How to check the PMIC on a DDR5 module: shorts, voltages in the PC and a bench start
This is part of the Corsair Vengeance DDR5 module repair — main article. Before any soldering you need to know two things: whether the power management chip (PMIC) is good and whether the memory chips are good. Here is how we checked the PMIC and what we got. On our dead module it is good.
What the PMIC is on a DDR5 module
On DDR4 the memory voltages came from the motherboard. DDR5 moved this onto the module itself: the motherboard only supplies 5 V input (VIN_BULK), and the PMIC on the module makes every working voltage from it. So a DDR5 module can “refuse to start” because of its own power supply — something DDR4 never did.
| Rail | Usual voltage | What it powers |
|---|---|---|
| VIN_BULK (input) | 5 V | the PMIC itself; comes from the motherboard |
| VDD | 1.1 V (higher with XMP, 1.35 V on ours) | memory chip core |
| VDDQ | 1.1 V (higher with XMP, 1.35 V on ours) | chip data outputs |
| VPP | 1.8 V | chip internal voltage |
| LDO 1.8 V | 1.8 V | SPD hub and PMIC logic |
| LDO 1.0 V | 1.0 V | auxiliary rail |
Step 1. Unpowered: rails for shorts
The module is out of the PC. Multimeter on resistance, black probe on ground (pin 6 or 150), red probe in turn on the output capacitors of each rail next to the PMIC and on the VIN_BULK pin. The reading rises at first — the capacitors are charging; write down the settled value and compare with a working module.
| What you see | What it means |
|---|---|
| under 5 Ω | a short: failed capacitor, PMIC or memory chip; do not put the module into a PC |
| much lower than on the working module | leakage: a degraded capacitor or chip |
| kilohms or more, rising over time | normal — the capacitors are charging |
On ours the 1.8 V rail against ground read 15 kΩ and kept rising to megohms — no shorts. A table for these readings is in the tool of the main article — the “Short-circuit check” block comes first there.
Step 2. In the PC: PMIC output voltages
If there are no shorts, the module can go into the PC, and during a boot attempt you measure the voltages on the PMIC output capacitors. Use thin needle probes: an ordinary probe slipping between neighbouring capacitors makes a short by itself.
| Mode | Two 1.8 V rails | VDD | VDDQ |
|---|---|---|---|
| JEDEC (XMP off) | 1.8 V and 1.8 V | 1.1 V | 1.1 V |
| XMP on | 1.8 V and 1.8 V | 1.35 V | 1.35 V |
These are readings from our dead module — every rail is fine, and VDD and VDDQ follow the BIOS settings (XMP). So the PMIC is not only alive, it also takes commands from the motherboard.
Step 3. On the bench, without a PC
The PMIC can be started without a motherboard — that is how we did it when reading the SPD with an ESP32. You need two supplies, 5 V and 3.3 V, with current limiting if your supply has it (start with a few hundred milliamps: with the memory idle the current is small).
| Pin | Signal | What to apply |
|---|---|---|
| VIN_BULK | PMIC input | 5 V |
| 151 | PWR_EN — PMIC enable | 3.3 V |
| 148 | HSA — module address | to ground (module acts as “slot 0”) |
| 6 or 150 | ground | negative of both supplies |
Where exactly the 5 V goes. According to the pinout datasheet we use, VIN_BULK is pins 1, 145 and 146. Our log shows 5 V applied to pin 3 — and the PMIC started on both modules, but pin 3 is not confirmed in that same pinout. Use the pins from your own module’s datasheet.
| What we measured | Result (both modules) |
|---|---|
| 5 V input | 5.0 V |
| PWR_EN | 3.3 V |
| LDO 1.8 V on the capacitor next to the SPD hub | 1.8 V |
| VDDQ on the capacitor next to the chips | 1.1 V |
| PWR_GOOD (pin 147) | not measured |
PWR_GOOD is an open-drain output: it does not drive a voltage by itself. To see it, pull pin 147 up to 3.3 V through a resistor: high means the PMIC considers every rail fine, low means it sees a fault. We did not do this.
One more bench check is whether the PMIC answers on the I²C bus (address 0x48 for a module with HSA on ground). The ESP32 firmware from the SPD reading article has the p command for this.
What these readings do not prove
- Behaviour under load. A voltage can be perfect with no load and sag during memory training. You only fully see that with an oscilloscope or by swapping the PMIC.
- Converter noise. A multimeter shows the average; it does not see ripple.
- The final test is swapping the PMIC between the working and the dead module: if the fault “moves”, the PMIC is to blame. We did not do this — the readings and the XMP behaviour were enough to look further.
Summary for our module
| Check | Result |
|---|---|
| 1.8 V rail ↔ ground, unpowered | 15 kΩ, rising — no shorts |
| voltages in the PC, JEDEC | all fine |
| voltages in the PC, XMP | VDD and VDDQ rose to 1.35 V — the PMIC follows the BIOS |
| bench start | starts, rails fine (both modules) |
| conclusion | the PMIC is good — the cause is elsewhere |
The next thing to know before a repair is whether the memory chips themselves are good: how we checked them.
Article changelog
| Date | What changed |
|---|---|
| 2026-11-02 | First version: rails for shorts, voltages in the PC (JEDEC and XMP), bench start of the PMIC, limits of these checks. |