[Success] AMD RX6000 Series working in macOS

[CaseySJ]

I just installed macOS Ventura on OC 0.8.5 and now I'm back to not having my Radeon RX 6900 XT being recognised, getting inverted/laggy OS + wrong card labeling in system info. I've dug through this thread for answers but no luck so far. I also have ID 73af and IGPU on my Designare Z390 is set to Enabled in BIOS F9i, which it was when it all worked in Monterey.

As far as I can remember I fixed it last time with this:


But it's not doing the trick anymore. Any help would be greatly appreciated!

I just looked at the config.plist you posted, and there's no SSDT-Z490-GPU-SPOOF.aml:

Were you using this SSDT with Monterey?

 

[Ponkass]

I just looked at the config.plist you posted, and there's no SSDT-Z490-GPU-SPOOF.aml:
View attachment 557287
Were you using this SSDT with Monterey?

Hah, wth. I'm sure I tried with it enabled on my initial tries and then it didn't work. Now it suddenly did! Well, thanks for that! I seem to be up and running again!

 

[shasen1235]

Just upgraded my Vega64 to 6800XT and here are some experiences I would like to share.

*SMBIOS and Final Cut rendering
My CPU is 10900K so naturally iMac20,2 SMBIOS should be the best choice. But after config my iGPU to headless, the system works but FCP stutters here and there and rendering time is bad, super bad, like 3 times slower than my M1 Pro MacBook. Something must be wrong.

*igfxrpsc=1 boot flag
Turns out, due to macOS's strange design, the iGPU inside a Hackintosh would work just at 0.35Ghz most of the time, adding this flag solves the issue. Some said igfxfw=2 works too, you can try either one just remember don't use them at the same time. With iGPU working at full speed, the rendering time cut down significantly yet it still takes 50% more time to render the same footage compared to M1 Pro. Way to go.

*Final optimize
By checking the task manager, usage on 6800XT is still very low at 10~15%. I've heard some people saying that macOS treat non Pro Radeon as external GPU in iMac SMBIOS. This issue can be solved by switching to MacPro or iMacPro and finally cut the render time to 35% faster than M1 Pro. But again, this means I will lose iGPU decoding for daily video viewing and create unnecessary load for CPU. After some research, injecting framebuffer in OpenCore device section solves the issue. For my MSI 6800XT Gaming Trio, ATY,Carswell works the best. GPU load when rendering is at 40~60% full time matches the speed as MacPro SMBIOS and I don't lose any feature provided by iGPU. Finally I can unleash full power if this beast without giving up any feature

**Side note
If all optimizations are applied and rendering is still slow, maybe it's the issue inside GPU hardware path.
Adding SSDT-BRG0.aml injection on top of them should do the job.

 

[MysticalOS]

I removed custom injection when 12.x fixed the bug with massive metal performance loss. but readig that makes me wonder if I should re-add it

 

[elgipsy]

Just upgraded my Vega64 to 6800XT and here are some experiences I would like to share.

*SMBIOS and Final Cut rendering
My CPU is 10900K so naturally iMac20,2 SMBIOS should be the best choice. But after config my iGPU to headless, the system works but FCP stutters here and there and rendering time is bad, super bad, like 3 times slower than my M1 Pro MacBook. Something must be wrong.

*igfxrpsc=1 boot flag
Turns out, due to macOS's strange design, the iGPU inside a Hackintosh would work just at 0.35Ghz most of the time, adding this flag solves the issue. Some said igfxfw=2 works too, you can try either one just remember don't use them at the same time. With iGPU working at full speed, the rendering time cut down significantly yet it still takes 50% more time to render the same footage compared to M1 Pro. Way to go.

