Tell you how to Setup CUDA for MacOS High Sierra

Support Technical Support
1

I have been searching for enabling CUDA rendering by Cycles for several days, finally I found the solution.

  1. you need nVidia latest driver for your specific MacOS version
  2. you need latest CUDA driver for your MacOS

In fact nVidia released new driver for every version for the MacOS, but I don’t know why you can’t search it on its website.

you just can search on Google, for keyword like :

nvidia quadro & geforce macos driver release 10.13.6

The 10.13.6 is your specific MacOS version, it must match yours
Then click link to go to download page

32%20PM
32%20PM2702×1796 886 KB

Then you go to CUDA site to download latest CUDA driver for your mac.

Then go!

(I’m a new user, can only post one image, else I have uploaded the driver version and blender render option screen shot)

2

Are you using an eGPU with Blender? Which one, and what are your experiences?

3

A little warning / note: Don’t update the OS too far or you’ll get install failures because the OS compatible driver is too new for the hardware and the hardware compatible driver is too old for the OS. :troll: :watermelon::smile::gun:

Edit: the snub got turned into a raygun, for real?

4

Yes

It rendered really faster using CUDA acceleration

5

On my mac book pro, with GeForce 750M, and my PC with a Geforce 1080

I’m using latest experimental blender 2.79 build for using both CPU + GPU.

Using CUDA, it is really fast to render than CPU only.

However, on my PC, the rendering process only take up only about 30% of GPU performance, and rendering frames one by one. it is not fully utilizing GPU horse power.

1 Like
6

You are right.

To this point, I think we should better using a linux based system, which is far more stable than using Windows or MacOS, especially Windows 10 will force you update to latest version

7

Even if it did would it prevent the installation of the old hardware compatible driver like OSX / MacOS does? Still, maybe there is some Hackintosh method to make them install?

I agree on using Linux, more for being the most flexible to enable CUDA and OpenCL features on Nvidia GPUs.