Constant Hard Crash, unable to start a game

GPU, PSU, and RAM can be ruled out (all tested/replaced), and other settings—PBO, PPT, SoC voltage, PCIe Gen3, C-states—had no effect. The key finding: disabling one CCD on the Ryzen 9 5900X completely fixes the issue.

This indicates that the B550M DS3H motherboard is likely the limiting factor. Its simpler power delivery, limited SoC VRM design, and less optimized Infinity Fabric routing struggle under BF6’s heavy multithreaded shader compilation across two CCDs. The fact that the system is stable with only one CCD active suggests a platform limitation rather than faulty CPU, GPU, RAM, or PSU.

Recommended solution: run BF6 on a single CCD (BIOS or CPU sets/ Process Lasso) or consider a motherboard upgrade for full dual-CCD support.

I park it on half my cores with a user.cfg file and it works wonders. In game i dont spike over 30% cpu usage anymore. No more cpu frame drops, just bad hitreg 

Exact same problem, only started when i upgraded from a 5500 to a 5900x. Its a cpu scheduling/latency problem that only happens in battlefield

GPU, PSU or RAM memory are definitely the problem, as I've changed them (Tested with RTX 4060, OCZ 750W PSU and Kingston RAM

I've tested all those options 

• Disable PBO

• Set CPU PPT ≈ 120 W

• Set SoC voltage ≈ 1.05 V

• Force PCIe Gen 3

• Disable Global C-States

None of them did any effect

But, i've messed with one thing that worked, Ryzen 9 5900X have 2 CCDs, disabling one (Using only one) in BIOS did worked
So this problem has got to have something with threads spread along those cores

Diagnostics → Windows Error Reporting (summary)

Multiple GPU timeout failures are recorded:

• Bugcheck 0x116 – VIDEO_TDR_FAILURE
• LiveKernelEvent 141 / 117 – GPU hang, reset failed
• Error code 0xC000009A – insufficient resources

Meaning:
The GPU stopped responding → Windows attempted recovery → the GPU never recovered → the system hard-locked or crashed.

Why this matches the symptoms

This explains:

• Complete system freeze
• No display updates
• No keyboard or mouse input
• Power button unresponsive
• No useful logs

Low-level TDR failures often end this way, especially on Ryzen + B550 systems.

Why this is not a driver or Windows issue

If it were:

• A bad NVIDIA driver
• Corrupt Windows install
• A DirectX issue

You would see:

• nvlddmkm errors
• Clean BSODs
• Recoverable crashes

Instead, the 141 → 117 → 116 chain points to hardware / power delivery instability, not software.

Important detail

The errors reference Windows 26200 (Insider), meaning they occurred before you rolled back.
Windows exposed the issue — it did not cause it.

Final diagnosis

Confirmed cause:
VRM / SoC instability under transient GPU load

System combination:

• Gigabyte B550M DS3H
• Ryzen 9 5900X
• RTX 5070

This motherboard is not designed for modern high-transient GPUs combined with simultaneous CPU boost.
Battlefield shader compilation is a perfect trigger.

Next steps

Confirmation test (recommended):

• Set GPU power limit to 75–80%
• Run Battlefield shader compilation
• If freezes stop → cause confirmed
• If freezes continue → rare GPU issue possible

Mitigation (if keeping the board):

• Disable PBO
• Set CPU PPT ≈ 120 W
• Set SoC voltage ≈ 1.05 V
• Force PCIe Gen 3
• Disable Global C-States

I hope this helps explain the issue :)