I am also running 387.10.10.15.15.108 without issue on two machines. I am keeping an archive of several recent driver releases on hand, just in case.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Frequently Asked Questions About NVIDIA PC (non-EFI) Graphics Cards
- Thread starter Asgorath
- WikiPost WikiPost
- Start date
- Sort by reaction score
- Status
- Status
I am also running 387.10.10.15.15.108 without issue on two machines. I am keeping an archive of several recent driver releases on hand, just in case.
Cool, can you come back in 2 years and remind us of this? TIA.
Just found a informative article from a very reliable news portal regarding mojave webdrivers.
The site is in german language, but you may translate it via google:
https://www.heise.de/mac-and-i/meld...stecken-angeblich-bei-Apple-fest-4217891.html
They are stating that nvidias driver kexts are stuck at apple where they need to be signed to pass the SIP.
The site is in german language, but you may translate it via google:
https://www.heise.de/mac-and-i/meld...stecken-angeblich-bei-Apple-fest-4217891.html
They are stating that nvidias driver kexts are stuck at apple where they need to be signed to pass the SIP.
Apple and NVIDIA really need to work it out. I get that Apple wants to promote their overpriced eGPU now. But c'mon...
It's technically not even theirs, it's Blackmagic Design... I'm sure Apple is getting a hefty cut.
Nvidia needs to work it out. They don’t support the full Metal feature set or 10 bit output and they keep their driver support for new cards in beta stage with no tech support. That means the drivers cause headache for Apple because users complain to them about bugs instead of complaining to Nvidia.
[doublepost=1541845346][/doublepost]
Apple only get the profit from selling it but it is a costly niche device and the money is a pittance for Apple. Probably the least money earner on the store.
Anyone knows if these webdrivers can work correctly with two GPU?
I know, now it's a known bug. Grrr
PS: I own two GTX 1070, a Founder Edition and a GTX 1070 Katana single slot.
Thank you for sharing your experience with this driver.Did You try new driver "40.108"?
I'm testing nvidia multi GPU combinations since Yosemite , and on high sierra please look at this post if x.x.x40.108 does not work for You let me know.
Post #189 in his thread :
https://forums.macrumors.com/thread...graphics-cards.1440150/page-189#post-26620848
to install multiGPU Nvidia cards even with different Chips This is only driver combination (drv+cuda)which works with multi Nvidia GPU configurations @HS.
As far as I know In Sierra the last update destroys the functionality as well.
The driver before worked with multi GPU configuration, going back with
SIP- disabled and using older driver is production proofed for me.
( if the research is too complicated for Sierra for you I can find out the build numbers if you write me a pm.
For HighSierra use WebDriver in linked post)
[doublepost=1541951951][/doublepost]
Flame 2019,1 supports HighSierra
Flame 2019,2 ->ADSK supports now even Mojave 4GB VRAM required
Maya 2019 is still Beta
Last edited:
Thanks Speedstar, I'm going to test the two solutions. 
Edit:
With the last webdriver, same graphic problems.
I'll test your second solution, disable SIP and install the old driver. I hope it will work with 10.13.6 17G3025.
Edit2:
I can't install 378.10.10.10.25.106 with High Sierra 10.13.6 (17G3025) (CUDA 387.128 yes).
I've rebooted with SIP disabled, but the installer is blocked with the last version of High Sierra.
Edit:
With the last webdriver, same graphic problems.
I'll test your second solution, disable SIP and install the old driver. I hope it will work with 10.13.6 17G3025.
Edit2:
I can't install 378.10.10.10.25.106 with High Sierra 10.13.6 (17G3025) (CUDA 387.128 yes).
I've rebooted with SIP disabled, but the installer is blocked with the last version of High Sierra.
Last edited:
You need to use the script from the webpage which modifies
Certification and Build-numbers,
It is not enough to try install older driver.
Did You use the script under
Command section?
https://vulgo.github.io/nvidia-drivers/
Rollover command:
(for example
bash <(curl -s https://vulgo.github.io/webdriver) 378.10.10.10.25.106
)
You need to copy and paste it
In terminal and it will download the driver and modify it for current build.after that install modified drv... Good luck !!!
( disabling SIP in this case , is not for the driver setup app.but for the command in Terminal which turns the nvidia webdriver into developer driver!
after the modification you get developer driver to be installed and loaded during boot if SIP is disabled )
Bon chance mon frère, merci pour like button
Last edited:
Ah, ok, I'll test this solution.
Thanks a lot Speedstar, for your precious help.
Edit 1: Ok, it's installed, both old webdriver and Cuda, I can login, I see the desktop 1 second then it quits and asks me the login and password ...
Edit 2: I can now login with one GTX 1070 in slot 4.
Edit 3: When I re-connect the GTX 1070 Founder Edition in Slot 1, with the Katana in slot 4, locked on login. :/
Edit 4: When I disconnect the HDMI cable from the GTX 1070 Katana, no login problem.
The two cards are recognized and finally work, with no lags, Open CL and Metal work on the two GTX, calibrated monitors profiles!
I'll re test with a different cable, with DVI for example.
Edit 5: I've installed a third Nvidia GPU, a GTX 1050 single slot from Inno3D. It works but without the HDMI cable (no cables connected), with HDMI same login problem. Need to test with DVI to see if it's the driver's fault.
