[SUCCESS] First documented bare-metal macOS Sequoia 15.7 on AMD EPYC server CPU (Zen 2 Rome, 16C/32T, SMT active) — Geekbench 6 multi 8207

[AGENCEPASEO]

Hello EliteMacx86 community,

After 40+ boot attempts and weeks of work, I got macOS Sequoia 15.7.7 booting bare-metal on an AMD EPYC server platform with SMT fully active. After extensive searches across forums in EN/FR/DE/CN/RU/VN/KR/JP, I couldn't find a single previously documented success of bare-metal Sequoia on AMD EPYC. The only EPYC + macOS reference out there is a Proxmox VM with PCIe passthrough, which doesn't count as bare metal.

Hardware:
- Supermicro H11SSL-i v2.0 (server board, IPMI/BMC ASPEED)
- AMD EPYC 7302P (Zen 2 Rome, 16C/32T)
- AMD RX 6800 16GB (RDNA 2)
- 64GB DDR4 ECC RDIMM
- Crucial P310 2TB NVMe
- macOS Sequoia 15.7.7 (build 24G720)
- OpenCore 1.0.5 RELEASE

Results:
- 32 SMT threads recognized by macOS (hw.ncpu=32, HTT flag active)
- Geekbench 6 Single: 1217 (+7.3% vs world average for this CPU)
- Geekbench 6 Multi: 8207 (+3.6% vs world average)
- GPU OpenCL RX 6800: 89761 (Mac Pro 2019 level)
- Native Ethernet, audio, USB, daily-driver stable

Public Geekbench result (verifiable):

https://browser.geekbench.com/v6/cpu/18196255

Full writeup with the technical story (failed attempts, dead ends, the 4 key conceptual blockers, BIOS settings, credits):

Hackintosh AMD EPYC Sequoia : Première mondiale bare metal

Premier hackintosh macOS Sequoia bare metal sur serveur AMD EPYC documenté au monde. 32 threads SMT actifs, 8207 Geekbench

paseo.cloud

Huge thanks to Micking (Michael Perche) who unblocked me on the T2/SMBIOS issue when I was about to quit, and to the Acidanthera + AMD_Vanilla maintainers without whom none of this works.

I'm not publishing my config.plist, EFI or kernel patches — this is a personal/educational experiment, not a product. macOS on non-Apple hardware violates Apple's EULA.

Happy to discuss the technical hurdles in the comments — especially EPYC IOHub high-MMIO, BMC firmware collisions, and SMT activation on AMD server boards (NOT solved by CpuTopologyRebuild, which is Intel-only and silently triple-faults on Zen 2, confirmed empirically).

— Jeremy "PASEO" Choulant

 

[EliteMacx86]

Hello EliteMacx86 community,

After 40+ boot attempts and weeks of work, I got macOS Sequoia 15.7.7 booting bare-metal on an AMD EPYC server platform with SMT fully active. After extensive searches across forums in EN/FR/DE/CN/RU/VN/KR/JP, I couldn't find a single previously documented success of bare-metal Sequoia on AMD EPYC. The only EPYC + macOS reference out there is a Proxmox VM with PCIe passthrough, which doesn't count as bare metal.

Hardware:
- Supermicro H11SSL-i v2.0 (server board, IPMI/BMC ASPEED)
- AMD EPYC 7302P (Zen 2 Rome, 16C/32T)
- AMD RX 6800 16GB (RDNA 2)
- 64GB DDR4 ECC RDIMM
- Crucial P310 2TB NVMe
- macOS Sequoia 15.7.7 (build 24G720)
- OpenCore 1.0.5 RELEASE

Results:
- 32 SMT threads recognized by macOS (hw.ncpu=32, HTT flag active)
- Geekbench 6 Single: 1217 (+7.3% vs world average for this CPU)
- Geekbench 6 Multi: 8207 (+3.6% vs world average)
- GPU OpenCL RX 6800: 89761 (Mac Pro 2019 level)
- Native Ethernet, audio, USB, daily-driver stable

