Resolved the WHEA-Logger Event 18 issue on Ryzen 3700X. No lasting negative<|pad|> can have been noted.
Resolved the WHEA-Logger Event 18 issue on Ryzen 3700X. No lasting negative<|pad|> can have been noted.
Here’s a revised version of your message:
Hello everyone,
I’m sharing this to both record the solution and pose a brief question regarding long-term impacts.
System Specifications: Ryzen 7 3700X, Gigabyte GA-AX370 Gaming, 5 GB DDR4 3200 MHz (XMP enabled), RTX 2070 Super Cooler Master MWE Bronze V2, 750W PSU.
The Problem: I encountered WHEA-Logger Event ID 18 errors, specifically:
> "A fatal hardware error has occurred. Error Type: Cache Hierarchy Error."
Symptoms included random crashes or reboots, often shortly after startup. There was no BSOD—just freezing and a reset. Initially, I thought the PSU, GPU, RAM, Windows, or even the USB RGB controller might be the issue. After removing the RGB controller and reinstalling Windows, the problems persisted.
What I think caused it: Instability in the CPU’s memory controller (Infinity Fabric) when running RAM at 3200 MHz. Over time, certain Ryzen 3000 chips become more sensitive to memory speed and SoC voltage. The motherboard’s auto-voltage settings sometimes dropped too low under certain conditions.
What I Changed (The Fix):
- RAM speed: Reduced from XMP 3200 MT/s to 2933 MT/s
- DRAM voltage: Changed from ~1.21–1.25 V (auto) to 1.35 V fixed
- SoC voltage: Switched from auto (~0.95–1.00 V) to 1.05 V fixed
- PBO: Disabled
Results: The system has remained stable for over a month, with no WHEA errors, crashes, freezes, or unusual behavior.
Why this worked:
- 2933 MHz slightly eases memory controller stress.
- 1.35 V DRAM provides more stable memory signals.
- Disabling PBO avoids unnecessary voltage fluctuations.
This aligns with the known tendency of some Ryzen 3000 chips to show minor aging effects on memory speed and voltage.
My Question: Are there any potential long-term drawbacks to these settings—using RAM at 2933 MT/s, SoC voltage at 1.05 V fixed, and PBO disabled? I’d appreciate any insights from experienced users who have worked with Zen 2 tuning over time. Thanks in advance!
Hope this helps someone else facing similar WHEA cache issues.
Was 1V really what was active before or after a Clear CMOS reset? 1V falls short of powering anything beyond around 2666MT/s, which likely explains your trouble. Motherboards typically set the SOC voltage to about 1.1V at 3200MT/s, so discrepancies might stem from a faulty BIOS that misbehaves with auto-voltages—Gigabyte’s X370 boards are a plausible example. Voltages you’re seeing are generally stable and safe, as long as you’re only concerned with basic operation. The SOC voltage remains acceptable up to roughly 1.2V, though most manufacturers avoid exceeding 1.15V for reliability. DDR4 is essential according to JEDEC standards to maintain 1.35V operation, meaning any memory failure points to a production issue rather than voltage problems.
No there's no downsides I think... I had similar issues, in the end how I fixed it is buy B-die RAM, lol! It wasn't really fixable with my old Corsair RAM, it's just trash ... So in the end it might have something to do with the controller but it really is the cheapo RAM some manufacturers use, it's just too sensitive (I think???) You should be able to run PBO btw, but no guarantees the issues won't come back either tbh... If you can get the B-die RAM it's like magic even overclocks *easily* (tldr: it's not your CPU fault)
I guess I had to reduce my SOC for the OC to work properly, not sure how much... RAM voltage set to 1.36 too—both my MSI boards hate that and they’ll either crash or fail to boot.
All right, focusing on the key points: yes, it's safe. My 3600, 5800X and 5700X3D models needed manual SoC voltage settings between 1100-1150mV to avoid errors above 1600MHz when clocks were involved.
Thanks for the feedback, this matches my expectations and confirms it with insights from experienced users. The main takeaways are: the SoC operating around 1.05 V is safe, the 1.35 V DRAM is typical for DDR4 XMP setups, and different boards manage auto-volt settings in varied ways. Gigabyte’s X370 Auto SoC ended up too low for my specific IMC and RAM combo. Some Ryzen processors require a bit more IMC as they age, which explains the gradual issue. Overall, the system has performed well for over a month, and performance differences between 2933 and 3200 are minimal for my use case. I might reconsider enabling PBO later, but right now stability and reliability are the priorities. Thanks for the discussion—it’s helpful to hear others’ perspectives on WHEA cache hierarchy problems on Zen 2.