*Final optimize
By checking the task manager, usage on 6800XT is still very low at 10~15%. I've heard some people saying that macOS treat non Pro Radeon as external GPU in iMac SMBIOS. This issue can be solved by switching to MacPro or iMacPro and finally cut the render time to 35% faster than M1 Pro. But again, this means I will lose iGPU decoding for daily video viewing and create unnecessary load for CPU. After some research, injecting framebuffer in OpenCore device section solves the issue. For my MSI 6800XT Gaming Trio, ATY,Carswell works the best. GPU load when rendering is at 40~60% full time matches the speed as MacPro SMBIOS and I don't lose any feature provided by iGPU. Finally I can unleash full power if this beast without giving up any feature

Are you still using igfxrpsc=1 boot flag after making your final optimization?

 

[shasen1235]

Are you still using igfxrpsc=1 boot flag after making your final optimization?

Yes, it seems iGPU still co-work with dGPU in FCP.
So far I haven't seen any negative effects so I still keep it.

 

[shasen1235]

I removed custom injection when 12.x fixed the bug with massive metal performance loss. but readig that makes me wonder if I should re-add it

Personally I think Geekbench score is not really that important.
After the optimization the score does not change significantly, but real world work difference is night and day.

 

[roraj19]

Just upgraded my Vega64 to 6800XT and here are some experiences I would like to share.

*SMBIOS and Final Cut rendering
My CPU is 10900K so naturally iMac20,2 SMBIOS should be the best choice. But after config my iGPU to headless, the system works but FCP stutters here and there and rendering time is bad, super bad, like 3 times slower than my M1 Pro MacBook. Something must be wrong.

*igfxrpsc=1 boot flag
Turns out, due to macOS's strange design, the iGPU inside a Hackintosh would work just at 0.35Ghz most of the time, adding this flag solves the issue. Some said igfxfw=2 works too, you can try either one just remember don't use them at the same time. With iGPU working at full speed, the rendering time cut down significantly yet it still takes 50% more time to render the same footage compared to M1 Pro. Way to go.

*Final optimize
By checking the task manager, usage on 6800XT is still very low at 10~15%. I've heard some people saying that macOS treat non Pro Radeon as external GPU in iMac SMBIOS. This issue can be solved by switching to MacPro or iMacPro and finally cut the render time to 35% faster than M1 Pro. But again, this means I will lose iGPU decoding for daily video viewing and create unnecessary load for CPU. After some research, injecting framebuffer in OpenCore device section solves the issue. For my MSI 6800XT Gaming Trio, ATY,Carswell works the best. GPU load when rendering is at 40~60% full time matches the speed as MacPro SMBIOS and I don't lose any feature provided by iGPU. Finally I can unleash full power if this beast without giving up any feature

**Side note
If all optimizations are applied and rendering is still slow, maybe it's the issue inside GPU hardware path.
Adding SSDT-BRG0.aml injection on top of them should do the job.

Would you be able to share your EFI or config.plist? I have a similar build (17-10700, 6800XT) but I'm having some performance issues. Specifically interested in the injecting frame buffer part.

 

[RurouniX]

Hello everyone, I just got the RX 6650 XT, and I am trying to patch it to work.
But unfortunately, I can find just 2 apci values on windows or on macOS.

The latest driver is installed.

I will appreciate for your help.

 

[etorix]

You have two unnamed bridges ('pci-bridge' in IOReg; 'PCI(0000)' in Windows). These need to have ACPI names for OpenCore to apply DeviceProperties to the GPU downstream.

Use SSDT-BRG.aml and then apply DeviceProperties:
PciRoot(0x0)/Pci(0x3,0x1)/Pci(0x0,0x0)/Pci(0x0,0x0)/Pci(0x0,0x0)
key device-id data <FF730000>

(Only device-id is required, further entries are purely cosmetic.)

Alternatively, a single SSDT-RX6650XT.aml may name the bridges and apply the spoof (no device property entries).

For the sake of clarity: Use either
a) SSDT-BRG and DeviceProperties entry or
b) SSDT-RX6650XT alone
but not both SSDTs together.

SSDTs and PCI path in DeviceProperties are specific to your board and to the slot used. Other users must adapt paths to their hardware.