Does 10.13.4 handle simultaneous Nvidia + AMD cards any better? FCX Improved?

[rovingbreeze]

@jyavenard seems you also have the similar setup. Can you get both card working at the same time? AMD card in slot 1 for everyday use and nvidia card for cuda (web driver has to be installed).

Thanks in advance!

 

[Scott1204]

@jyavenard seems you also have the similar setup. Can you get both card working at the same time? AMD card in slot 1 for everyday use and nvidia card for cuda (web driver has to be installed).

Thanks in advance!

Don't have 10.13.5 yet but in .4, was not able to get both working simultaneously on the Hack.

The AMD worked solo as long as the Nvidia drivers were not installed. In my case, this was helpful because I could keep both cards in the machine and use the Nvidia for games under Windows 10 while the AMD was used under MacOS for Final Cut Pro.

As eGPU support matures, hopefully all cards will work under MacOS.

 

[rovingbreeze]

@Scott1204 Thanks Scott, I get it. Will probably delay my plan to buy an AMD card then...

 

[rovingbreeze]

that's sad. thanks.

 

[Drinahn]

At risk of resurrecting the ancient....

I recall reading that to do AMD in Mac, Nvidia in Windows, I would need to put the AMD card in the first PCI slot.
Is that still the case or has the situation changed?

I have a Radeon WX Pro 7100 (second hand) on the way and was expecting to have to shuffle my 1080 down a slot. This thread gives me hope it has got simpler?

 

[Drinahn]

If both a nvidia and amd are installed and you use 10.13, it boots to a black screen. I still semi-regularly boot into 10.13 and each time I do so, I must first uninstall my 1080Ti

Really? But others on this thread report that both can be installed - as long as the nvidia drivers are not installed the system should boot and AMD card work?

My WX Pro 7100 has arrived, so busy disk cloning...

 

[Drinahn]

So, was delayed a bit, as I realised after the card arrived that I had thrown out my PSU modular cables! So I had to order one in.

Radeon PRO WX 7100 installed in second PCI-E slot after I first removed Nvidia Drivers from macOS High Sierra 10.13.6. Plugged it into my display using second DisplayPort - so both cards are attached to the same 4K display using two DisplayPort cables.

Booted but appeared to hang with the Apple Logo and the progress bar incomplete. Enabled verbose in clover args and it stopped at:

"ConsoleUsers: gIOScreenLockState 3, hs 0, bs 0, now 0, sm 0x0"

This is I don't think much of a concern... out of interest, this was using the DisplayPort from the GTX 1080 in PCI-E 1. I used the display OSD to switch to the other input - and voila - I see my High Sierra background.. but no login screen. I have a go at logging in using keystrokes, and as far as I can tell - it works, my HDD whirs.... but the UI doesn't move on in any useful manner.. hmm.

Next step, I wanted to pass nv_disable=1 via Clover Boot Args. However for some reason I couldn't type _ in clover. There are a few threads about this, but none helpful. So I booted using macOS recovery partition and was able to mount my EFI partition using terminal, then vim the config.plist to add my nv_disable=1 to the boot args.

After this, it boots fine - leaving my main display at the same stage as above (ConsoleUsers: gIO.... etc) - HOWEVER when I switch inputs, I find the macOS login screen with the accounts displayed etc. Login - all good!


So my next steps I think will be to tell my bios to start initial display on the Radeon rather than the 1080 - I am hopeful that maybe that'll let it start up by default on the right input. {EDIT: Tried this, actually makes my machine not boot at all! } I guess I'll look at a DDST edit to disable the GTX for macOS... though I'm not clear if a DDST edit affects only macOS or whether it would also flow through to Windows?

If all else fails on the final niggle, I'll live with switching inputs when I boot to macOS. Catalina install is near in my future....

 

[jaymonkey]

So my next steps I think will be to tell my bios to start initial display on the Radeon rather than the 1080 - I am hopeful that maybe that'll let it start up by default on the right input.

@Drinahn,

Try disabling the Nvidia GPU using ACPI code as detailed in @Rehabmans guide which is linked in this post :-

RehabMan SSDT trick used to disable unsupported Nvidia and keep it in pcie slot while using Mojave

For people like me who want to keep their unsupported videocard in pcie slot 1 (or 2, or 3, I suppose) while testing Mojave with a second supported videocard in another pcie slot (or integrated igpu), take a look at @RehabMan Guide : [FIX] "Window Server Service only ran for 0 seconds" with...

www.tonymacx86.com

Many users with mixed GPU's have used that method to disable a Nvidia GPU when running Mojave or Catalina, if done correctly it should power down the Nvidia GPU which may allow your monitor to auto switch to the AMD.

Cheers
Jay

 

[Drinahn]

@Drinahn,

Try disabling the Nvidia GPU using ACPI code as detailed in @Rehabmans guide which is linked in this post :-

RehabMan SSDT trick used to disable unsupported Nvidia and keep it in pcie slot while using Mojave

For people like me who want to keep their unsupported videocard in pcie slot 1 (or 2, or 3, I suppose) while testing Mojave with a second supported videocard in another pcie slot (or integrated igpu), take a look at @RehabMan Guide : [FIX] "Window Server Service only ran for 0 seconds" with...

www.tonymacx86.com

Many users with mixed GPU's have used that method to disable a Nvidia GPU when running Mojave or Catalina, if done correctly it should power down the Nvidia GPU which may allow your monitor to auto switch to the AMD.

Cheers
Jay

Cheers Jaymonkey. Followed that but had a hard time identifying any _OFF methods in my DSDT. Realised of course it was because I’m on desktop. Anyway - rehabman does go on in the thread to point out that clover Devices/AddProperties and InjectNVidia to set IOName and class-code are usually enough on desktops.

Did that and now my login screen starts up on the Radeon display output reliably.

Bit the bullet and successfully upgraded from High Sierra to Catalina. No problems. Am going to research if I can used the feature described for rev5051 of Clover here https://www.insanelymac.com/forum/topic/304530-clover-change-explanations/?page=7 to make it start up and send no signal via Nvidia output.

 

[jaymonkey]

had a hard time identifying any _OFF methods in my DSDT. Realised of course it was because I’m on desktop.

@Drinahn,

The method does work for desktops as you can see if you read the first post in the link i gave you and the responses to that post in that thread. However it does very much depend on how well the GPU ACPI code is implemented in your native DSDT, some motherboards have the code some do not.

rehabman does go on in the thread to point out that clover Devices/AddProperties and InjectNVidia to set IOName and class-code are usually enough on desktops.

Did that and now my login screen starts up on the Radeon display output reliably.

Yes that is one way of doing it but i don't believe that it completely soft powers down the GPU as is the case when calling the _OFF ACPI method which is what would be called to power down the GPU when entering sleep mode.

Bit the bullet and successfully upgraded from High Sierra to Catalina. No problems. Am going to research if I can used the feature described for rev5051 of Clover to make it start up and send no signal via Nvidia output.

Let me know how you get on with that new feature, i read about it here a while back but have not had chance (or the need) to try it myself, if it really works as advertised it should simplify things for users with mixed GPU who need to disable Nvidia in MacOS considerably.

Cheers
Jay