Hey all,
I bought the mobo and the CPU in May 2024, along with a ROG STRIX LC II 360 ARGB. I always kept the default BIOS settings (AI Optimized, Auto everything, etc.), only adjusted the fan curve to be a bit more quiet. With this curve, the CPU temperature never exceeded 80 degrees for longer than a few seconds. All was well until a couple of months ago, when:
- Games under windows started crashing;
- Windows itself started BSODing with different messages every time;
- Building AOSP under Linux started failing with segmentation faults at random stages;
- Building AOSP even resulted in kernel panics so that the reset button was the only way out.
Now, I don't care that much about the games and windows as a whole, but the AOSP issue is directly connected to my work and it was a problem. Even building with -j1 resulted in segfaults.
I consulted Gemini, I Googled a lot, and at first suspected the usual suspect - RAM. I was running a 2x16GB Corasir Vengeance at 6000 MHz (stupid, I know), so I bought and replaced it with a 2x16GB Kingston Fury kit. Unfortunately, this didn't solve the issues, even though I rat it at base 4800 MHz. As a side node, both kits passed memtest86+, multiple times
I started rolling back BIOS versions until, in v2202, I found an option "Load Intel optimized defaults". This seemed to ease the pain a bit, resulting in less frequent segfaults and kernel panics, but they were still there. You'll agree that this was not a reliable solution.
I then started reading the overclocking forums, because I noticed that the segfaults and kernel panics happened when the build was in full swing (i.e. the cpufreq was at its max 5.6GHz), and I couldn't find an easy option in BIOS (v2801) to simply limit the frequency. I found this great post and started playing with "IA AC Load Line" and "IA VR Voltage Limit" (IAA and IAV for short, respectively).
Setting IAA to 0.2 mOhm and IAV to 1234 mV effectively limited the cpufreq to 5.2 GHz with which AOSP was building with -j20 without hiccups, plus kernel and u-boot. Also, the CPU temperature never exceeds 80 degrees celsius during the whole ~1hour build, even at the end when java and javac processes take over the earth.
I then decided to push the limits a bit, and raised IAA to 0.3 mOhm and increased IAV to 1300 mV to see if it will be still stable. Well, the moment the cpufreq reached 5.6 GHz (1 or 2 minutes into the build), I got a kernel panic and the system hung. I didn't see a temperature spike before that.
Little background: All settings in BIOS (except IAA and IAV) are on Auto, everything is "Intel default" or relative. The BIOS is currently at 2503 but with the stability I've reached with the former IAA and IAV values I simply do not dare update back to 2801 (yes, I know 0x12B fixes a lot, but my Debian's 6.12.12 kernel forces 0x12C early on boot, so it should be even better, right?).
So, now that I've told my story, I have to ask my question, and I hope the specialists reading and posting here will have a proper answer.
Is it possible that running the CPU at [BIOS-default] max speeds and [BIOS-default] max voltages all those 8 months has degraded the CPU to a state that its max frequencies are unstable? And if so, what should I do about it? Whom may it concern and am I eligible to RMA?
Thanks for reading and thanks in advance for any reply.
ROG STRIX Z790-F GAMING WIFI ROG STRIX LC II 360
