Identifying RAM problems when XMP is active
Identifying RAM problems when XMP is active
I've noticed some problems with my RAM. Turning XMP on seems to trigger issues, and I consistently fail Test 8 with Memtest86. When I run the test without XMP enabled, it completes successfully. My tests with a single stick in slots A2 and B2 showed mixed results—Stick 1 passed in both, while Stick 2 passed in A2 but failed in B2. I repeated the tests multiple times and got the same outcomes. At first, I thought the B2 slot on my motherboard might be faulty, but since Stick 1 worked there, that’s less likely. I’ve also tried the RAM with both Asus ROG STRIX Z790-E GAMING WIFI and the upgraded version, but XMP causes a blue screen and memory errors. When XMP is off, the system has been stable for about two months. System details: motherboard model, case, power supply, CPU, GPU, RAM type, storage. I’m unsure what to do next and am weighing an RMA against possible similar experiences from others. The RAM I own is Corsair Vengeance 32GB (SN CMH32GX5M2E6000C36), and my motherboard’s QVL lists it correctly.
Have you considered refreshing the motherboard BIOS? If you're comfortable doing so safely, I'd check if a small voltage boost could assist. Otherwise, look for guides on this topic, as it's often how I resolved similar problems. Since XMP can boost performance, higher speeds usually mean more power. If your system lacks sufficient power, that might be the cause. If you've tried updating the BIOS before, that worked three times. For DDR4 systems, these settings apply; for DDR5, let it auto-XMP without issues. If not, increasing voltages slightly could help. One of the two methods consistently resolved the issue. The settings I used were: I may note these are for DDR4—people below mentioned I only used DDR5—and for others, just let it auto-XMP without problems. Drm: 1.38V SOC: 1.1V
Two overclock tutorials are available to assist with achieving stable performance at XMP levels. One focuses on boosting DDR5 RAM speeds, while the other explains general overclocking principles. Both emphasize checking the newest BIOS, resetting CMOS, and starting with the manufacturer's listed speed and timing parameters before gradually raising voltages.
I’d likely retest using these sticks with TestMem5 on the 1usmus_v3 preset, since 13th gen tests are generally tougher than MemTest86. This approach could cause both units to fail in slot B2 or have one stick failing in both slots. It might be wise to try another BIOS version or two to check if memory support improves, especially given the Corsair 6000 CL36-44-44 kit’s reputation for inconsistent BIOS updates. If you’re still within the return window and can trade it in for a 6000 CL30 model for under $10, that would be a solid option, though I doubt you’ll do that now after two months of use. I should note these are AMD DDR4 systems, not Intel DDR5—Intel doesn’t support SOC at 1.1V and the voltage levels there tend to clash with many CPUs, making them unreliable at that setting.
I just changed my BOIS to 1301 and plan to test XMP with memtest86 later tonight because it’s warm here. I’ll confirm everything is cool in my room before running the test. I’ll update you once I see if memtest86 works with the new BOIS version. So far, I haven’t adjusted any timing or voltage settings; first I’ll try XMP using all defaults with the updated BOIS. It seems ASUS has finally enabled Intel Default Settings in the BOIS defaults. I spotted some new Performance Preferences that show extreme values—on extreme it reached 86°C during startup, but with Preferences it dropped to 70°C at boot. Also, my coolant was about 37°C after a full day on the machine. I know the i9 13900K gets hot, but 86°C at startup is unusual. My cooler is an Ice Link H170I LCD.
Hey, the test results show Memtest86 passed runs 1 and 2 but failed runs 3 and 4, showing instability with BIOS version 1301. I switched back to version 1201 because 1301 caused unpredictable jumps in CPU multipliers—specifically from 55 to 8 on cores 1-9 and from 43 to 8 on cores 10-24. This behavior was odd and possibly linked to Intel defaults. After reverting, the problem disappeared. I’m planning to stick with the standard 4800 MHz RAM for now and monitor future BIOS updates or consider replacing it if needed. When XMP was active, the performance difference between 4800 MHz and 6000 MHz wasn’t noticeable, especially since my typical gaming is around 60 FPS at 1440p. I’m unsure if 6000 MHz would help much.