Memory issue detected, stability compromised.
Memory issue detected, stability compromised.
After two weeks without issues, I tried Ryzne Master per core curve optimization but the system crashed at the end. The GPU didn’t show anything on the screen. Normally it resets the CMOS, but that didn’t help. Waiting was the only option. Now I’m running at 3200mhz. I remember enabling spread spectrum earlier—maybe I should try it again. It’s possible the auto settings aren’t optimal anymore. My SOC voltage is 1.016. I might increase it to 1.1v. I’ve already tried raising it to 1.36 and disabling spread spectrum, then re-enabled auto global control. If the motherboard supports XMP, I can dial in timings up to XMP, even if the chipset set them lower than recommended (RRD RRDL and TFAW). Settings that were reduced to XMP: command rate 2→1 and tRC...
Based on my observations, 5800X and 5950X don't significantly impact performance when using these CPUs.
1. Consider increasing the SoC voltage to 1.05–1.1V, keeping it below 1.15V each day.
2. Using VDDG for CCD/IOD around 0.95–1.0V can improve stability with mixed capacity kits such as 2x8 + 2x16.
3. Ensure the gear down Mode is activated; this may assist with irregular command rates.
4. When running four sticks, the imc might struggle with XMP; adjusting a slightly looser tRFC or tFAW manually can help stabilize performance.
5. Verify you’re using the most recent BIOS for your b450; updates from Agesa often enhance memory training. Running at 3200MT/s stably is sufficient for a Ryzen 5000 with four DIMMs, but fine-tuning the SoC and VDDG settings may bring XMP closer to its target.
Right now I'm using just two dual-rank sticks, each holding one 16GB drive, giving me a combined storage of 32GB.
Using a 2×16GB dual rank typically works better than with four sticks. Even then, I’d still consider adjusting the SoC to around 1.05–1.1V. Reducing Vddg ccd/iod to 0.95–1.0V can also assist in dual rank scenarios. Ensure gear down mode is active during unusual timings. You might also experiment with a slightly looser tRFC/tFAW manually, as dual rank can be sensitive there. If you can maintain stability at 3200MT/s, that’s already strong. With some manual tuning, you should reach your desired XMP 3600 speed again.
Soc was running at 1.006v. I added a +0.1v offset and configured vdddg to 1v, which caused it to boot at 1.224v. After restarting and setting it to auto, it stabilized around 1.12. Still experienced errors within seconds. I switched on gear down mode in the command rate settings. It appeared to be hidden under misc, so I captured a screenshot for guidance. Later, I noticed it was also disabled in command rate. Gear down mode seems to offer lower latency compared to 2T command rate. After changing it from 1T to gear down in command rate, the p95 test ran smoothly without errors for the last 3-5 minutes. I’m curious about the power usage impact of the +0.1v setting.
Gear Down Mode offers a balanced approach, pushing values beyond typical limits and fixing the command rate around 1T while allowing more flexible internal timing. This makes it simpler for stability compared to enforcing strict 1T, though it may be tougher on the imc. At 1T with gdm, latency stays near ideal levels but stability leans toward 2T. Enabling gdm can help memory issues at 1T run smoothly. The power difference from a +0.1V SoC is almost undetectable in practice, so it’s not a major concern. If your system remains stable in P95 with gdm enabled, it’s a solid choice to maintain.