HP EliteDesk 800 G4 / G5 Mini with RX560x dGPU (hackintosh)

[deeveedee]

I am experimenting with some minor changes to my posted EFI based on things I learned while hacking my HP Elitebook 850 G7 (which is an awesome, inexpensive hack, by the way).  The changes with which I'm experimenting are:

• Remove boot-arg darkwake=2 (which may impact sleep)
• Add DeviceProperty "pci-aspm-default = <02000000>" to each NVMe m.2 SSD device for NVMe power management as explained by @MaLd0n here.

In my macOS Tahoe testing, I find that NVMe ASPM is disabled without the addition of the pci-aspm-default DeviceProperties.  A sample plist is attached with my changes for those who may want to experiment.

 

After adding the pci-aspm-default DeviceProperties, Hackintool shows NVMe ASPM enabled in Tahoe:

Spoiler

 

Sleep/wake was working perfectly for me before and continues to work perfectly after these changes.

config.plist.zip

[deeveedee]

14 hours ago, ird said:

 

So, inspired by you I went a bit wild and enabled default ASPM for both PCIe bridges 0x0 and 0x1 root nodes (and by association to all attached end points). I'm going to test a bit if this causes instability in the next 2 days before I'm away for the rest of this year. Basic testing, I found no differences in CPU, GPU geekbench 6 scores. My dGPU scores were a bit low to start with before this change (I never got the opportunity to debug it). I'll report back if I see any issues enabling L1 ASPM.

 

  Reveal hidden contents

 

 

 

 

 

 

 

 

Edit: Sorry, I didn't realize this is the non-dGPU thread! I'll leave the dGPU Geekbench 6 scores as-is in the post anyway. The rest of PCIe devices are common between the two configs.

 

@ird Replying here in response to your post in the other thread (no worries).  I am also seeing reduced  GeekBench 6 dGPU benchmarks after upgrading to macOS Tahoe 26.2.  My metal benchmark is consistently ~19K (down from ~24K) after repeated reboots.  pci-aspm-default setting for the dGPU doesn't seem to affect the benchmark as you have observed.  After examining the pci-aspm-default properties for real Macs, I am currently adding the property for NVMe SSDs, HDEF (HDAS), Ethernet, HDAU, dGPU.

Spoiler

 

EDIT: @ird I installed macOS 26.1 in an additional APFS volume so that I can A/B performance testing between 26.1 and 26.2 (booting both with the same EFI).  In repeated tests, macOS 26.1 consistently produces GB6 Metal Benchmarks = 23K-24K while macOS 26.2 consistently produces GB6 Metal Benchmarks = ~19K.

Edited December 14, 2025 by deeveedee

[deeveedee]

New version of IntelMausiEthernet.kext is available from Mieze.  Look for the latest release starting here.

 

EDIT: I am currently testing macOS 26.2 with IntelMausiEthernet.kext 3.0.3.  I no longer need to set Kernel > Quirks > DisableIoMapper = True for fully working Ethernet.  AppleVTD is enabled and Tahoe Ethernet works.

 

Edited December 15, 2025 by deeveedee

[deeveedee]

Very easy update to Tahoe 26.3 Beta (25D5087f).  Prior to update, I disabled BluetoolFixup.kext (for incremental update) and added boot-arg -wegnoegpu (required because I'm still using Acidanthera's WhateverGreen.kext 1.7.0).  After update, I re-enabled BluetoolFixup.kext and deleted -wegnoegpu boot-arg.

 

After rebooting a few times, GeekBench6 Metal benchmark is still around 19K vs. the 24K observed with macOS Tahoe 26.1.

 

 

EDIT: If someone is interested and has time, it might help us to compare the IORegistries for macOS 26.1 and 26.2 / 3 to see if there are device property differences for the dGPU.

Edited December 17, 2025 by deeveedee

[ird]

On 12/14/2025 at 5:59 AM, deeveedee said:

 

@ird Replying here in response to your post in the other thread (no worries).  I am also seeing reduced  GeekBench 6 dGPU benchmarks after upgrading to macOS Tahoe 26.2.  My metal benchmark is consistently ~19K (down from ~24K) after repeated reboots.  pci-aspm-default setting for the dGPU doesn't seem to affect the benchmark as you have observed.  After examining the pci-aspm-default properties for real Macs, I am currently adding the property for NVMe SSDs, HDEF (HDAS), Ethernet, HDAU, dGPU.

  Reveal hidden contents

 

EDIT: @ird I installed macOS 26.1 in an additional APFS volume so that I can A/B performance testing between 26.1 and 26.2 (booting both with the same EFI).  In repeated tests, macOS 26.1 consistently produces GB6 Metal Benchmarks = 23K-24K while macOS 26.2 consistently produces GB6 Metal Benchmarks = ~19K.

 

Interesting, I'm still on Sequoia 15.7.3 and I'm seeing lower dGPU GB6 scores, this has been the case for quite some time now. I wonder if it has to do with iMac19,2 vs MacMini8,1 SMBIOS differences.

[deeveedee]

@ird I'm traveling and away from my HackMini8,1, so I don't have my current DeviceProperties to post.  I recall that you were focused on power saving while I was focused on performance.  Somewhere earlier in this thread, I had posted the dGPU performance DeviceProperties that I was testing.  Maybe those DeviceProperties are making a difference?  Also, you have my MacMini8,1 config.plist, so you can test to see if SMBIOS is affecting performance.  I am very happy with SMBIOS MacMini8,1 (which gives me access to all three video ports).

[ird]

3 hours ago, deeveedee said:

@ird I'm traveling and away from my HackMini8,1, so I don't have my current DeviceProperties to post.  I recall that you were focused on power saving while I was focused on performance.  Somewhere earlier in this thread, I had posted the dGPU performance DeviceProperties that I was testing.  Maybe those DeviceProperties are making a difference?  Also, you have my MacMini8,1 config.plist, so you can test to see if SMBIOS is affecting performance.  I am very happy with SMBIOS MacMini8,1 (which gives me access to all three video ports).

 

Great point (and memory!), @deeveedee. Yes, I think I have modified PowerPlay tables that may well be the reason for the lower GB6 scores.

[deeveedee]

A fix to Lilu.kext is coming which will address the Radeon dGPU install issues that we have been experiencing in Tahoe.  With this Lilu fix, there are no changes needed for WhateverGreen.kext.  After this Lilu change is released and the new Lilu.kext is included in our EFIs, there will be no need for boot-arg -wegnoegpu when updating Tahoe.

 

Thank you @Avery B !!!

Edited January 14 by deeveedee

[deeveedee]

For those who are maintaining their Intel-based hacks in order to delay the inevitable migration to Apple Silicon, the wait is going to be well worth it.  And even if you don't jump immediately to an M5 because of the price, the release of Apple's newest silicon is going to make a purchase of the older M3 or M4 (they will be the older models when M5 is released) an extremely compelling value.  We're fortunate that these hacks are working so well to carry us into the next generation of Macs.

 

 

[deeveedee]

The Lilu fix that resolves problems when upgrading Tahoe with our dGPU is now available in a Lilu artifact and will be officially available in Lilu.kext 1.7.2.

[deeveedee]

The new Lilu.kext 1.7.2 works perfectly.  I upgraded my EFI with Lilu.kext 1.7.2 Beta prior to upgrading from Tahoe 26.3 Beta 2 -> Beta 3.  The upgrade proceded without issues and without needing boot-arg -wegnoegpu.  macOS upgrades are back to "normal" with the new Lilu.kext.  Well done Avery B!  Note that I still needed to temporarily disable BluetoolFixup.kext to allow the incremental update.

 

 

EDIT: Note that after the upgrade from Tahoe 26.3 Beta 2 -> Beta3, there are graphics artifacts around some cursors.  This is a known issue with Tahoe 26.3 Beta 3 and affects systems with dGPUs and/or iGPUs.  

Edited January 29 by deeveedee

[deeveedee]

Tahoe 26.3 Beta 3 has fixed the reduced Polaris GeekBench 6 Metal performance that I was seeing with previous versions of Tahoe 26.3.  Metal performance as measured by GeekBench 6 is back to ~24K.

Spoiler

 

[deeveedee]

Tahoe 26.3 (25D122) fixes the cursor artifacts while preserving the excellent GB6 Metal benchmark performance.  Flawless update.  More details here.

[deeveedee]

Updated to Tahoe 26.3 (25D125) Release.  No issues.  GB6 Metal benchmarks continue to be excellent.  This has become a perfect little hack for running macOS Tahoe.  I'm still using VoodooHDA.kext 3.0.3 for audio and no Wi-Fi, so this hack needs no root patches (SIP csr-active-config = <01000000> for VoodooHDA.kext).

 

About This Hack

Spoiler

 

GeekBench 6 Metal

Spoiler

 

[ird]

On 2/12/2026 at 6:36 PM, deeveedee said:

Updated to Tahoe 26.3 (25D125) Release.  No issues.  GB6 Metal benchmarks continue to be excellent.  This has become a perfect little hack for running macOS Tahoe.  I'm still using VoodooHDA.kext 3.0.3 for audio and no Wi-Fi, so this hack needs no root patches (SIP csr-active-config = <01000000> for VoodooHDA.kext).