Edit 6: it's not a problem of connector. Tested with HDMI and DVI, of two different GPU.
With the webdriver 378.10.10.10.25.106 High Sierra 10.13.6 (17G3025) and CUDA 387.128 , connect your monitors to your primary GPU, and use your other(s) GPU for computing, without monitors.
Thanks a lot Speedstar, for your precious help.
Edit 1: Ok, it's installed, both old webdriver and Cuda, I can login, I see the desktop 1 second then it quits and asks me the login and password ...
Edit 2: I can now login with one GTX 1070 in slot 4.
Edit 3: When I re-connect the GTX 1070 Founder Edition in Slot 1, with the Katana in slot 4, locked on login. :/
Edit 4: When I disconnect the HDMI cable from the GTX 1070 Katana, no login problem.
The two cards are recognized and finally work, with no lags, Open CL and Metal work on the two GTX, calibrated monitors profiles!
I'll re test with a different cable, with DVI for example.
Edit 5: I've installed a third Nvidia GPU, a GTX 1050 single slot from Inno3D. It works but without the HDMI cable (no cables connected), with HDMI same login problem. Need to test with DVI to see if it's the driver's fault.
Edit 6: it's not a problem of connector. Tested with HDMI and DVI, of two different GPU.
With the webdriver 378.10.10.10.25.106 High Sierra 10.13.6 (17G3025) and CUDA 387.128 , connect your monitors to your primary GPU, and use your other(s) GPU for computing, without monitors.
Last edited:
Now It's time for you to experiment with HDMI
https://forums.macrumors.com/threads/guide-hdmi-audio.1499797/
Maybe the behavior will change with log-in problems after this patch applied it worked on my GTX Mac edition with nv Quadro
I have booted on first (PCIe 1)GPU connected and then attached second (PCIe 3) screen on other GPU or vice versa
But never boot with all connected screens on different GPUs after driver install.
Once your configuration is established it may become very buggy if you try to experiment with different screens sometimes it breaks all the configuration .
In my case after attaching six different screens on two GPUs The configuration become unusable and breaked with one of the screens disconnected and changed resolutions, so be careful with synchronization of different screens !
This synchronization features changed during the OS updates
(<13.4) so it's not The problem of the driver.
This driver was introduced before eGPUs operating system feature leaved BETA
In this scenario SwichResX becomes very handy.
I have login problems with all other drivers, and did not understand your results.
1Question 2x1070 but different brand do not log in with both connected to hdmi ?
2Question 3 different 1070 fe, katana, inno3d 2GB vram work only with one card connected per dvi ?
And finally make a screenshot of your McP Cuda POWERHORSE with benchmark like Vray that would be interesting!!!
https://forums.macrumors.com/threads/guide-hdmi-audio.1499797/
Maybe the behavior will change with log-in problems after this patch applied it worked on my GTX Mac edition with nv Quadro
I have booted on first (PCIe 1)GPU connected and then attached second (PCIe 3) screen on other GPU or vice versa
But never boot with all connected screens on different GPUs after driver install.
Once your configuration is established it may become very buggy if you try to experiment with different screens sometimes it breaks all the configuration .
In my case after attaching six different screens on two GPUs The configuration become unusable and breaked with one of the screens disconnected and changed resolutions, so be careful with synchronization of different screens !
This synchronization features changed during the OS updates
(<13.4) so it's not The problem of the driver.
This driver was introduced before eGPUs operating system feature leaved BETA
In this scenario SwichResX becomes very handy.
I have login problems with all other drivers, and did not understand your results.
1Question 2x1070 but different brand do not log in with both connected to hdmi ?
2Question 3 different 1070 fe, katana, inno3d 2GB vram work only with one card connected per dvi ?
And finally make a screenshot of your McP Cuda POWERHORSE with benchmark like Vray that would be interesting!!!
Last edited:
For the two questions.
1st GPU Asus GTX 1070 (1 monitor connected in DVI)
2nd KFA2 GTX 1070 (1 monitor connected in HDMI, even with the DVI, forced log out, very frustrating. )
When I don't connect a monitor on the second GPU, no problems.
I'll test with the HDMI patch.
PS: yes I've made some benchmarks with one, two and three GPU, (Octane Bench).
I'll post the screenshots.
1st GPU Asus GTX 1070 (1 monitor connected in DVI)
2nd KFA2 GTX 1070 (1 monitor connected in HDMI, even with the DVI, forced log out, very frustrating.
)When I don't connect a monitor on the second GPU, no problems.
I'll test with the HDMI patch.
PS: yes I've made some benchmarks with one, two and three GPU, (Octane Bench).
I'll post the screenshots.
For the two questions.
1st GPU Asus GTX 1070 (1 monitor connected in DVI)
2nd KFA2 GTX 1070 (1 monitor connected in HDMI, even with the DVI, forced log out, very frustrating.
When I don't connect a monitor on the second GPU, no problems.
In my case the first is connected Via hdmi and second DP-DVI and VGA
Disastrous like their webdrivers?
No, because of these issues (german and english): https://www.tomshw.de/2018/11/14/de...lle-sein-nvidia-rtx-graphics-cards-are-dying/
FAQ updated with my recommendation to not buy an NVIDIA GPU until all these issues get sorted out, and we see timely web driver releases that enable all NVIDIA architectures like Volta and Turing.
- Status