diagnostics

Gpu Card Diagnostics

GPU Hardware Diagnostic & Repair

This guide follows a structured sequence for diagnosing graphics card failures, from visual inspection to advanced power injection and logic signal analysis.

Section 1: GPU Diagnostics

Role: General hardware assessment and client intake.

  • GPU Make & Model Info: __________
  • Visual Inspection & Initial Condition:
    • Physical damage (cracks, missing components)
    • Liquid damage / Corrosion
    • Thermal pad/paste degradation
    • Blown/Discolored MOSFETs or Caps
  • Notes: _________________

Section 2: Circuit & Power Testing (Passive)

Testing for faults with the card disconnected from power (Multimeter in Resistance mode).

2.1 Resistance Measuring Reference

| Rail | Expected Resistance | Notes | | :— | :— | :— | | GPU VRM (VCore) | 0.2Ω - 3.0Ω | Low resistance is normal for high-current cores. | | Mem VRM | 20Ω - 120Ω | Varies by VRAM type (GDDR6 is lower than GDDR5). | | PEX Rail | 10Ω - 80Ω | Power for PCIe communication logic. | | 1.8V Rail | 500Ω+ | Logic standby rail; often a 1117 regulator or small buck. | | 5V Rail | kΩ Range | Internal regulator rail. | | 3.3V Rail (PCIe) | kΩ Range | Power from the motherboard slot (Pin A9/A10). | | 12V Rail (Slot/Ext) | kΩ Range | If <100Ω, there is a major short on the input. |

Section 3: In-System Voltage Measuring (Active)

GPU installed in a test bench and powered on. Measure at the inductors/coils.

3.1 The GPU Power Sequence (The “Bucket Brigade”)

Rails must turn on in this specific order. If one fails, the sequence stops.

  1. 12V (Main Input) $\rightarrow$
  2. 5V Rail $\rightarrow$
  3. 1.8V Logic $\rightarrow$ 4. VCore $\rightarrow$ 5. VMem $\rightarrow$ 6. PEX

3.2 Target Voltages

  • GPU Core (VRM): ~0.8V - 1.1V
  • Memory VRM: ~1.35V - 1.55V
  • PEX Rail: ~1.0V
  • 3.3V Rail: 3.3V
  • 5V Rail: 5.0V

Section 4: Logic & Communication Diagnostics

Use these checks if voltages are present but there is No Display or Code 43.

Signal/Component Role Diagnostic Check
Crystal Oscillator The “Heartbeat” Use Scope; verify 27MHz sine wave.
BIOS Chip (SPI) Instruction Manual Check Pin 1 (CS) for data “chatter” at startup.
Differential Pairs PCIe Data Lanes Diode mode on PCIe fingers; must match within 5%.
Reset (PERST#) Wake-up call Pin A11 should jump from 0V to 3.3V at boot.

Section 5: Fault Finding & Power Injection

Used when a short circuit (~0Ω) is detected on a main rail.

5.1 Injection Safety Settings

| Target Circuit | Starting Voltage | Max Safe Current | | :— | :— | :— | | Short Tracing | 0.5V – 1.0V | 0.5A – 1.5A | | Logic (3.3V/5V) | 1.0V – 2.0V | 0.5A – 1.0A | | Power Rails (12V) | 2.0V – 3.0V | 1.0A – 2.0A |

5.2 Component Heat Observation

  • Injected Power Into: _______________
  • Current Draw: [ ] Steady [ ] Spiking [ ] OCP Tripped
  • Heat Detected: [ ] MOSFET [ ] Capacitor [ ] PWM IC [ ] GPU Die (Dead Core)

Section 6: Diagnostics Outcome & Notes

6.1 VRAM & Artifacting

  • Software Test: Use NVIDIA MATS / AMD T-Script to identify failing modules.
  • Push Test: Gently press VRAM chips during stress test to check for cold solder joints.

6.2 Results Summary

  • Repairable: Blown external component (Cap/FET/Fuse).
  • Inconclusive: Potential BIOS corruption or failing Crystal.
  • Beyond Economic Repair (BER): 12V leaked into Core; internal silicon short.

Detailed Diagnostics Notes: __________________________

Pro Tip: The 12V-to-Core Death Check

Measure resistance between the 12V Input and the VCore Inductor. If you see 0Ω, a high-side MOSFET has failed “closed,” sending 12V into the 1V Core. The GPU is almost certainly dead. * [ ] Thermal pad/paste degradation * [ ] Blown/Discolored MOSFETs or Caps