Contribute
Register

5K Resolution (5120x2880) via NVIDIA Drivers on Dell UltraSharp 27" Monitor

Status
Not open for further replies.
Crosspost from https://www.tonymacx86.com/threads/gpu-for-a-5k-monitor-dell-up2715k.204510/page-4#post-1452137:

I believe I've finally discovered the magical incantation for getting stable 5K support on the Hackintosh.
You MUST use the MacMini6,2 system definition. For some reason, this is the only sysdef that appears to allow hackintosh's to correctly support Multi-Stream Transport displays. I tried many combinations of editing framebuffers, AGDC settings, overrides, etc. None of it appears to matter. Just this sysdef. Wish I knew what it was doing as perhaps with the right edits we could use only a specific setting with any sysdef.

You need to use an ATI card. The FirePro W7000 works out of the box and has 4 displayports. I picked it because it is essentially the D300 on the Mac Pro 2013. I believe the W9000 will also work out of the box, tho I haven't tested it. But it is far more expensive.

Because of the weird ATI sleep/wake issues on Sierra, you will need to "sacrifice" an nVidia card to the Apple gods in your first PCI-E slot and set it to your primary boot display in the BIOS. The ATI card can work out of your secondary PCI-E and will stay active after boot completes. Sleep/wake should also work fine (obviously only as long as it was working fine before). In theory, doing this with the IGPU as primary should work too, but the computer and display attached to it will never truly go to sleep and you may have a phantom monitor as well if you do this. Maybe I can work around this with more tweaking.

I've only tested so far with the Dell UP2715K on my card ... i may test with the HP Z27q later as I do have one but the cables don't quite reach. We know monitors matter

The good news here ... no half screen cursor issues, no having to unplug the monitor and plug it back in when you boot, and hopefully no unstable sleep/wake issues with the nVidia drivers, but it will take time to verify the long term stability.
Well the thing is that 5k on ATI cards is another beast. I had an ATI card and got the 5K to work but I ultimately need CUDA performance so I've switched to Nvidia.

Are you saying that the MacMini6,2 would fix MST on Nvidia?
 
Well the thing is that 5k on ATI cards is another beast. I had an ATI card and got the 5K to work but I ultimately need CUDA performance so I've switched to Nvidia.

Are you saying that the MacMini6,2 would fix MST on Nvidia?
There's a possibility. I haven't tried it yet but it wouldn't be hard for me to test. I can't really explain how 5K works on certain Macbook's with the GeForce GT750M otherwise because as far as I know that driver hasn't been updated in ages. I think there's a special setting in these sysdef's that is causing the OS to natively handle the MST display. But the glitches with the cursor would still remain an issue.
 
Further details:

Thanks to some extensive googling I now realize that it's not really the Macmini6,2 definition that's special, but more the fact that the board id for that sysdef is not present in the AppleDeviceGraphicsPolicy kext (which is responsible for so much other grief with video cards). Apparently the mere presence of a board id in the Info.plist causes some default policy code to be run, even if the board id has "none" for the config. The kicker is that this default code apparently kills multi stream transport display support! So by removing the board id, MST displays on ATI cards start working!

Therefore, you can either use a sysdef like Macmini6,2 ... or use something like iMac14,2 and eradicate the matching board id from the AppleDeviceGraphicsPolicy Info.plist. I went with the latter for the moment.

The remainder of the issues left involve the standard ATI card problems that began with El Capitan ... sleep/wake being broken and FCPX crashing. You can either have FCPX working with a phantom display and a computer that doesn't sleep properly (this by getting a display running that has no kext support), or you can sacrifice a card and display with native driver support, but FCPX will try to use that card and crash when it starts up. Haven't found a good workaround here yet.

P.S. I did try my nVidia GTX 980 Ti and that did not work in 5K without nv_spanmodepolicy=1 set. Still had all the usual problems.
 
Last edited:
Hey guys,
thank you so much for all the help. The monitor works great and has been for a while. I wanted to ask you if you think this setup is possible. So I have the 980ti, 5K monitor, 4K monitor on HDMI and am thinking about getting a 4K monitor for the DP. Can the 980ti handle it? Thank you.
 
Oh wow, I totally missed that thread.

