Mac Pro 2013 eGPU set ups
I'd love to hear how you succeeded with the El Capitan setup.
and your process for a single eGPU on the nMP
If I remember correctly, I downloaded the script via the instructions from Goalque.
chmod +x ~/Desktop/automate-eGPU.sh
I had the GPU plugged in while I was doing this.
I guess couple questions. What card are you using to connect to the Mac and where are you getting hung up?
I can run the script ok. When I restart, the progress on the Apple boot screen gets to around 45-50% then freezes.
The system reboots with the kernel panic error message and the cycle starts again.
It then loops like this until I force shutdown (Holding down the power button).
I tried -a mode and it booted once, but the card was not recognised by name only by NVIDIA Chip Model.
When I restarted it went back to the freeze loop.
I would try and SMB reset. shut down the Mac, unplug the power for 15-20 secs. This resets the thunderbolt ports if there is something electronically hung up. Then start he Mac without the Akitio plugged in. Plug the Akitio in, run the script again. Then use -a mode.
sudo ./automate-eGPU.sh -a
restart. If that doesn't work, I would try sudo ./automate-eGPU.sh -uninstall and restart and then repeat the above process.
Is dual Titan Z setup still workable if you reinstall Mavericks from the Internet recovery and then upgrade to Yosemite or El Capitan? Firmware updates possibly have an effect on multiple eGPU detection as well.
The early boot process has been changed significantly in macOS Sierra. Browsing the system log and comparing it to earlier OS X versions is one way to troubleshoot this problem. You can also use verbose mode (Command-V) on startup.
@imcalledandy: did you try GTX 980Ti?
I've had to put it off until Monday unfortunately. Work got in the way!
I will let you know how I get on. I'm trying to establish a rock solid process:
I think starting with the SMC reset is a good move. Then maybe reset PRAM. Disable SIP and follow the steps to install and run the script. Then, before restart turn on -a mode then restart and keep fingers crossed.
Anything to add to that?
Sounds a good plan. It’s also worth to try without -a mode in Bus 0. Good luck with the GTX 980Ti!
Interesting post at MacRumors:
@goalque - The 980Ti worked, quite literally out of the box 🙂
As you suggested I have tried without -a mode and on bus 0.
I have a Thunderbolt display which I need to have unplugged during boot or the system tends to hang. I wonder if using -a mode and switching to a different bus would solve this? Edit - This issue has resolved itself with a restart!
If anyone is interested in my process, let me know and I'll respond.
The only issue I have so far is that Final Cut Pro 10.3.2 does not open - it crashes upon launch.
FCP X now works!
I'm running macOS 10.12.2
Is Titan-X on Pascal architecture? Meaning if Titan-X works on MacOS any GTX 10XX will work? Am I missing something?
@devild - I'm using the card without a display connected, so mainly for CUDA rendering in C4D. I think I've just got lucky to be honest.
But this is what I did...
First, I opened FCP X and it crashed. I sent the report to Apple and restarted the machine.
I then opened loads of apps; Photoshop, Illustrator, After Effects, C4D and FCP and it opened straight away.
I closed it and opened again - all fine.
I've just tested opening it as I type and it worked - I didn't have to open other apps etc (I doubt that had anything to do with it anyway).
I'm sorry this isn't a great answer.
Mac Pro spec - 3.3GHz 8-core (I changed the processor), 64GB RAM, AMD D700s, EVGA 980Ti
eGPU is connected to TB Bus 0 (lower left) and is not powering a display
Also, I am not using @goalque's script in -a mode i.e. I ran it as default when setting up the eGPU.
Just out of curiousity with FCPX, did you make a new timeline? I.e. I can get FCPX to open but as soon as a create a new timeline or load a project it crashes consistently
also did the crash have anything to do with " CGLSetVirtualScreen failed - didn't expect to get here." I have tried both -a mode and standard mode and FCPX is crashing as soon as I open a timeline or a clip. It is in Bus 0. Tried it with 980Ti and Titan Z. No Luck. Every other App works fine. Just FCPX. Any thoughts?
Hi @nickmilitello, sorry for the delay in coming back to you. I can confirm that I can use FCPX normally (creating new projects etc) with the eGPU connected.
Have you tried re-installing FCPX?
How is your dual eGPU setup working generally? Do you daisy chain the devices?
Also, do you use C4D?
Hey, thanks for getting back to me. Tried reinstalling OS and FCPX. Even if I start empty, i.e. No project open, then create a timeline, I get the crash every single time along with the same error message. Tried single GPU, one Titan Z and then tried a smile 980ti. Still crashes.
wasnt ever able to get daisy chaining to work with dual GPUs. Yep using c4d. Definitely odd that the exact same set up works fine on my MacBook Pro. Definitely odd.
Tried single GPU, one Titan Z and then tried a smile 980ti. Still crashes.
wasnt ever able to get daisy chaining to work with dual GPUs.
Perhaps it's the brand of graphics card. EVGA cards seem to be the most compatible.
So you haven't been able to daisy chain dual eGPUs? This is what I was hoping to try next.
What brand and exact card are you working with? It would be awesome to get FCPX to work. The dual GPU hasn't worked, i.e. able to boot since I upgraded to sierra. On El Capitan I used both ports on bus 0 to get it to work. The daisy chaining was never success on either operating system.
I have the EVGA GTX 980 Ti FTW ACX 2.0+ 6GB
In El Capitan, I'm assuming once one eGPU worked it was simply a case of plugging in the other and it was detected?
Yeah I believe I tried the a single 980ti ftw evga with no luck in sierras and FCPX working. Will take another look. However one note, multiple maxwell cards i.e. 900 series don't work. From my experience I was only able to get titan and 700 series to work in multiple configuration. Also I did have to modify the script a bit and have the lines regarding "install GPU one at a time removed". Goalque helped me with that. Then it was working fine with two Titan z, each plugged into the two ports on Bus 0.
Thanks for the reply. Please can you elaborate on the dual setup you had when it was working, just so I'm clear...
i.e. Will I have to run the script again even if I'm using the exact same card and setup, albeit plugging into a different port on bus 0?
Also, I'm on 10.12.2 - I wonder if that makes a difference to FCPX and eGPUs?