A friend's PC, built about 1–2 years ago, has started turning on and off repeatedly without reaching the BIOS. The system uses a Ryzen 5 7600X, MSI B650 Gaming Plus WiFi motherboard, ASRock Radeon RX 5700 XT, four 16GB DDR5-5200 modules, a Kingston NVMe drive, a 2TB SATA SSD, and a Corsair RM650 power supply.
The machine power-cycles even with the graphics card removed and the monitor connected to the motherboard's video output. I've reseated the storage, RAM, graphics card, power cables, and other connections; cleared CMOS; inspected and cleaned the connectors; and even removed and reseated the CPU. The motherboard's diagnostic LEDs have shown white for the graphics card, yellow for DRAM, red for the CPU, and occasionally red plus yellow during what may be memory training. A known-working replacement graphics card did not resolve the issue.
I'm beginning to suspect either the motherboard, the memory configuration, or possibly the CPU's memory controller. The system has four DDR5 sticks, which I now realize can be more difficult for AM5 platforms to run reliably. What would be the best next steps to isolate the failed component without access to another DDR5 system for testing?
5 Answers
The diagnostic LEDs are useful, but the light shown can change depending on how far the board gets through POST. A DRAM failure can also produce a CPU light because the memory controller is built into the CPU, while the graphics LED may appear only after the earlier stages succeed. Since a different graphics card also failed and the system cannot POST with the GPU removed, the graphics card is unlikely to be the primary cause. The motherboard’s memory circuitry or the CPU’s memory controller are more plausible.
If the minimum-configuration test fails with every RAM stick, I’d focus on the motherboard next, followed by the CPU. Reseat both ends of the power connections and, if possible, test with another known-good power supply, but the fact that it also loops without the graphics card makes the GPU’s power demand less likely to be responsible. A POST diagnostic card may provide additional codes, although the motherboard’s own LEDs and systematic component testing will usually be more useful here.
Test the memory one stick at a time in slot A2, which is normally the recommended slot for a single DIMM. Try each stick individually and give the system several minutes after clearing CMOS, because AM5 memory training can leave the screen blank for a while. If one stick works, enter the BIOS, disable EXPO or any memory overclock, and run the memory at a conservative default speed before adding anything else. Four DDR5 modules put more strain on the memory controller, so the system may be failing during training rather than having a graphics problem.
A single failed DIMM can cause repeated restarts, so testing every module separately is important. If none works alone, test a known-good basic DDR5 module if one can be borrowed; that helps separate a memory issue from a board or CPU issue.
Start with a bare-minimum POST test outside the case. Put the motherboard on its box and connect only the CPU and cooler, 24-pin power, CPU EPS power, one RAM stick, and the motherboard’s power switch connection. Remove both storage drives and the graphics card, then use the 7600X’s integrated graphics. This rules out a case short and unnecessary components. Also inspect the AM5 socket carefully for bent pins, especially since the CPU was removed.
The MSI board supports BIOS Flashback, so reflashing the latest stable BIOS is worth trying even if the computer cannot reach the BIOS screen. Follow the board manual exactly: use a FAT32 USB drive, the correct BIOS file name, the designated Flash BIOS USB port, and only the required power connections. Afterward, repeat the one-stick memory test. If the flash process fails immediately, try another small USB drive because some boards are picky about them.
I also suspect the motherboard at this point. I’ll try the one-stick test, BIOS Flashback, and an out-of-case minimum build before replacing anything.