My four-year-old system has started randomly failing to POST. It uses an MSI Z790 Tomahawk motherboard, an Intel i9-13900K, four 32GB Kingston Fury RGB DDR4 modules, and an MSI RTX 4090.
When the problem occurs, the PC powers on but never displays anything. The memory lighting stays on its default rainbow pattern instead of switching to the Windows profile, and the system repeatedly shuts itself off and tries to start again. The motherboard's diagnostic LEDs briefly progress through CPU, RAM, and VGA, but the BOOT light never turns on. The CPU light may flash red again just before shutdown.
Previously, removing three memory sticks and clearing the CMOS would eventually let the machine start. I could then reinstall all four modules and use the PC normally, including under heavy load. That workaround has now stopped being reliable, although after several hours of trying with one stick installed, it unexpectedly booted and worked again with all four sticks installed.
The failure happens after both sleep and a complete shutdown, so it does not appear limited to waking from suspension. I have no spare hardware for testing and cannot afford to replace parts without narrowing down the cause. Could this be faulty RAM, memory training with four DIMMs, the CPU's memory controller, the motherboard, or the power supply? What tests should I perform first?
4 Answers
Start by testing the memory properly with a bootable memory tester such as MemTest86 or MemTest86+. Run several complete passes, ideally with one module installed at a time and then with the full set. Test each stick in the recommended slot from the motherboard manual. A system can be stable in Windows and still fail during POST or memory training, so the absence of blue screens does not rule out a memory problem.
Four large DIMMs put more strain on the CPU’s integrated memory controller than two modules. Clear CMOS, update the motherboard BIOS if possible, and test with default settings only—no XMP, memory overclock, or aggressive timings. Begin with one stick in the board’s recommended slot, then add the modules one at a time. If two sticks work reliably but four do not, the issue may be memory training, the memory controller, or the board rather than one obviously dead DIMM.
The diagnostic LEDs are useful: if they stop around CPU or RAM and the machine never reaches the boot stage, focus on power delivery, BIOS settings, the DIMMs, and the CPU socket or memory controller before blaming Windows. Reseat the RAM and power connectors, inspect the socket for bent contacts, and check that the 24-pin motherboard and CPU EPS cables are firmly connected. If the problem continues at completely stock settings after individual memory testing, the motherboard, CPU, or PSU will need to be isolated with known-good hardware or tested by a repair shop.
Fully discharge the system when it gets stuck: hold the power button until it shuts off, switch off or unplug the PSU, hold the case power button for about ten seconds, then restore power and try again. Also disable fast boot and avoid sleep or hibernation temporarily. Since the same behavior occurs after a cold shutdown, this probably is not the whole explanation, but it can eliminate a power-state issue.
That would explain why removing most of the memory sometimes helps, but I’ll need to test each module and slot separately before assuming the RAM is bad.