Contribute
Register

Intel Hades Canyon NUC?

Status
Not open for further replies.
Yes… another chance for a working hackintosh if this doesn't work, would be Intel releasing an updated NUC in 2019 with some "Coffee Lake-G" (or a "G" version of the continuously postponed Cannon Lake) with say, a Vega 12 GPU instead of an RX Vega M. But nothing in the news about this (whereas a lot is known about Intel's roadmap). Either Intel has abandoned this path, or they are very secretive about it!
 
The current Hades Canyon can be considered very "custom" since there is absolutely no other desktop PC with an RX Vega M, and very few laptops. Plus this is the first and unique Intel processor merging their own CPU with an AMD GPU in one single package, so this is VERY custom. Yet it already exists, so I don't really understand your point. Especially as it seems the new Vega 12 GPUs will be more widespread used in the industry, than the restricted RX Vega M. Therefore, a hypothetical Coffee Lake-G with Vega 12 would be LESS "custom" than the current Kaby Lake-G with RX Vega M.

There is a market for this, since Intel UHD graphic chips are awful.
 
Understood. I agree that the current one looks somewhat "custom". But the "fault" is not on Intel's side. It is the business of Apple or AMD to create drivers for macOS, not Intel. Apple would never do it for the hackintosh community obviously, so the only chance would have been the Mac mini using this processor, which many were expecting, but didn't happen sadly. AMD does not make independent "web drivers" like NVIDIA for macOS, so this too is a dead-end :(
 
Great. But I am a bit pessimistic. The announced Vega 12 GPU is the chip inside the Radeon Pro Vega 16 and 20 indeed, available in the new high-end MacBook Pro 15". It has a Vega architecture from start to end (GCN 5.0).

The Radeon RX Vega M in the Hades Canyon is marketed as being a "Vega" GPU, but it is not. It is recognized internally as a Polaris 22 as you know. But it is not a pure Polaris chip neither. It is somewhat a hybrid with a Polaris architecture (GCN 4.0) and a Vega memory controller (to keep things simple).

Still, it is a very good GPU* but Polaris 22Vega 12

So I think it won't work alas. But I'd be pleased to be wrong (especially as I'd like to buy this NUC for macOS!)
Keep up the good work and let us know.

* I notice that the low-end Polaris 22 GPU (Radeon RX Vega M GL) has the same number of shading and texture mapping units as the high-end Vega 12 GPU (Radeon Pro Vega 20).
Glad to know!
It very important information so we should try fakeid 0x67DF1002 and inject ATI use amd9500controller.kext

This gets us closer, but something is still not right.

First, using Clover for this inject does not work (I will need to do some investigation as to why that is the case). Clover just hangs. Maybe NoDefaultProperties would have helped, didn't test it.

But if you inject just device-id=0x67df using config.plist/Devices/Properties/PciRoot(0)/Pci(1,0)/Pci(0,0)/device-id=<df670000>, you will be able to boot with acceleration. WhateverGreen.kext (+Lilu.kext, of course), were also required. More investigation required to discover what WhateverGreen is actually doing in this scenario, and whether there are any tweaks there to help finish it.

The desktop looks ok,... transparent menu bar and dock make you think it is actually working.

But you have glitches. For example, Safari is a mess. Alt-Tab window is scrambled, and even simple menu popups (such as right click on a dock item) are scrambled. You can use Firefox (didn't test Chrome).

I will come back later tonight with screen shots, config.plist examples, a codeless injector kext, etc.
 
Great. But I am a bit pessimistic. The announced Vega 12 GPU is the chip inside the Radeon Pro Vega 16 and 20 indeed, available in the new high-end MacBook Pro 15". It has a Vega architecture from start to end (GCN 5.0).

The Radeon RX Vega M in the Hades Canyon is marketed as being a "Vega" GPU, but it is not. It is recognized internally as a Polaris 22 as you know. But it is not a pure Polaris chip neither. It is somewhat a hybrid with a Polaris architecture (GCN 4.0) and a Vega memory controller (to keep things simple).

Still
This gets us closer, but something is still not right.

First, using Clover for this inject does not work (I will need to do some investigation as to why that is the case). Clover just hangs. Maybe NoDefaultProperties would have helped, didn't test it.

But if you inject just device-id=0x67df using config.plist/Devices/Properties/PciRoot(0)/Pci(1,0)/Pci(0,0)/device-id=<df670000>, you will be able to boot with acceleration. WhateverGreen.kext (+Lilu.kext, of course), were also required. More investigation required to discover what WhateverGreen is actually doing in this scenario, and whether there are any tweaks there to help finish it.

The desktop looks ok,... transparent menu bar and dock make you think it is actually working.

But you have glitches. For example, Safari is a mess. Alt-Tab window is scrambled, and even simple menu popups (such as right click on a dock item) are scrambled. You can use Firefox (didn't test Chrome).

I will come back later tonight with screen shots, config.plist examples, a codeless injector kext, etc.

The Polaris 22 might be a problem if it’s really a Lexa architecture based card. RX 550 is the Lexa based card which is Polaris 12/22? GFX804 like Polaris 10/20/30, Polaris 11/21 etc..

So getting the Lexa RX 550 working may be the key to this Vega/Polaris/Lexa card issue.

25973373-BD54-48C9-BF15-90F1E3EDA71F.png


Hopefully, if Apple is using the Lexa drivers for the new Vega 12 cards it can be worked out. But if Bega 12 is truly a Vega architecture based card then the Lexa based chips probably won’t be feasible on these NUCs.
 
Last edited:
It is known that
  • Polaris 22 (e.g. Radeon RX Vega M GL or GH) is GCN 4.0 / gfx804
  • Vega 12 (e.g. Radeon Pro Vega 16 or 20) is GCN 5.0 / gfx904
Two different generations and architectures. Still, RehabMan's discovery looks promising :)
 
Can't believe it. It works <3. Thank you Rehabman. You always save the day. It is in a perfectly usable state for my workflow.
Here's my config.plist if someone else is also looking for it.


Yg1pcVS.png
 

Attachments

  • config.plist
    8.9 KB · Views: 553
Can't believe it. It works <3. Thank you Rehabman. You always save the day. It is in a perfectly usable state for my workflow.
Here's my config.plist if someone else is also looking for it.


Yg1pcVS.png

Very Nice, can you run an openGl bench like Uningine Valley?

Are you using fakepciid kexts and injection?
 
Status
Not open for further replies.
Back
Top