Laptops with a dedicated graphics chip carry a second power system alongside the processor's: a large regulator for the GPU core, separate rails for the graphics memory and a handful of smaller supplies. On most modern laptops this whole section can be switched off when the integrated graphics are enough. That makes GPU power faults distinctive — the laptop may work perfectly until something asks for the dedicated GPU, or it may refuse to start at all. This guide explains the circuit and how to approach it.
The rails a dedicated GPU needs
| Rail | Typical names (vary) | What it feeds | Typical regulator |
|---|---|---|---|
| GPU core | NVVDD, VDDC, GPU_CORE | The graphics processor itself | Multi-phase buck |
| Memory supply | FBVDDQ, MVDD, VRAM rail | Graphics memory and the GPU's memory interface | Single- or dual-phase buck |
| PCIe / I/O | 1V0, 1V8, PEX rails | GPU interfaces and auxiliary logic | Small buck or LDO |
| Memory termination and reference | VTT, VREF | Some memory types | Small tracking regulator or divider |
Names depend on the GPU vendor and the board designer; the GPU's power pages in the schematic list them. The core rail is described under gpu core rail and the memory under vram.
Power gating on switchable graphics
On laptops with switchable graphics, the dedicated GPU is powered only when needed. A control signal — often from the chipset or a GPIO, named something like DGPU_PWR_EN — switches the GPU's rails on, usually in a fixed order:
- Auxiliary and I/O rails.
- Memory rails.
- Core rail.
- Power-good signals back to the system, then the GPU's reset is released.
When the GPU is not in use, the sequence runs in reverse and the rails go to zero. So measuring zero on the GPU core with the laptop running is not automatically a fault — the GPU may simply be off. To check the rails, make the system use the dedicated GPU (start a demanding program, or use a BIOS setting that forces it) or measure during the start-up phase, when many laptops briefly power the GPU.
Laptops whose display is wired only to the dedicated GPU keep it powered whenever the screen is on. On those, a GPU power fault means no display at all.
How the GPU core voltage is controlled
Like the processor, the GPU sets its own core voltage. Depending on the design, it sends a PWM signal or a serial command to its controller, and the voltage changes with load. The controller has an enable input and a power-good output, and the core regulator runs from the system rail through its own power stages and inductors.
Common GPU power faults
- Shorted core rail — one of the most common GPU faults. Causes include a cracked ceramic capacitor under or around the GPU, a failed power stage, or the GPU itself.
- Missing memory rail — the GPU starts but fails to initialise; the system may hang or show no image.
- Enable never arrives — a firmware, chipset or control-path fault keeps the whole section off.
- Rails present, GPU not detected — the GPU itself, its reset, its clock or its PCIe link.
Diagnosing a shorted GPU core
With all power removed, measure the core rail to ground. As with the processor core, a healthy rail can read a low resistance because the GPU is a large load, so compare with a reference. A reading of a fraction of an ohm where the reference is much higher, or near zero in diode mode, is a short.
To locate it:
- Lift each phase inductor to separate the power stages from the rail. A shorted power stage stays shorted on its own side.
- If the short is on the rail side, inject a low voltage — well below the normal core voltage — through a current-limited supply.
- Look for the hot spot with a thermal camera, freeze spray or rosin, as described in Finding Shorts with Freeze Spray, Thermal Camera and Rosin.
If the GPU package itself warms evenly and nothing else does, the GPU is likely shorted internally. Replacing a GPU is BGA work and not always worthwhile; on some designs a laptop can run with the dedicated GPU disabled in firmware, which is a compromise to discuss with the owner.
Diagnosing a missing rail
When the GPU should be on and a rail is missing:
- Check the rail's regulator: input, enable, power-good, switch node.
- Check whether the enable signal arrives at all — trace it back to its source.
- Check the order: a missing auxiliary rail stops the memory and core rails from being enabled.
On laptops where the screen depends on the dedicated GPU, this diagnosis overlaps with the general no-display routine in Laptop Powers On but No Display: Where to Start.
Graphics memory rail checks
The graphics memory rail feeds both the memory chips and the GPU's memory interface. When it is missing or wrong, the GPU may power up but fail to initialise, producing a hang, a black screen or artefacts. Measure it with the GPU active, compare it with the value the schematic gives for the memory type, and check its regulator's feedback network if the value is off. With power removed, compare the resistance to ground with a reference board; a shorted memory chip or capacitor on this rail behaves like any other short.
Heat and the GPU section
GPU power stages, inductors and the GPU itself run hot under load. Cracked solder joints from repeated heating are a long-term failure mode, especially on older designs. A GPU section that works cold and fails when warm points at joints, power stages near their limit, or a thermal pad that no longer contacts the heatsink. Check the cooling before replacing parts.
Example: works on battery saver, freezes in games
A laptop runs normally for office work but freezes as soon as a game starts. The dedicated GPU powers up when the game launches; at that moment, the core rail rises, then drops back to zero within a second, and the system hangs. With power removed, the core rail reads normally, but injecting a low voltage at limited current draws more than expected, and a thermal camera shows one small capacitor next to the GPU warming. It fails short only under voltage. Replacing it lets the core rail stay up, and games run normally.
FAQ
Why is my GPU core rail at zero with the laptop on?
On switchable-graphics laptops the GPU is powered only when needed. Make the system use it, then measure.
Can a laptop run without its dedicated GPU?
Many can, if the display is connected to the integrated graphics and the firmware allows the GPU to be disabled. Some cannot.
Is a shorted GPU core always a dead GPU?
No. Capacitors and power stages fail often. Isolate them before condemning the GPU.
Do GPU memory rails differ from system memory rails?
They serve a different memory type and are controlled by the GPU section, but they are diagnosed the same way.
What tools do I need for GPU power faults?
A multimeter, a current-limited bench supply and a way to find hot spots — thermal camera, freeze spray or rosin. The gpu itself needs BGA equipment to replace.
Bottom line
A dedicated GPU has its own core, memory and auxiliary rails, often switched off when not in use. Make sure the GPU should be powered before calling a rail missing, check enables and sequence, and treat a shorted core rail like a processor core short: isolate the phases, inject low voltage and find the hot spot. Capacitors and power stages fail more often than the GPU itself.