28
CUDA - 0:04.54

I expected one to be faster than other in both cases, but this is strange, why is this happening?

  • $\begingroup$ Can I ask how you enable GPU rendering via Python? My Blender never uses GPU on arch linux. $\endgroup$ Dec 29, 2016 at 23:58
  • $\begingroup$ i just change settings in blender to cuda, then in rendering tab set render to gpu, nothing else, it works $\endgroup$
    – avi9526
    Dec 30, 2016 at 0:47
  • $\begingroup$ Right, with GUI it's that simple. I am asking if you know how to do it via Python. Seems not. Thanks anyways! $\endgroup$ Dec 30, 2016 at 0:55
  • $\begingroup$ Live rendering is preformed on the whole screen area at once, or a large part of it at least; and actual rendering is performed in smaller tiles, the way Cycles works that leads to very different management usage and bandwidth requirements, which gives different performance results. Also viewportRendered view generally uses a much lower number of samples suitable for a quick preview, while final rendering requires "production quality" sample count $\endgroup$ Dec 30, 2016 at 3:08
  • $\begingroup$ i tried set samples for rendering same as for preview (50), still same $\endgroup$
    – avi9526
    Dec 30, 2016 at 3:29

1 Answer 1


If you are rendering with a CPU smaller tile values like 16x16 ( keep them square ) will be fastest. With a GPU larger ones like 256x256 work better. This is due to the way they make calculations. So just ticking the box and set it to GPU without changing this value will make it slower than it can be. Since viewport render is the whole screen the GPU will be faster because the "tile" is bigger enter image description here


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .