[SUCCESS] ASUS ROG Crosshair X870E Extreme / Ryzen 9 7950X / RX 6950XT / Dual Boot macOS Ventura+Windows 11

[KWODO]

UPDATED ON 4 JANUARY 2026



MY SECOND AMD HACKINTOSH
and its working perfectly - big thanks to this awesome community







MAINBOARD

ASUS ROG Crosshair X870E Extreme (BIOS 1715)
ROG-X870E-X
Realtek SupremeFX USB Audio (ALC4082) √
Realtek 5Gbit Network Adapter (RTL8126) X (not supported in macOS) (Disabled in BIOS)
Aquantia/Marvell FastLinQ Edge 10Gbit Network Adapter (AQC113CS) √
MediaTek Wi-Fi 7 BT5.3 (MT7927) X (not supported in macOS) (removed from mainboard)
ASMedia USB4 (ASM4242) X (not supported in macOS)

VRM-Fan @ 3000 RPM / 10% fixed (PWM-Mode)


USB-Mapping

1x Front USB Type-C Port with switch (USB3.2)
U20G_C6 (1)

4x Front USB Type-A Ports (USB3.0)
U5G_E1234 (18)

2x Rear USB Type-C Port with switch (USB3.1)
U10G_C23/U10G_C22 (14)

8x Rear USB Type-A Ports (USB2.0)
U10G_3/U10G_4/U10G_18 (2) / U10G_19 (11) / U10G_20 (13) / U10G_9/U10G_8/U10G_21 (5)

1x Internal USB 2.0 Port for Bluetooth
M.2(WIFI) (15)

1x Internal USB 2.0 Port for USB Audio
USB Audio (6)(16)






CPU

AMD Ryzen 9 7950X 4.5 GHz AM5
100-100000514WOF
iGPU AMD Radeon Graphics X (disabled in BIOS)
PBO-Enhanced (Level 2 80°C 170W) -> temps @ idle 37°C / max 81,6°C



RAM

Kingston FURY Beast DDR5-5200 MHz 288-Pin 128 GB-Kit (4 x 32 GB)
KF552C40BBK4-128
(Default (JEDEC) configuration for 4 modules in dual channel running @ 3600 MHz / CL 30-29-29-58 / 1.10v)



PCI EXPRESS CARDS

Sapphire TOXIC AMD Radeon RX 6950 XT Limited Edition 16 GB √
11317-01-20G
(Riser PCIe-Slot)

OC-BIOS Mode running @ 2565 MHz Core speed


Thermaltake PCIe 4.0 Dual 90 Degree Riser Cable (130mm / black)
AC-071-CO1OTN-C1
(PCIe-Slot1) PCIEX16(G5)_1 (A)


Broadcom BCM94360CD √
6871658863422
(M.2 NGFF Slot) M.2(WIFI) (15)



DISPLAY

Dell 34“ Curved Monitor
S3422DW
WQHD 3440 x 1440


HP Series 5 Pro 34“ Curved Monitor
534PM
WQHD 3440 x 1440



STORAGE

Internal

WD Black SN850X NVMe SSD 2TB (Windows 11)
WDS200T2X0E-00BCA0
(Storage Interface Socket 3) M.2_1 (7)


WD Black SN850X NVMe SSD 1TB (macOS Ventura)
WDS100T2X0E-00BCA0
(Q-DIMM.2 Slot) DIMM.2_1 (8)


WD Black SN850X NVMe SSD 2TB (macOS Beta/Test)
WDS100T2X0E-00BCA0
(Q-DIMM.2 Slot) DIMM.2_2 (8)


KingSpec P3 SSD 2TB SATA-3 (Grub/OC/OC-Debug/OC-Test/Ubuntu Linux/Acronis Rescue Boot macOS+WinRT)
B0BKL4PQM2
(SATA-3 Port 1) SATA6G_1 (9)


KingSpec P3 SSD 4TB SATA-3 (Games ExFAT)
B0BKL4PQM2
(SATA-3 Port 2) SATA6G_2 (9)


Seagate Exos 7E8 8TB SATA-3 (System-Backups ExFAT)
ST8000NM000A-2KE101
(SATA-3 Port 3) SATA6G_3 (9)


Seagate Exos 7E10 8TB SATA-3 (Data ExFAT)
ST8000NM017B-2TJ103
(SATA-3 Port 4) SATA6G_4 (9)



External

Seagate Exos X22 22TB SATA-3 (Backups ExFAT)
ST22000NM001E
(USB-A 3.0 Enclosure - Sabrent EC-KSL3)


WD Green SN350 NVMe 2280 2TB (Data ExFAT)
WDS100T2G0C-00CDH0
(USB-C 40gbps - Thunderbolt 3/4 & USB 4.0 Enclosure - Yottamaster YACOM2-U4)


Crucial P5 Plus NVMe 2280 512GB (Data ExFAT)
MTFDKBK512QFM
(USB-C 40gbps - Thunderbolt 3/4 & USB 4.0 Enclosure - ORICO M234C3-U4-GY-BP-V1)


Sabrent Rocket Nano NVMe 2242 SSD 512GB (Data ExFAT)
SB-1342-512
(USB-C 10gbps - USB 3.1 (Gen2) Enclosure - IcyBox IB-1817MC-C31)



CASE

Thermaltake CTE C750 Air Snow Full Tower Chassis
CA-1X6-00F6WN-00



CABLES

AsiaHorse 16AWG Sleeved Cable Kit (300mm / white)
B0B4V8T8BD
Sleeved Cables for CPU (4+4-Pin) & Mainboard (24-Pin)


deleyCON 18AWG Sleeved PCIe Cable Kit (300mm / black & grey)
MK5570
Sleeved Cables for GPU (8+8+6-Pin)



PSU

NZXT C1000 Gold
PA-0G1BB-EU



CPU COOLING

Thermaltake TH420 V2 Ultra ARGB AIO Liquid Cooler Snow Edition (420mm)
CL-W407-PL14SW-A
(Back-Intake with 3x Thermaltake CT140 White ARGB PWM Fans)

Waterpump @ 2000 RPM / 50% fixed (PWM-Mode) -> Liquid-temps @ idle 24°C / max 35°C
Radiator-Fans @ 500 RPM / 20% fixed (PWM-Mode)


Thermal Grizzly AM5 Contact Sealing Frame
TG-CSF-AM5



GPU COOLING

Sapphire TOXIC AMD Radeon RX 6950 XT Limited Edition Build-in Asetek AIO Liquid Hybrid Cooler (360mm)
11317-01-20G
(Front-Intake with 3x Sapphire branded Asetek ARGB 120mm Fans)

ZeroRPM disabled (BIOS Mod)



SYSTEM COOLING

Thermaltake CT140 White PC Cooling Fan
CL-F152-PL14WT-A
(3x Bottom-Intake + 2x Top-Exhaust)

Chassis-Fans @ 500 RPM / 20% fixed (PWM-Mode)



AUDIO INTERFACE

Focusrite Scarlett Solo 4th Generation USB √ Output: Logitech Speaker System Z523 / Input: E-Guitar
PCM0017717-000


Creative Sound Blaster PLAY! 3 USB DAC AMP and External Sound Card √ Plantronics RIG 400HX XBOX Headset with Microphone
0054651191273



KEYBOARD/MOUSE/CONTROLLER

Logitech Craft Wireless Keyboard (Win/Mac)
920-008484


Logitech G502 LIGHTSPEED Wireless Gaming Mouse
MR0076


Xbox Wireless Controller (Robot White)
QAT-00009
(with Xbox Wireless Adapter for Windows)

​

check my Github page for files and more information

If anyone finds mistakes or has suggestions for improvement, please post them here, thank you

 

[fabiosun]

Realtek SupremeFX USB Audio (ALC4082) X

This audio world perfectly
No needs any kext or driver only a correct usb mapping

 

[KWODO]

This audio world perfectly
No needs any kext or driver only a correct usb mapping

Thanks Fabio for the tip. USB Audio is working without AppleALCU.kext. I have changed it.

 

[dalexhz]

Hi. Did anyone have a success with the latest bios (i.e. 1804), by chance? It would seem that they broke something once again.

 

[fabiosun]

On my pc @dalexhz 1804/1805 need new acpi patches to boot
But for both bioses i had serious problems with sata disks controller
I reverted to 1715 which works perfectly

 

[dalexhz]

On my pc @dalexhz 1804/1805 need new acpi patches to boot
But for both bioses i had serious problems with sata disks controller
I reverted to 1715 which works perfectly

Well it would seem that no release is perfect. 1715 is quite buggy when it comes to memory training. It's fatally inconsistent. Like in incredibly, hellishly inconsistent. Got itstable at 6400 MT/s CL28 then made a reboot only to see that it fails Karhu RAM test in two seconds... Reboot and it's "stable" again. Let them (Asus and their employees) be thrice damned for their disastrous incompetency. It looks like they aren't capable of making a single release without blowing that thing up their faces at least in some way. This behaviour is so bizarre that I'm finding myself questioning my own perceprion of reality.

 

[KWODO]

Well it would seem that no release is perfect. 1715 is quite buggy when it comes to memory training. It's fatally inconsistent. Like in incredibly, hellishly inconsistent. Got itstable at 6400 MT/s CL28 then made a reboot only to see that it fails Karhu RAM test in two seconds... Reboot and it's "stable" again. Let them (Asus and their employees) be thrice damned for their disastrous incompetency. It looks like they aren't capable of making a single release without blowing that thing up their faces at least in some way. This behaviour is so bizarre that I'm finding myself questioning my own perceprion of reality.

I also have zero problems with the 1715 BIOS. My RAMs are only running at 3600 (4 sticks), though.

 

[dalexhz]

@KWODO I've had all of my trouble solved by switching back to the good old KVM-based hardware virtualization. Been using that for a lifetime, since I'm using it whenever it is possible. Now I can just flash the latest bios whenever it is getting released, without caring about consequences. No need for all that platform-specific nonsense when you can setup a virtual machine and use PCIe passthrough for necessary devices (GPUs, NVMe SSDs etc).







The benefits are obvious.

First, I no longer need to care about those pesky bios updates, DSDT patches and whatever nonsense. Even better, I can boot the very same macOS partition that had been originally used on the bare metal, without any modifications. I had to add a few kernel patches though. The macOS kernel doesn't like being virtualized, so the kernel patches are needed to make sure that Bluetooth and Handoff are functional. On the positive side, you will only need to make it once and there will be no need to think about this again.

For the second, I'm able to run docker and whatever on the linux virtual machines without resorting to the binary translation, which is painfully slow.

Oh and there's no need to use AMD-related CPUID patches since the KVM itself is able to fool the virtualized OS into believing that it's running on Intel, regardless of the real hardware.

The downsides are obvious too.

1) I had to sacrifice some RAM for the linux host OS. One gigabyte had been enough for all that disk cache and whatever stuff. Doesn't matter much for me, but that can be sensitive for those who aren't ready to install the dual-rank memory modules.
2) There are some bugs which are frustrating. For some unfathomable reason, it is not possible to use the original Broadcom-based Wi-Fi adapter natively, unless you're restricting yourself to the linux kernel 5.15.x on your host. Any attempt to use the newer linux kernel version will lead to the kernel panic after entering the Wi-Fi password in your macOS. Curiously, though, the unencrypted WI-Fi connections aren't causing the kernel panic and are working fine. Not that this is useful in any way, since I'm not going to fallback to unencrypted Wi-Fi in 2026.

I think it would be a good idea to try some other virtualization solutions and see whether they're better or worse. FreeBSD's bHyve comes to mind. But that will require a lot of research and experimentation. KVM works perfectly for now, though.

P.S. If anyone would like to ask, I'd be happy to provide the copies of my opencore config as well as the libvirt XML files. That should be enough to reproduce my configuration. Not going to publish it here, though, since libvirt is fairly difficult for the general people, especially for those that aren't used to it.

Perhaps, I should consider making it work "natively", but I'm too lazy to actually care about losing 1-2% of performance in exchange for bothering about the BIOS updates.

 

[dalexhz]

My RAMs are only running at 3600 (4 sticks), though.

That's bad news actually. It may be enough for the likes of 9700X or 9800X3D, but not nearly enough to feed the 16 cores. I'd try to look for the single-rank 32 Gb modules (those are rare, but are said to exist) in order to get higher frequency and bandwidth.

I’ve done some testing with Linpack Extreme (6th option when it asks you for the problem size) and 9950X3D.

Look for yourself:

4800 MT/s: 809 GFlops
5600 MT/s: 865 GFlops
6200 MT/s: 920 GFlops
6400 MT/s: 980 GFlops
8200 MT/s: 991 GFlops
8400 MT/s: 1001 GFlops

As you can see, 6400 MT/s is a sweet spot while 6200 MT/s is a second best. Higher frequencies will require you to restrict yourself to 48 GB (2x24 GB) or lower. That is why I have two 48Gb modules running at 6400 MT/s in my own system. There’s no point in trying any better.

I didn't even bother to test at lower frequency since it is obvious that it will be painfully slow. I might try to check it out but I’m sure that it will give me some ridiculous results.

My conclusion is simple. Unless you're able to run your memory at 5600 MT/s or faster, there's literally no point in wasting your money on the 12 or 16-core AM5 CPUs. It's far better choice to buy Threadripper or Xeon in that case. Same is correct for Intel by the way… There’s no point in buying 13900K or 14900K unless you’re able to make sure that your memory is stable at 6400+ MT/s.

Yes, you’ve got it right. The consumer CPUs are really THAT handicapped. Both Intel and AMD are managed by the greedy bastards who are busy with handicapping their own hardware.

 

[fabiosun]

For people interested to understand why 1715 is good and laterst not so good this is a screenshot where is visible that 2001 looses some drives in my rig :

 

[KWODO]

@KWODO I've had all of my trouble solved by switching back to the good old KVM-based hardware virtualization. Been using that for a lifetime, since I'm using it whenever it is possible. Now I can just flash the latest bios whenever it is getting released, without caring about consequences. No need for all that platform-specific nonsense when you can setup a virtual machine and use PCIe passthrough for necessary devices (GPUs, NVMe SSDs etc).

View attachment 18625

View attachment 18626

View attachment 18627

The benefits are obvious.

First, I no longer need to care about those pesky bios updates, DSDT patches and whatever nonsense. Even better, I can boot the very same macOS partition that had been originally used on the bare metal, without any modifications. I had to add a few kernel patches though. The macOS kernel doesn't like being virtualized, so the kernel patches are needed to make sure that Bluetooth and Handoff are functional. On the positive side, you will only need to make it once and there will be no need to think about this again.

For the second, I'm able to run docker and whatever on the linux virtual machines without resorting to the binary translation, which is painfully slow.

Oh and there's no need to use AMD-related CPUID patches since the KVM itself is able to fool the virtualized OS into believing that it's running on Intel, regardless of the real hardware.

The downsides are obvious too.

1) I had to sacrifice some RAM for the linux host OS. One gigabyte had been enough for all that disk cache and whatever stuff. Doesn't matter much for me, but that can be sensitive for those who aren't ready to install the dual-rank memory modules.
2) There are some bugs which are frustrating. For some unfathomable reason, it is not possible to use the original Broadcom-based Wi-Fi adapter natively, unless you're restricting yourself to the linux kernel 5.15.x on your host. Any attempt to use the newer linux kernel version will lead to the kernel panic after entering the Wi-Fi password in your macOS. Curiously, though, the unencrypted WI-Fi connections aren't causing the kernel panic and are working fine. Not that this is useful in any way, since I'm not going to fallback to unencrypted Wi-Fi in 2026.

I think it would be a good idea to try some other virtualization solutions and see whether they're better or worse. FreeBSD's bHyve comes to mind. But that will require a lot of research and experimentation. KVM works perfectly for now, though.

P.S. If anyone would like to ask, I'd be happy to provide the copies of my opencore config as well as the libvirt XML files. That should be enough to reproduce my configuration. Not going to publish it here, though, since libvirt is fairly difficult for the general people, especially for those that aren't used to it.

Perhaps, I should consider making it work "natively", but I'm too lazy to actually care about losing 1-2% of performance in exchange for bothering about the BIOS updates.

With my old x670 motherboard, I once tried virtualization as well. Proxmox KVM with GPU passthrough. I only had occasional problems with starting (black screen), but once it booted, everything ran smoothly. However, that was some time ago (2022) and back then there were apparently issues with the Proxmox Linux kernel. I once tried an older one but then went back to the "old way." And that works. At least now with the Asus motherboard. The old Gigabyte had catastrophic BIOS problems, basically. Which KVM are you using? I'm curious.

By the way. Nice wallpaper in your room! I like it.
And cheers (Hans Baer Riesling) tasty

 

[KWODO]

That's bad news actually. It may be enough for the likes of 9700X or 9800X3D, but not nearly enough to feed the 16 cores. I'd try to look for the single-rank 32 Gb modules (those are rare, but are said to exist) in order to get higher frequency and bandwidth.

I’ve done some testing with Linpack Extreme (6th option when it asks you for the problem size) and 9950X3D.

Look for yourself:

4800 MT/s: 809 GFlops
5600 MT/s: 865 GFlops
6200 MT/s: 920 GFlops
6400 MT/s: 980 GFlops
8200 MT/s: 991 GFlops
8400 MT/s: 1001 GFlops

As you can see, 6400 MT/s is a sweet spot while 6200 MT/s is a second best. Higher frequencies will require you to restrict yourself to 48 GB (2x24 GB) or lower. That is why I have two 48Gb modules running at 6400 MT/s in my own system. There’s no point in trying any better.

I didn't even bother to test at lower frequency since it is obvious that it will be painfully slow. I might try to check it out but I’m sure that it will give me some ridiculous results.

My conclusion is simple. Unless you're able to run your memory at 5600 MT/s or faster, there's literally no point in wasting your money on the 12 or 16-core AM5 CPUs. It's far better choice to buy Threadripper or Xeon in that case. Same is correct for Intel by the way… There’s no point in buying 13900K or 14900K unless you’re able to make sure that your memory is stable at 6400+ MT/s.

Yes, you’ve got it right. The consumer CPUs are really THAT handicapped. Both Intel and AMD are managed by the greedy bastards who are busy with handicapping their own hardware.

I actually wanted to switch to 2x64GB or wait to see if maybe 128GB sticks would come out in 2026 or later. But then the price increase happened, so I'm sticking with the 4 sticks and 3600 for now.

 

[dalexhz]

I actually wanted to switch to 2x64GB or wait to see if maybe 128GB sticks would come out in 2026 or later. But then the price increase happened, so I'm sticking with the 4 sticks and 3600 for now.

When it comes to content creation, Core Ultra is a good alternative and their memory controllers are far more tolerant. I’ve seen the people being able to stabilize the two dual-rank modules on 7800 or even 8000 MT/s with CL32. Means that you can have up to 96 Gb of RAM operating at the reasonably high frequency and making it possible to get over 1 TFlops in Linpack without much of a ruckus.

That would require the dual-slot motherboard and the water cooling though. Considering a fact that dual-rank modules are quite hot even at the JEDEC settings, any attempt to do so with the air cooling will inevitably fail. There’s no point in even trying such a thing.

 

[dalexhz]

With my old x670 motherboard, I once tried virtualization as well. Proxmox KVM with GPU passthrough. I only had occasional problems with starting (black screen), but once it booted, everything ran smoothly. However, that was some time ago (2022) and back then there were apparently issues with the Proxmox Linux kernel. I once tried an older one but then went back to the "old way." And that works. At least now with the Asus motherboard. The old Gigabyte had catastrophic BIOS problems, basically. Which KVM are you using? I'm curious.

To tell you frankly, I don’t take all those “new way” lunatics seriously. They seem to be using the virtualization for a sole purpose of using the virtualization and nothing else. None of them are extracting the benefits that are coming from it. Furthermore, they’re preaching about the OpenCore and its “unreliability” while “forgetting” to mention that they’re using the OpenCore as well. That’s a donation-collecting hypocrisy at its finest. Most of them aren’t even bothered to use something other than the emulated SATA controller which is frankly saying a lot about them.

I’m running the latest Ubuntu release as a host. The only thing that’s different is a kernel of course. I had to replace it with the old Proxmox kernel in order to get the encrypted Wi-Fi working. There are no black screens, no hanging up and whatever. Everything just works as it is.

However, I eventually found myself thinking of trying the FreeBSD as a host. If I could make the bHyve-based hackintosh config up and running that that would be awesome. Because the old Linux kernel is only able to last for so long. It will eventually get stuck on some newer hardware, making this approach limited in terms of possible lifetime.

P.S. I’ve got the VMware ESX working by the way. It doesn’t have any trouble with Wi-Fi and whatever. But it’s proprietary and I don’t like such a thing for the host. It might be tolerable for Microsoft, since their products are being used anyway, but there’s no point in shackling yourself up to the yet another corporation when this can be avoided. I’m sure that making the Hyper-V to work would be even easier, but I didn’t bother trying yet. Might be a usable option for those who need Windows running on the regular basis though.