I recently solved a problem with [my home PC](https://milohax.gitlab.io/memex/notes/w8cdbg4a7evgm32bzsokb2c/) failing to switch on. I thought I was facing a significant hardware fault, because I've never before faced a problem where the PC _doesn't turn on at all_. The computer is named _Theseus_ because it has had parts replaced for over a decade, and I wondered whether the fault was something serious — I have had a failed Voltage Regulator Module in the past, causing a melted ATX header on the main circuit board, and having to replace main board, PSU, and CPU/RAM, because my existing ones were too dated for any available "motherboards". Or maybe the problem was more benign — perhaps the 13-year-old case has a faulty power switch? But simply bridging the PWR header on the main board did not resolve the power-up failure, so I put off further diagnosis until this last weekend. The final answer was at once simple, but also surprising! I learnt that ***a badly seated NVMe M.2 device can cause a short circuit on the 3.3V rail*** in such a way that the main board/PSU protections cause power delivery to scram, _even before POST_. I honestly don't understand how the NVMe became unseated, since it is held in place by a screw. But I am happy to have found and fixed the fault without having to replace any components. This would have been difficult and expensive in the current computer economics situation. A full diagnostic report follows, for future me. --- ## PC Diagnostic Report - **System:** Self-built desktop - **Date:** March 2026 - **Presenting symptom:** System would not power on _at all_ — no fans, no lights, no drives, no POST ### System Specifications --- | Component | Detail | |--- | ---| | CPU | AMD Ryzen 5 3600 (AM4, no integrated graphics) | | RAM | 2×16GiB TForce Dark DDR4 3200MHz (slots A2 and B2) | | GPU | NVidia GeForce GTX 1660 Ti | | Main board | Gigabyte Aorus B550 Pro | | Boot drive | 120GB SanDisk Extreme SATA SSD (OpenSUSE + GRUB) | | OS drive | 256GB ADATA NVMe M.2 SSD (slot M2A, CPU-attached) | | PSU | Cooler Master MWE Bronze V2 (all black cables) | | Case | Cooler Master Silencio 650 (circa 2013) | **Boot chain:** UEFI → GRUB on SanDisk SSD → OpenSUSE or Windows (chain-loaded via GRUB) --- ## Diagnostic Process ### Phase 1 — PSU verification - Observed brief VRM LED flash upon mains reconnect, confirming +5VSB standby rail was live - Performed multimeter test on [24-pin ATX connector](https://www.smpspowersupply.com/connectors-pinouts.html) with PSU isolated (paperclip bridging PS_ON to COM/GND) - All rails within spec: 12V, 5V, 3.3V confirmed healthy - **Result:** PSU cleared I had to count the pins on the ATX to identify each one, since the cables are not colour-coded on the MWE Bronze V2. ### Phase 2 — CMOS battery - Removed CMOS battery - Measured CR2032 battery: 3.02V - Attempt start without battery: system still does not start - **Result:** Battery healthy, and not a contributing factor ### Phase 3 — Minimum POST configuration Stripped system to bare minimum: - Disconnected GPU PCIe power, and removed GPU circuit board - Removed NVMe M.2 device - Removed one RAM bank (B2), leaving single bank in A2 - Disconnected all SATA data cables **Observations on power-on:** - System powered on — drives spooling up, fans and lights active! - POST debug LEDs: DRAM lit briefly then extinguished (memory init OK), VGA LED remained lit - Speaker beep code: 1 long, 2 short — video card memory failure (expected: no GPU present, Ryzen 5 3600 has no integrated graphics) ### Phase 4 — GPU reseated - Reseated GTX 1660 Ti, connected PCIe power, connected monitor - **Result:** Clean single POST beep, all LEDs extinguished, monitor showed POST output - UEFI reported settings reset (expected after earlier CMOS battery removal) - Only complaint: no boot drive found (expected — NVMe and SATA still disconnected) - **Main board, CPU, GPU, and single RAM bank all cleared** ### Phase 5 — Second RAM bank - Reseated B2 RAM bank - **Result:** System continued to POST cleanly - **Both RAM banks cleared** ### Phase 6 — NVMe reseated into M2A - Visually inspected NVMe and M2A slot — no visible damage - Reseated NVMe into original M2A slot - **Result:** System powered on and POSTed cleanly - UEFI could see the ADATA NVMe but reported no bootloader — expected, as GRUB resides on the SATA SSD ### Phase 7 — SATA drives reconnected - Reconnected all SATA drives - Set UEFI boot drive to SanDisk Extreme SSD/openSUSE - **Result:** GRUB loaded, OpenSUSE booted, Windows chain-load confirmed working --- ## Root Cause **Poorly seated NVMe M.2 device in slot M2A.** The device appeared correctly installed (clicked into connector, secured with a screw) but was likely microscopically misaligned, causing a high-resistance or intermittent connection sufficient to prevent the system from powering on. --- ## Notes for Future Reference - The Ryzen 5 3600 has no integrated graphics — a GPU is required for any display output - UEFI settings are stored in SPI flash, not solely CMOS-backed RAM; the CMOS battery primarily maintains the real-time clock on this board - The B550 Aorus Pro POST debug LEDs sequence: CPU → DRAM → VGA → BOOT; a LED remaining lit indicates that stage failed - M.2 devices can appear correctly seated while being subtly misaligned — if symptoms point to storage, reseat before assuming device failure - After a UEFI settings reset, SVM Mode (AMD-V hardware virtualisation, required for WSL2, VirtualBox, KVM etc.) will be disabled - re-enable SVM (Secure Virtual Machine) at **Settings → Advanced CPU Settings → SVM Mode**