Error with Prime95, crashing during DOCP execution
Error with Prime95, crashing during DOCP execution
I’m dealing with some persistent problems that have lingered for over a week. After reviewing numerous forum discussions, performing several tests, I finally reached out for assistance. The situation involves a PC built roughly two weeks prior, upgraded from an older system. All components were new at stock settings—CBR23 was run once, then I switched to PBO2 in BIOS with negative curve offset and DOCP enabled. I also remembered CTR2.1RC5 but encountered log errors and couldn’t display CPU/MB details (LLC set to L3 for CTR). I tried Ryzen Master’s curve optimizer instead. During stress tests with P95 blended, the system ran smoothly for more than an hour. However, I started experiencing random crashes, which were detailed in the WHEA logs (Processor Core, Machine Check Exception, Cache Hierarchy Error) and mostly pointed to APIC IDs 6 through 11. I gradually reduced the curve offset from -30 to around -5 on core 3, then continued stress tests with P95 for hours without issues. After a week away, I returned and noticed two or three crashes during idle time. I reset the BIOS to Optimized Defaults (except disabling Armoury Crate, changing Boot Option to Windows UEFI, and enabling SVM). Still, spontaneous reboots persisted. I became uncertain whether the issue lay in the CPU or RAM, as errors appeared as CPU faults in WHEA. I ran an AUTO RAM profile under P95 with Large FFTs for 15 minutes without problems. Switching to DOCP introduced errors within five minutes (Fatal Error: Rounding). I attempted to raise the RAM voltage to 1.375 for DOCP, hoping for better stability, but this caused the PC to cycle repeatedly. Testing two or three sticks at a time sometimes led to longer delays or more than six errors in the first two minutes. Eventually, I replaced all four sticks and started a P95 test, which triggered a BSOD. After clearing CMOS and reverting to the AUTO profile, a worker error appeared in under eight minutes. I suspected the crashes were random, though I now see patterns—sometimes 20s, sometimes 8–10 errors, occasionally up to 32 or 31 per minute. The inconsistency made me lean toward RAM as the likely culprit. I also realized I might have used non-standard RAM not listed in QVL (SN850, 870EVO/870QVO SSDs, HX1000 PSU), assuming they’d be fine since the list was outdated. It’s worth noting that errors during P95 were sporadic and sometimes affected many workers before I could even open HWiNFO. I tested a Small FFT to check for CPU errors and found none for half an hour, though Large FFTs showed issues within five minutes. Overall, I’m still unclear on the root cause. Key Takeaways: - Random PBO2 curve opt issues resolved with stress testing - DOCP introduced quicker error visibility - No errors with Small FFT, but P95 errors were frequent - RAM compatibility and voltage settings mattered - Cooling and thermal paste application need review - BIOS version and configuration play a role - Patterns are inconsistent, making diagnosis harder.
Recently many users encountered issues with their Ram that wasn't on the approved list. Remember this point. I had a Gigabyte Mobo with unstable RAM and needed to adjust one BIOS setting. ---> XMP High Frequency Support ---- > set it to level 2 This brings up another concern: is your BIOS current? Have you attempted to configure your DRAM voltage to 1.4? Updated August 21, 2022 by SamClan additional feedback
Attempted disabling D.O.C.P. and setting timings/settings manually for RAM. The latest BIOS from April 29 that updates AGESA is available.
You received the poorest RAM setup, which explains why it crashes immediately when Corsair uses aggressive binning and ASUS models have terrible RAM layouts. The B550 + X570 combo struggles with only 4533 MB, and adding four sticks only worsens the situation due to poor memory topology. It’s running dual-rank but is very slow, so the motherboard isn’t worth saving right now. You might want to remove two sticks and test—if it works, discard that board since the issue is likely the RAM. If not, just replace the RAM, as Corsair seems to have consistently bad binning, and other RAM problems usually come from ASUS boards. Just pick the cheapest option on the QVL list and move forward.
Using two kits together may lead to complications. Purchasing separate sets can result in two distinct memory chips per stick (RAM makers aren't great unless you buy from an OEM, which gives whatever fits the speed category). A 3600MHz CL16 can support various chips, some compatible, others not. Corsair RAM provides version numbers that reveal chip details—like v4.31, v3.41, v5.49, etc. These codes indicate the manufacturer and density (2^B). The density tells you if it's dual or single rank: 8GB with v4.31 is dual, while 8GB with v3.41 would be single. The specific die also matters; some follow a 0-1-2-3 pattern, others deviate (e.g., v4.49 is Samsung M die). This helps identify whether your sticks are mixed or high-speed. Knowing this clarifies performance and compatibility.
Determining if you have mixed memory and quad rank support is straightforward. Different version numbers mean mixed chips; identical ones suggest uniformity. For quad rank, a B number of 3 indicates quad rank, 4 suggests dual. However, with the Ryzen controllers handling many ranks, 3600MHz often fails to deliver consistent results. Adjusting the voltage to 1.15V can help, but you might need to lower the speed to 3533MHz or 3466MHz. Changing the BIOS revision is another option—especially on AMD boards where memory support fluctuates with updates. On my X570 Master, BIOS F35 is faulty for memory, while F32 and F36e work fine. Switching revisions could resolve the issue.
The latest BIOS on the ASUS site updates AGESA to ComboV2PI 1.2.0.7. I’m not sure if my motherboard supports this—this seems like a Gigabyte-only feature. I haven’t tried it yet, but I’m planning to. Honestly, I’m aiming for a straightforward setup that works right away so I can get back to work before the month ends. I know computer components can be tricky, but I’d have liked manufacturers to deliver their products as advertised out of the box. Still, it feels like part of the blame lies with me for not following the recommended configuration.
Thank you for the assistance. I pushed my Vsoc to 1.15V without any problems like random crashes or P95 RAM errors, whether using Auto or DOCP.