PSA: XMP and EXPO are basically a manufacturer-approved RAM overclocking. They are not guaranteed to work. I got XMP (not EXPO) RAM because it was cheaper, and then I had to wait for 7950X3D for 1.5 months. So once I put it all together and RAM refused to work at XMP settings, I could not return it, so I had to make it work, as it’s expensive to just discard. It wouldn’t even boot without BSOD. Logically, when memory doesn’t work you up the voltage until it does, and deal with the heat. But raising the voltage didn’t work at all, in my case. It’s a G.SKILL S5 6000-CL32-38-38-96 but it didn’t work at 6000 at XMP settings (1.4v) even with higher voltage. I was desperate and tried everything. But it seemed to be getting better with lower voltage! So after some research, I kept lowering the primary voltages, and raising secondary voltages, until I found a sweet spot. To my surprise, I got it rock-stable and passing all MemTest86 and TestMem5 tests at a whopping 1.3v, which is very low. VDD_SOC, VDDG etc. had to be raised. As a result it runs great, and very cool, and even at CL30 (38-38-96) and 2033Mhz Infinity Fabric (supposed to be better than default 2000Mhz), which was my “best case” scenario. I could not tighten secondary timings without triggering errors in stress-tests, but it runs CL30 on CL32 timings. So maybe everyone who has problems getting their memory to run at XMP or EXPO, if just raising the primary voltage doesn’t help - try that approach.
1 Like