I'm on a Dell UP2715K with GTX970 for around 9 months now and beeing annoyed by the ugly cursor bug. I just read in this thread that driving the display in 4k works well. I should really try that.

After updating my BIOS (ASUS Hero VIII) yesterday I'm currently fighting to get 5k back at all. Things have changed and it looks like NVRAM support is broken. I had to use the Clover driver loading method to get Webdrivers working at all.

Maybe I should fetch an mDP cable and try the 4k mode.

Does anyone know of the kinds of problems i'm fighting now to get 5k back at all? May this have to do with the EmuVariableUefi Drivers? Any other known fixes to that?
 
I don't think there's any hope of 5K working properly on a hackintosh with Nvidia cards. I got it to work nicely with a 280x and thought it would be much better with a nvidia card. I was wrong.

I'll stick to Nvidia though, the performance difference in Premiere Pro is huge.

I'll probably just get a 4K monitor to replace this one.
 
Is there a thread that describes the cursor issue? I haven't seen it with my Nvidia Titan X and 5K output.
There's a thread at https://www.tonymacx86.com/threads/mouse-pointer-mouse-cursor-issues-in-sierra.204215/
Crosspost from https://www.tonymacx86.com/threads/gpu-for-a-5k-monitor-dell-up2715k.204510/page-4#post-1452137:
The good news here ... no half screen cursor issues, no having to unplug the monitor and plug it back in when you boot, and hopefully no unstable sleep/wake issues with the nVidia drivers, but it will take time to verify the long term stability.

Ok, I finally noticed the cursor problem while playing World of Warcraft. Each half of the screen has it's own hardware cursor. And the setting of the hardware cursor seems to be one setting behind the correct setting on one of the screen halves. When the cursor is moved to the horizontal center of the screen, you can see each half of the screen has a different cursor. This affects other apps, including Safari and Finder (the cursor is supposed to change when you mouse over the resize areas of a window). Actually, Safari and Finder stopped being affected when I quit World of Warcraft. Actually, I don't know what causes the problem, since when I restarted World of Warcraft, the cursor appeared to behave correctly...

A software cursor might be a possible solution, but a software cursor lags behind a hardware cursor.
 
Last edited:
Hey LostVector

I had the same realisation when trying to make a MST 4K display work, that Macmini6,2 was the only one that would work for it. I now have a HP Z27q 5K display as well, and trying to make it work with 2 DP cables. So far it doesn't, but the reason is that in my nvram despite setting it, nv_spanmodepolicy=1 does not get applied - kinda like how the nvda_drv=1 stopped working as well, and had to be injected differently.

What's wrong with nv_spanmodepolicy? Isn't that what I need to make this work? I have a GTX 980. How do I enable that when sierra ignores the boot args set in Clover config?

Thanks for any help you can be.

Further details:

Thanks to some extensive googling I now realize that it's not really the Macmini6,2 definition that's special, but more the fact that the board id for that sysdef is not present in the AppleDeviceGraphicsPolicy kext (which is responsible for so much other grief with video cards). Apparently the mere presence of a board id in the Info.plist causes some default policy code to be run, even if the board id has "none" for the config. The kicker is that this default code apparently kills multi stream transport display support! So by removing the board id, MST displays on ATI cards start working!

Therefore, you can either use a sysdef like Macmini6,2 ... or use something like iMac14,2 and eradicate the matching board id from the AppleDeviceGraphicsPolicy Info.plist. I went with the latter for the moment.

The remainder of the issues left involve the standard ATI card problems that began with El Capitan ... sleep/wake being broken and FCPX crashing. You can either have FCPX working with a phantom display and a computer that doesn't sleep properly (this by getting a display running that has no kext support), or you can sacrifice a card and display with native driver support, but FCPX will try to use that card and crash when it starts up. Haven't found a good workaround here yet.

P.S. I did try my nVidia GTX 980 Ti and that did not work in 5K without nv_spanmodepolicy=1 set. Still had all the usual problems.
 
it's probably not being ignored, it's just that nVidia did some super hacky stuff to make the Dell UP2715K work and the hack doesn't work with the HP. I have both monitors and the HP has never worked with my nVidia cards under any circumstances in 5K.
 
Status
Not open for further replies.
Back
Top