How to build your own iMac Pro [Successful Build/Extended Guide]

[shael]

@DSM2 can you please share how youve implemented your vega cards? is the ssdt section for the GFX0 blank?

 

[kgp]

@DSM2 can you please share how youve implemented your vega cards? is the ssdt section for the GFX0 blank?

You have even a fully described 2x Vega 64 SSDT implementation of @apfelnico in Section E.9.2) of my guide !

 

[shael]

You have even a fully described 2x Vega 64 SSDT implementation of @apfelnico in Section E.9.2) of my guide !

i understand and its already implemented. however i believe the frame buffer may be set wrong for the card, as i dont believe that is how apple natively implemented them

@shael : The best way is to use @kgp 's guide... Also for compatibility issues...

i thought i saw you mention that you currently are not using any ssdt for the card to function..

 

[TECH_ATTACK]

@shael Yes, but the method I use is not for "normal-users"!
The implementation by Apple is correct, the problem is a different one:

1) The Cooling solution is different
2) iMac Pro can handle up to 6!!! screens (6 outputs, check iMac Dump for this) depending on the screen size (resolution) and thats why all Vegas running with full fan speed.
Even if you would tell the controller that you only have 4 outputs it will not change anything.

The only solution to get the cards quite is to water-cool it.

Hmm, not exactly. I have e Vega Frontier Air Cooled and a Vega Frontier Water Cooled. The Problem is not the temperature, its just a strange usage even in IDLE. Play a video on youtube in safari, and the fan is getting very quiet. (I always start a video, pause it and minimize. Fan is very quiet then. This strange usage did come with the update to 10.13.2. In 10.13.0 this "problem" was not present. The drivers for Vega in OS X (and RX cards) are still buggy. Just test it with Blender, many scenes are rendered wrong (no hair etc). I think Apple is updating their drivers for Polaris/Vega soon.

 

[kgp]

There should be no frame buffer set for the Vega frontier edition. There should also be no ssdt.
You will see high fan speeds and a lot of GPU activity at idle, but this is a flaw in the driver itself, not the hackintoshing. We'll probably have to live with it until Apple configure the driver for egpu use rather than the imac pro use.
I have the EK waterblock so don't suffer the fan thing, and the GPU activity doesn't appear to affect the performance.

As far I understand the Vega SSDT implementation is of purely cosmetic nature (PCI device implementation in Apple's System Report) and has no influence on the Vega performance... Thus if you implement the VEGA in the SSDT or not is a question of taste and nothing else..

I know that you neglect to use a SSDT, but that's your personal decision and likely also your personal problem.

All the best,

KGP

 

[kgp]

I use Lilu + Whatevergreen for cosmetic reasons - it gives me the Vega description in "about this mac".
Sorry, I meant to say no SSDT for the Vega FE. On account of it's native to the Mac.
My system is flawless

Nobody recommends to use Lilu.kext + Whatevergreen.kext with a natively implemented Vega 64 or Vega Frontier. I don't know where you take this nonsense from! Carefully read my guide instead of stating and distributing b*****t

Many thanks for the additional EDIT of your post above and the added flowers We always have a weird way of communication but deep inside I love you and your comments and recommendations..
Thus absolutely nothing for bad, ok?

Hugs,

KGP

 

[shael]

@shael Yes, but the method I use is not for "normal-users"!
The implementation by Apple is correct, the problem is a different one:

1) The Cooling solution is different
2) iMac Pro can handle up to 6!!! screens (6 outputs, check iMac Dump for this) depending on the screen size (resolution) and thats why all Vegas running with full fan speed.
Even if you would tell the controller that you only have 4 outputs it will not change anything.

The only solution to get the cards quite is to water-cool it.

isnt the first half of the code kgp has in the guide (prior to method _DSM) non cosmetic?

further more with my current setup i have 1 port that doesnt work and when i update or reach an installer with multiple monitors i reach black screens

 

[kgp]

isnt the first half of the code kgp has in the guide (prior to method _DSM) non cosmetic?

further more with my current setup i have 1 port that doesnt work and when i update or reach an installer with multiple monitors i reach black screens

Well my nvidia GPU implementation is purely cosmetic, despite that I use it for the HDAU PCI implementation..

I did not study @apfelnicos Vega 64 SSDT implementation in all it's detail. Thus there might be really some part which is none-cosmetic. If you think so, delete that part from the VEGA PCI device implementation and that's it.

We agree??? Sometimes the solution is more easy than you think and has not to fill the next twenty pages..

@shael, you are an expert for opening the box of Pandora... Try to fix some of your issues or doubts by yourself and don't ask always other people to do the work for you.

What do you think about this suggestion?

 

[shael]

Well my nvidia GPU implementation is purely cosmetic, despite that I use it for the HDAU PCI implementation..

I did not study @apfelnicos Vega 64 SSDT implementation in all it's detail. Thus there might be really some part which is none-cosmetic. If you think so, delete that part from the VEGA PCI device implementation and that's it.

We agree???

i tried that, and even while compiling correctly and checking the code multiple times. it breaks all pci implementation.

im self righteous.... its easier for me to get someone to do it for me lol

 

[kgp]

i tried that, and even while compiling correctly and checking the code multiple times. it breaks all pci implementation.

Because your modifications are wrong! If one device is badly implemented, the PCI implementation is broken for all devices. In your case your mistakes in modifying the Vega 64 implementation collapse the entire PCI device implementation ...

Thus, try to properly modify the part of the Vega 64 implementation and everything will just work flawless. Once you have the in your opinion correct and working Vega 64 PCI device implementation approach, distribute your awesome work to others.

No further posts and comments to this topic until then please...

Good luck, man!