6
$\begingroup$

I have a GTX 1060 6GB in my current build, and I also have a GTX 960 2GB in storage.

I was wondering if I got one of these PCI Express X1 to X16 Extension Adapter, and got both of my GPUs in one build (not SLI or CrossFire, just in the same system) would I be able to use both GPUs for rendering simultaneously?

I don't care if this means having them both render separate tiles in one instance of Blender, or rendering separate frames in two instances. I'm using Cycles for everything I do.

$\endgroup$
11
$\begingroup$

You can, as long they are from the same brand (AMD, or NVidia) you should be able to use them simultaneously to render in Blender Cycles.

If you keep them out of any proprietary GPU pairing technologies setup like SLI or Crossfire, the operating system detects them as two discrete GPUs and see both graphics cards, Blender should be able to see them as discreet GPUs in the user preferences.

This will virtually decrease render times linearly, proportionally to each additional compute device performance, as opposed to pairing technologies which have significant performance penalties and lose efficiency with each additional GPU unit.

You can then prior to rendering select which one to use, or both simultaneously, if you want. You can also have each instance of Blender use its own GPU independently, although you will likely experience performance penalties caused by bottlenecks anywhere else in your system, namely available memory, and CPU resources for pre-processing your scene.

Under User Preferences > System > Cycles Render Device you can activate CUDA processing and pick from the list of available devices below to chose which ones to use for each running session. From the Properties Window > Render Properties you can then set the device to GPU while Cycles is set as active render engine.

Have in mind that when using multiple GPUs for rendering a single scene you will always be bound by the available memory on the card with the least VRAM amount. Since GPUs still can't share memory access, a copy if the whole scene must be fully loaded into each individual graphics card memory, so it must completely fit into the available RAM of every GPU, failing any of them will abort rendering.

| improve this answer | |
$\endgroup$
  • 4
    $\begingroup$ Worth noting: available vRAM for rendering will be limited to the size of the smallest card. So if one of the cards has 6Gb and the other one 2, the 1060 will only use 2 for rendering. $\endgroup$ – user1853 Nov 30 '16 at 4:01
  • $\begingroup$ Yes, that's right, forgot to mention that. That is true for rendering the same scene with bots cards only, I believe. If you render two separate files at the same time each card will be able to fully utilize it's total amount of memory individually, as far as I know. $\endgroup$ – Duarte Farrajota Ramos Nov 30 '16 at 5:34
  • $\begingroup$ @cegaton I don't mean to bump an old question, but does that mean I could have 3 Instances of blender Rendering at once? On on my 1060, one on the 960, then one on CPU? $\endgroup$ – Josh Silveous Dec 7 '16 at 1:24
  • 2
    $\begingroup$ @JoshSilveous You can have many instances at once. Be aware that at some point you'll get diminishing returns when you hit the limit of what your computer can handle efficiently. Do keep in mind that even when you render on GPU, the CPU is not idle: there are still a lot of processing going on with it. If what you are after is efficiency look into more gpus of the same capacity or a render farm service (much more affordable than getting your own hardware) $\endgroup$ – user1853 Dec 7 '16 at 2:39
  • $\begingroup$ @cegaton I was not aware that GPU rendering used CPU, so then Just rendering seperate instances with two GPUs would be better? $\endgroup$ – Josh Silveous Dec 7 '16 at 2:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.