Public Geekbench result (verifiable):

https://browser.geekbench.com/v6/cpu/18196255

Full writeup with the technical story (failed attempts, dead ends, the 4 key conceptual blockers, BIOS settings, credits):

Hackintosh AMD EPYC Sequoia : Première mondiale bare metal

Premier hackintosh macOS Sequoia bare metal sur serveur AMD EPYC documenté au monde. 32 threads SMT actifs, 8207 Geekbench

paseo.cloud

Huge thanks to Micking (Michael Perche) who unblocked me on the T2/SMBIOS issue when I was about to quit, and to the Acidanthera + AMD_Vanilla maintainers without whom none of this works.

I'm not publishing my config.plist, EFI or kernel patches — this is a personal/educational experiment, not a product. macOS on non-Apple hardware violates Apple's EULA.

Happy to discuss the technical hurdles in the comments — especially EPYC IOHub high-MMIO, BMC firmware collisions, and SMT activation on AMD server boards (NOT solved by CpuTopologyRebuild, which is Intel-only and silently triple-faults on Zen 2, confirmed empirically).

— Jeremy "PASEO" Choulant

This is nice. Have tried Threadrippers so far but never got hands on EPYC. Would be nice to know the Cinebench scores if you can post it here.

Also, insights on the SMT activation would be useful for the community as a whole. Could you clarify what you mean by the IOHub high-MMIO and BMC firmware collisons?

 

[MickinG]

Technical Record and Statement of Facts​

The technical and chronological reality of any computing project relies entirely on verifiable data. The objective facts are as follows:

1. Concept and Post-Installation Handover​

To ensure complete accuracy, it is confirmed that the original idea to attempt a Hackintosh configuration on this platform, as well as the subsequent post-installation tasks (such as USB mapping, Bluetooth setup, and autoboot configurations), were entirely his own initiative. I never endorsed or validated the concept of turning this architecture into a Hackintosh; my intervention was strictly limited to solving the system blockages to make the machine bootable. However, the public narrative concerning the project's overall timeline, the hardware selection process, and the core kernel debugging phase contains significant inaccuracies.

2. Hardware History and Architectural Origin​

The configuration used (Supermicro H11SSL-i v2.0, AMD EPYC 7302P, ECC RAM) replicates component-for-component the exact architecture of the enterprise infrastructure server that I personally advised on, selected, and configured two years ago for its original deployment under Proxmox VE. Whether this project represents the physical recycling of that exact machine or the identical repurchase of a second identical datacenter platform for Hackintosh purposes, the choice of this highly specific hardware stems directly from that initial infrastructure, which was technically validated at the time.

3. The Failure of Automated Methods (AI & Prompts)​

The recent attempt to transition this platform to bare-metal macOS Sequoia utilizing automated templates and prompt engineering resulted in a complete technical impasse, causing dozens of failed boot attempts and systematic black screens. Automated tools were fundamentally incapable of diagnosing or resolving the primary blocker: the physical memory collision at high-MMIO address 0xCFC000, which is natively reserved by the server board's BMC ASPEED administration firmware.

4. Current Project State and Admission​

As explicitly stated in writing on international specialized forums, the project was "about to quit" prior to my engineering intervention. The surgical countermeasures implemented during my low-level debugging session—specifically resolving the kernelmanagerd process panic, applying the correct SMBIOS profile switch to bypass Apple T2 security dependencies, and injecting DummyPowerManagement—are what allowed the platform to achieve its very first successful boot.

5. Final System Structure​

A finalized OpenCore EFI folder capable of natively booting macOS Sequoia on a datacenter architecture while utilizing only 6 base kexts and zero on-the-fly ACPI patches is the definitive signature of a structured, human engineering intervention.

The complete history of remote access sessions, before-and-after crash logs, and the exact chronology of technical file modifications are fully archived and secured. Source code and timestamps cannot be rewritten; they simply record the reality of the work accomplished.