AI-rewritten: This is a summary of an article from Tom’s Hardware, rewritten by AI (Qwen, running locally) to make it easier to read. The facts come from the original article – read it for the full story.

---

Tony from NorthWestRepair has released a video detailing the repair of two high-capacity DDR5 memory sticks valued at $16,000. These are ECC server modules, one containing 128GB and the other 256GB of DRAM. Tony noted that initial physical inspections did not reveal the cause of the problems, so he relied on electrical testing instead. He used flux and hot air to reflow the solder on the first module, a process simpler than reballing where memory ICs are removed and resoldered. After this repair, the first DIMM began functioning correctly.

The second module presented more severe issues, with Tony indicating it may have a faulty Power Management Integrated Circuit (PMIC) chip. A PMIC regulates power to the memory chips. When powered on, the system showed no signs of life, and thermal camera checks confirmed the module did not warm up. To address this, Tony repaired non-DRAM components like the PMIC and SPD EEPROM before attempting to reflow other chips while the board was warm. He also reballed some memory ICs after inspection and identified a short circuit using voltage injection and thermal monitoring. This short was fixed by replacing a capacitor and repairing the damaged edge of the printed circuit board (PCB). A new PMIC chip ultimately allowed the second DIMM to power on and train successfully.

Tony highlighted that spending hundreds of dollars on labor to save thousands in components is now sensible due to the high demand for memory driven by AI applications. Previously, replacing a damaged DIMM was often cheaper than repairing it, but the cost difference has narrowed significantly. This shift suggests that repairing memory modules could become a viable business opportunity, preventing valuable hardware from becoming electronic waste.