 

About This Hack

  Reveal hidden contents

 

GeekBench 6 Metal

  Reveal hidden contents

 

 

This machine is equally good on previous version (I guess that's true by association :)). Just upgraded to BIOS 02.33.00 Rev.A and Sequoia 15.7.4 and everything works just great. After a long time, I've stopped tinkering with my computers and starting to "just use it"

[deeveedee]

I've been using this HackMini running Tahoe 26.4 Beta (25E5218f) all day today as my main work platform.  It has operated perfectly ... Quiet, fast, flawless.  Tahoe is a fitting conclusion to this awesome little hack.

[deeveedee]

I am traveling, so I haven't had a chance to test VoodooHDA.kext 3.1.0 on this HackMini, but I have tested it on my HackBookPro laptop and it works well as I noted here, here and here.  VoodooHDA.kext 3.1.0 does not have the AppleGFXHDA problem that required loading AppleALC.kext as a AppleGFXHDA blocker.  I am currently using VoodooHDA.kext 3.1.0 on my HackBookPro with AppleALC's MaxKernel set to 24.99.99 (Sequoia).

 

EDIT: I am testing VoodooHDA.kext 3.1.0 on my HackBookPro with Tahoe 26.4 Beta (25E5223i).

 

EDIT2: I was wrong about VoodooHDA.kext 3.1.0 and its need for a AppleGFXHDA blocker.  On this HP EliteDesk 800 G4/G5 Mini, I still need to load AppleALC.kext with VoodooHDA.kext 3.1.0, so that AppleALC blocks AppleGFXHDA.  Otherwise, AppleGFXHDA binds to HDAS (HDEF) preventing VoodooHDA.kext 3.1.0 from loading.

Edited March 16 by deeveedee

[deeveedee]

I was wrong about VoodooHDA.kext 3.1.0 and its need for a AppleGFXHDA blocker.  On this HP EliteDesk 800 G4/G5 Mini, I still need to load AppleALC.kext with VoodooHDA.kext 3.1.0, so that AppleALC blocks AppleGFXHDA.  Otherwise, AppleGFXHDA binds to HDAS (HDEF) preventing VoodooHDA.kext 3.1.0 from loading.  Note that on my HP EliteBook 850 G7, I do not need AppleALC to block AppleGFXHDA when using VoodooHDA.kext 3.1.0 with macOS Tahoe.

Edited March 16 by deeveedee

[deeveedee]

I have uploaded a new EFI to Post #1.  This new EFI is based on Open Core 1.0.7 and has changes previously discussed in this thread as listed below.  When using this EFI, note the following:

• csr-active-config = <00000000> (SIP fully enabled).  Change this as follows based on your use case:
  • csr-active-config = <01000000> for using and updating VoodooHDA.kext for audio
  • csr-active-config = <03000000> for installing VoodooHDA.kext for the first time
  • csr-active-config = <03080000> for using OCLP
• This EFI does not enable Bluetooth.  If you are using Bluetooth, you will need to enable Bluetooth in BIOS, modify USB Map to add Bluetooth USB port (HS14), add NVRAM properties for Bluetooth and add required kexts.  BrcmPatchRAM 2.7.2 is available from Acidanthera.
• I am currently using VoodooHDA.kext 3.1.2 available here for Audio in macOS Tahoe.  I have SIP csr-active-config = <01000000> to allow VoodooHDA.kext in /Library/Extensions.  This version of VoodooHDA.kext is working very well for me with outstanding audio quality and no root patches.  Since I am not using Wi-Fi on this hack, I don't need to re-apply any root patches after updating macOS Tahoe.
• If you are using BluetoolFixup.kext (not included in my posted EFI), you will need to temporarily disable BluetoolFixup.kext to permit incremental macOS updates and then re-enable BluetoolFixup.kext after applying the incremental macOS update.  Since this EFI includes Lilu.kext 1.7.2, there is no longer any need to disable WEG for macOS updates.

