My roughly four-year-old PC has started randomly failing to POST. It has an MSI Z790 Tomahawk motherboard, an Intel Core i9-13900K, four 32GB Kingston Fury RGB DDR4 modules, and an MSI RTX 4090.
When the problem occurs, the system powers on but never displays anything. The memory lighting stays on its default rainbow pattern instead of switching to the Windows profile. The motherboard debug LEDs briefly progress through CPU, DRAM, and VGA, then turn off; the BOOT light never comes on. The computer shuts down after a few seconds and repeatedly tries to start again until I switch off the power.
Previously, clearing the CMOS and removing three memory sticks would eventually let it POST. I could then reinstall the other modules and use the computer normally, including under heavy loads. That workaround has now stopped being reliable, although after several hours of trying with one stick, it unexpectedly started again and currently works with all four installed.
The issue happens after both sleep/resume and a normal shutdown, so it does not seem limited to waking from sleep. The CPU and motherboard look physically undamaged, and I do not have spare components for testing. I would like to determine whether the problem is faulty RAM, unstable memory settings, the CPU's memory controller, the motherboard, the power supply, or something else before buying expensive replacement parts. What diagnostic steps should I try?
4 Answers
Clear CMOS properly, load BIOS defaults, and leave XMP disabled while troubleshooting. Four 32GB DIMMs put much more stress on the memory controller than two modules, and a configuration that worked for years can become unreliable after a BIOS change, aging, or marginal training. Once the system is stable at default speed, update the motherboard BIOS using its built-in flash utility and only then experiment with a lower memory speed or XMP.
Do a complete power reset when it locks up: switch off the PSU, unplug it, hold the case power button for about 10 seconds, then reconnect power and try again. If that does not help, test with the minimum hardware possible—CPU, cooler, one RAM stick, and graphics output—and disconnect drives and USB devices. Because the CPU and VGA debug stages complete, RAM training or the motherboard/CPU memory path is more suspicious than Windows or a BOOT-device problem, but a failing PSU or motherboard still cannot be ruled out without substitution testing.
Reseat the RAM, graphics card, motherboard power connectors, and CPU power connector. Inspect the CPU socket carefully for bent contacts, since the memory channels connect through the socket. Also check that the cooler is not applying unusual pressure. A marginal socket contact can cause intermittent memory-channel or POST failures even when Windows is perfectly stable once it finally starts.
Start by running MemTest86 from a bootable USB. Test each memory module individually in the motherboard’s recommended slot, then test a matched pair, and finally all four together. Let the test complete several passes; a single error is significant. If individual sticks pass but four sticks fail, the problem may be memory training, an overly aggressive XMP setting, the CPU’s memory controller, or the board rather than a defective module.
So MemTest86 runs outside Windows from a USB drive? I’ll test the modules separately and with XMP disabled.
Even if the RAM is rated for its advertised speed, four modules may not reliably reach that speed. Running two sticks or a lower speed can help identify whether this is a memory-training issue.