WHEA uncorrectable error Ryzen 9 + 3070

[qus]

Hello,

 

This is a new PC, which seems to run pretty stable with some low-end office Radeon, but when the Radeon gets replaced with 3070, WHEAs start to pop. The config:

 

Ryzen 9 5900X, 3.7 GHz, 64 MB, BOX (100-100000061WOF)

ASUS B550MPlus

Gigabyte GeForce RTX 3070 Ti Gaming OC 8GB GDDR6X (GV-N307TGAMING OC-8GD)

Ripjaws V, DDR4, 32 GB, 4400MHz, CL19 (F4-4400C19D-32GVK)

 

And the crashes (bug check codes are constant):

 

Crash dump file: C:\Windows\Minidump\032023-8156-01.dmp (Minidump)

Bugcheck code: 0x124(0x0, 0xFFFFD00E785D8028, 0xB4800800, 0x60C0859)

Bugcheck name: WHEA_UNCORRECTABLE_ERROR

Driver or module in which error occurred: authenticamd.sys (authenticamd)

Bug check description: A fatal hardware error has occurred. This bug check uses the error data that is provided by the Windows Hardware Error Architecture (WHEA). This bug check is typically related to physical hardware failures. It can be heat related, defective hardware, memory or even a processor that is beginning to fail or has failed.

Analysis:

Google query: authenticamd WHEA_UNCORRECTABLE_ERROR

Crash dump file: C:\Windows\Minidump\032023-4890-01.dmp (Minidump)

Bugcheck code: 0x124(0x0, 0xFFFFB60470A9D028, 0xB4000800, 0x1010135)

Bugcheck name: WHEA_UNCORRECTABLE_ERROR

Driver or module in which error occurred: authenticamd.sys (authenticamd)

Bug check description: A fatal hardware error has occurred. This bug check uses the error data that is provided by the Windows Hardware Error Architecture (WHEA). This bug check is typically related to physical hardware failures. It can be heat related, defective hardware, memory or even a processor that is beginning to fail or has failed.

Analysis:

Google query: authenticamd WHEA_UNCORRECTABLE_ERROR

 

Is there anything I could do? Is the CPU bonkers? GPU? Mobo?

[RONOTHAN##]

WHEA errors are almost always a memory issue. Given you're trying to run your RAM at 4400 CL19, I'm gonna assume that's probably the issue. Try setting the memory to something more pedestrian for settings like 3600 CL18-22-22-42 1.35V and see if that stops the WHEA errors. 

[191x7]

Ryzen 5000 Zen3 can't do 4400. 

3600 is optimal. Many can do 3800 to 4000 but have lower performance than with 3600.

[qus]

OK, I set the memory to 3600 and system to 1800 (that should be correct, right?). But any hints where should I look for these timings and voltage settings? There's a lot of voltage settings and none of them sound like memory voltages to me...

[RONOTHAN##]

49 minutes ago, 191x7 said:

Ryzen 5000 Zen3 can't do 4400. 

3600 is optimal. Many can do 3800 to 4000 but have lower performance than with 3600.

Technically none of this is true. Zen 3 can do 4400 if you desync the infinity fabric and set it up right, if you've got a good enough board and CPU DDR4 5000 is doable as well, though performance will be worse than a tightish timing 3600 setup.

 

Almost every single Ryzen 5000 series CPU can do 3800MT/s with synced FCLK though, and performance does scale to that point, with some chips that scale all the way up to 4000 though those are very rare. Heck, even AMD said that the sweet spot for Ryzen 5000 was 3733 (conservative for the handful of chips that can't do 1900MHz FCLK), it's just that so few kits exist at that XMP that no one recommended going for it. 

 

9 minutes ago, qus said:

OK, I set the memory to 3600 and system to 1800 (that should be correct, right?). But any hints where should I look for these timings and voltage settings? There's a lot of voltage settings and none of them sound like memory voltages to me...

Memory timings should have a their own dedicated menu usually called something like advanced memory configuration. The timings you need to set are in the order tCL-tRCD-tRP-tRAS. Set the Rd and WR versions of tRCD to the same value. Since that's an ASUS board though, they have a tendency to spell out the timings, so look for Cas latency for tCL, RAS to Cas delay for tRCD, RAS to Pre for tRP, and RAS act timing for tRAS. For voltages, there should be one called DRAM voltage, that's the memory one. Set that to 1.35V. 

[qus]

Hmm, my current settings in BIOS and CPU-Z. Still crashing.

 

[duckargent]

4 hours ago, RONOTHAN## said:

WHEA errors are almost always a memory issue.

Not always a memory issue. You get WHEA error also due to a faulty CPU. An unstable CPU overclock always leads to WHEA error.

 

Since he gets the error when he installs the 3070, it might be a problem with the board.

[qus]

Well, actually tried it with 1060 (or something similar), it also crashes, but takes longer to blue screen. I think it's hopeless, I'm sending whole PC back for RMA.

[duckargent]

Not sure if it helps, they may be too general, but you can try these:

 

https://appuals.com/authenticamd-sys-bsod-windows/

 

https://www.howto-connect.com/fix-authenticamd-sys-bsod-blue-screen-error-windows-10-or-11/