 

EFI Changes since my last posted EFI

 

EFI-MM81-RX560x-OC107-01

• Upgrade OC 1.0.6 -> OC 1.0.7
  • Upgrade BOOT/BOOTx64.efi
  • Upgrade OC/OpenCore.efi
  • OC/Drivers
    • Upgrade OC/Drivers/*.* (except HfsPlus.efi)
  • Upgrade OC/Tools/*.*
  • Upgrade OC/Kexts
    • AppleALC.kext 1.9.6 -> 1.9.7
    • Lilu.kext 1.7.2 Beta -> 1.7.2 Release
  • OC/ACPI
    • Add SSDT-FWHD
  • OC/config.plist
    • Add ACPI > Add > SSDT-FWHD

  

EFI-MM81-RX560x-OC106-03 (never uploaded)

• OC/Kexts
  • Upgrade Lilu.kext 1.7.1 -> 1.7.2 Beta (resolve Tahoe upgrade panic with Polaris dGPU) - no longer need boot-arg -wegnoegpu for Tahoe upgrades
• OC/ACPI
  • Add SSDT-MATH to enable Math device in macOS
• OC/config.plist
  • Add ACPI > Add > SSDT-MATH
  • Add ACPI > Patch > MATH._STA -> MATH.XSTA
  • Add pci-aspm-default DeviceProperty to HDAU audio, HDAS audio, NVME m.2 SSDs, SATA, GLAN
  • Remove boot-arg darkwake=2

 

EFI-MM81-RX560x-OC106-02 (never uploaded)

• OC/config.plist
  • Set Kernel > Quirks > DisableIoMapper = False (no longer need to enable with IntelMausiEthernet.kext 3.0.3)
• OC/Kexts
  • Upgrade IntelMausiEthernet.kext 2.5.5d0 to 3.0.0 Build 3.0.3
• OC/ACPI
  • Remove unnecessary Else conditions from SSDTs
     

 

EDIT: Using this new EFI, I upgraded to macOS Tahoe 26.4 RC.  Fast and easy with no issues.  This hack is working perfectly.

Spoiler

 

EDIT2: A couple of observations about this hack

• RX560-accelerated graphics are very responsive with Tahoe 26.4 RC.   No graphics lagging and instant screen draws.  I still have System Settings > Accessibility settings set to reduce Display Transparency and reduce Motion.
• Fans are quiet.  I don't notice any excessive fan noise from this little unit and hardly any heat from the cover.
• For its size, this little hack is amazing.  The footprint for an Intel-based device is about as small as it gets and it houses an Nvidia RX560x, 32GB RAM (expandable to 64GB), two M.2 NVME SSDs (with the mod described in this thread), Wi-Fi / Bluetooth (for those who need it) and an external SATA SSD (with the mod described in this thread).  Not bad. Not bad at all.

Edited March 24 by deeveedee

[deeveedee]

I really like using VoodooHDA.kext 3.1.2 for audio with this hack.  I listed my reasons for using VoodooHDA (instead of AppleHDA root patches) here.

[deeveedee]

I upgraded this hack to macOS Tahoe Release 26.4 (25E246).  GeekBench 6 Metal benchmarks remain consistent with the best synthetic benchmarks that I've seen on this little hack.

Spoiler

 

EDIT: Unfortunately for real MacMini 8,1 owners, OCLP Devs are still struggling with the T2 chip.  Tahoe patching for the real MacMini 8,1 is significantly delayed (maybe permanently?).  This HackMini has turned out to be a way better investment than a real MacMini 8,1.  Very happy to still be running the latest macOS on Intel.

Edited March 25 by deeveedee

[deeveedee]

After upgrading to macOS Tahoe 26.4 Release, I am finding that Ethernet may be slow to negotiate a connection after booting macOS.  See here and here for details.  If this bothers you, set Kernel > Quirks > DisableIoMapper = True in the OC config.plist.

 

EDIT: Also see mieze's recommendations here.

Edited March 30 by deeveedee

[deeveedee]

If you want to test VoodooHDA with audio over DP for the RX560x in this hack, try this new version.  If this new version of VoodooHDA works, there is no longer any reason to apply AppleHDA root patches to macOS Tahoe.

 

EDIT: Note that VoodooHDAEnableVolumeChangeFix=True (VoodooHDA.kext/Contents/Info.plist) is required for proper operation of the master Volume slider on this hack.

 

EDIT2: If you experience any sleep/wake issues when using previous versions of VoodooHDA.kext, try this version 3.3.4.

 

EDIT3: See my VoodooHDA.kext 3.3.4 test results here.

Edited April 10 by deeveedee

[deeveedee]

My next posted version of the Open Core EFI for this hack will set Kernel > Quirks > DisableIOMapper = True (déjà vu?).  The Ethernet negotiation delay is annoying (for me) and setting DisableIOMapper=True eliminates this problem.  

[deeveedee]

I performed a clean installation of macOS Tahoe 26.4.1 in a new test volume.  Much to my pleasant surprise, GeekBench 6 Metal benchmarks are among the best I've observed on this little hack!

 

Spoiler

 

Maybe the incremental installs had retained garbage from my previous experiments with the RX 560x.  I don't know for sure, but I'll take it!

Edited April 14 by deeveedee