euqlaog

Screen output works but no acceleration yet.

Script’s options are documented on GitHub. I haven’t tested 10.11.5, but you can try, sudo ./automate-eGPU.sh -uninstall sudo ./automate-eGPU.sh -url http://any-valid-web-address-here.pkg

The script finds only officially announced drivers and the latest at the moment is 10.11.4 in Nvidia’s dropdown list. Humans can use google and search those leaked beta links and apply the -url option.

@SignatureInferno As the script says, no web driver yet available. It downloads only official drivers and doesn’t support beta builds.

@Proph No difference when you do it. Using an external ATX PSU is also worth a try.

Thanks for the info. SMC reset and erase & reinstall OS X also tried? Please describe your power configuration since III-D provides only one 6-pin connector? Your III-D is likely causing this problem and requires the right timing in TB connection.

Yes, the only stable R9 thus far with the AKiTiO. Faster than a GTX 980 which is in the same price category.

@Proph He has exactly the same problem and the same enclosure: https://github.com/goalque/automate-eGPU/issues/24 AKiTiO users have not reported this issue. Did you try another TB port?

Through a self-made 8-pin-to-barrel adapter from ST45SF-G. It didn’t help with the R9 390 but this energy efficient Nano stayed stable. Mostly under 200W, max 206W peaks in 3DMark13. So small and light that it sits upside down on my desk I wish I had a Dell DA-2. Somehow I got an automatic Crimson software update within Windows. http://www.3dmark.com/3dm/11689114 http://www.3dmark.com/3dm11/11181494

Go and grab a R9 Nano (Asus). It’s a perfect match with the AKiTiO on Windows 10. No sign of crash at full load or idle. I have completely removed the 4-pin cable and soldered a wire on the back side and cannot revert back to original factory circuit but I guess it should work just as well.

@Proph If the nvram command says <00> just before the next “hang” reboot, I don’t have a clue why it works only every other time. @juniordiscart The release is delayed because I’m not sure about the licensing. It will be a free demo app and in binary form. GNU-EFI doesn’t contain the necessary protocol so I had to include a header file from the EDK2. They are both under BSD. The current MIT license allows commercial purposes, but that may not be the case with this EFI app.

@Xaularis The only solution is to change the card, https://www.techinferno.com/index.php?/forums/topic/7989-script-automating-the-installation-of-egpu-on-os-x-inc-display-output/&do=findComment&comment=131501

@Xaularis and @Jojo2805 Please discuss hardware related issues preferably here: https://www.techinferno.com/index.php?/forums/topic/6224-diy-egpu-macbook-experiences/ @Xaularis From your post history I can see that you have chosen a GTX 750Ti without the 6-pin connector. If you do a search, you will realize that you are not alone with this issue. AKiTiO is not stable with cards without auxiliary power connector.

Completely correct. The daemon process is launched only in special cases. This is the problem. If normal reinstall didn’t work, you should do NVRAM reset, SMC reset, erase your SSD and reinstall from the OS X Internet Recovery and upgrade to OS X El Capitan. Script v0.9.8 and MacBookPro11,1 is verified to work on OS X 10.11.4. No boot issues.

@Proph Since the daemon plist is created and the service is running, this issue became more difficult. Try these steps, without eGPU monitor: 1) sudo ./automate-eGPU.sh -uninstall and restart the Mac 2) Comment lines 1076, 1077, 1078 and 1082 (# character in front of the line) 3) Change the line 62 of the script to: config_board_ids=() 4) sudo ./automate-eGPU.sh (eGPU plugged) 5) Restart the Mac (eGPU plugged) 5.1) If the boot process hangs with default mode [-m] - Restart without eGPU, hot-plug the eGPU, sudo ./automate-eGPU.sh -a - Type “nvram tbt-options” and tell what it says. If not exactly “<00>”, we found the problem. Exit and report, otherwise go to 5. 5.2) If the boot process hangs with automatic mode [-a] - Restart without eGPU - Type “nvram tbt-options” and tell what it says. If not exactly “<00>”, we found the problem. Exit and report, otherwise go to 5. 5.3) If the step 5 was successful, go to 5 and repeat a couple of times. 6) If went through, the problem is fixed (without eGPU monitor). Try also another TB port.

@Alchemistake Set the external monitor as the primary in System Preferences/Displays.

@MacGTX The error message under Hardware/PCI is normal. Did you try Erase & Reinstall OS X? What does “sometimes” mean? @Proph Ok. So the first restart works with eGPU [-a] mode and never with [-m] mode? What's the output of (4) after reboot (try first eGPU unplugged)? And do you have automate-eGPU-dameon.plist in /Library/LaunchDaemons/ ?

@Proph The problem is your multi-slot enclosure and script’s [-a] mode is especially designed to resolve boot issues with them but doesn’t take into account new iMacs. Would you open Terminal, type “sudo ./automate-eGPU.sh -a” without eGPU, and provide the output of these commands: (1) sw_vers -productVersion (2) sw_vers -buildVersion (3) ioreg -c IOPlatformExpertDevice -d 2 | grep board-id | sed "s/.*<\"\(.*\)\">.*/\1/" (4) sudo su root -c 'launchctl list | grep automate-egpu-daemon' If (4) says nothing, it’s very likely that automate-eGPU-dameon.plist doesn’t exist in the folder /Library/LaunchDaemons/ It’s also good news because I know how to fix it. The script takes backup of kexts into /Library/Application Support/Automate-eGPU/backup so please don’t do anything manually. Since you already did, boot into recovery (Command+R) and choose “Reinstall OS X”.

I assume you meant to say Late 2013 or Mid 2014 rMBP. Both have the same model identifier. The script requires untouched S/L/E folder and it’s automated to download only official drivers that match with the OS X build. If you have modified kexts manually, your only option is to reinstall OS X or restore from a Time Machine backup.

@fisch3r R7 260X is not supported by Apple and AKiTiO is not stable with newer AMDs. Better to stick to cards whose device ids are ready in the kexts. I also remember that XFX’s VBIOS is not Mac compatible. I personally prefer Sapphire and Asus. Thanks! The old good days when web browsing was done in Lynx, emails composed in Pine and everyone had a Red Hat Unix tools are still powerful, the most straightforward way to get the job done.

@shirazigs Don’t edit kexts manually. If you do so, the script backups them and cannot recover original kexts. 1) Reinstall OS X 2) If the SIP is re-enabled, boot into recovery mode and type csrutil disable 3) Run the script and take into account the following, https://github.com/goalque/automate-eGPU/issues/17

Ok, the script validates if SIP is turned off. OS X update might re-enable it. I can verify that 10.11.4 works

SIP is System Integrity Protection. All the updates should be done by rerunning the script. It will download the correct driver and makes backup folder. Uninstalling is not necessary, but as you already clicked yes for Nvidia’s prompt window, it’s recommended to uninstall everything: 1) sudo ./automate-eGPU.sh -uninstall (click yes this time) 2) restart 3) sudo ./automate-eGPU.sh (eGPU plugged) 4) restart

@Kr4nG Don’t update via Nvidia’s driver manager. Disable the SIP and run the script again.

@Kane There is no universal solution. It is a game programmer’s responsibility to choose the acceleration API, an active GPU for OpenGL/Metal view, and adapt to renderer changes. OS X is not a good gaming platform. Nvidia has to keep up with the Apple’s changes and it results in bugs. The future is Metal, but the API needs refinement, http://us.battle.net/wow/en/forum/topic/20043464897?page=1