Darwin Service
GPU Repair Darwin
Component-level graphics card repair: artifacting, dead cards, blown VRMs, fan and thermal service, BGA reflow / reball, VRAM upgrades, mining and AI rig refurbishment, and vBIOS recovery — from our Winnellie workshop.
When a GPU goes dark, throws artifacts, or shows "no signal" after a driver update, most repairers tell you to buy a new card. We don't. We trace the failure to the actual cause — a dead memory chip, a failed VRM phase, a popped MOSFET, a cracked solder ball on the GPU package, dried thermal paste cooking the die — and decide whether repair is economic before quoting.
We also handle work most shops won't touch: VRAM chip replacement, GPU memory upgrades on cards where the layout supports it, BGA reflow and reball with proper preheat and profile control, vBIOS recovery on bricked cards, and refurbishment of ex-mining GPUs for resale, AI rigs, or workstation use.
Service scope
Diagnose at the chip level, repair what's worth repairing, document the rest.
Common GPU Faults
- Artifacting, checkerboarding, lines, or driver-crash on load — usually VRAM or VRM-related
- Card not detected, "no signal", fans spin but no output — GPU power-up sequence trace
- Card detected but stuck at low clocks or thermal-throttling early — thermal / power delivery
- Dead fans, screaming bearings, melted shroud — cooler service and replacement
- Bricked after vBIOS flash — SPI flash recovery and reflash
Component-Level Board Repair
- VRM phase repair — MOSFETs, drivers, inductors, bulk caps
- VRAM chip replacement — GDDR5, GDDR5X, GDDR6, GDDR6X where parts are sourceable
- BGA reflow and reball on GPU package with proper preheat, profile, and inspection
- Trace, pad, and via repair on damaged PCBs
- Failed PCIe interface and display-output circuit repair
VRAM Upgrades
- GPU memory upgrade where the PCB and vBIOS support higher-capacity modules — e.g. select NVIDIA and AMD generations
- Module sourcing, install, and strap modification where required
- vBIOS modification and validation post-upgrade
- Stability and stress-testing across the full memory range before handover
- Honest "this card isn't a candidate" call when the design doesn't support it — we don't waste your money
Mining & AI Rig GPUs
- Ex-mining card refurbishment — cooler tear-down, thermal pad replacement, fan service
- VRAM and core re-paste with appropriate compound for sustained 24/7 thermal loads
- Fault diagnosis across rig-scale batches — "two of twelve are dropping, which two and why"
- vBIOS clean-up and reset where needed
- Refurbishment for resale, gaming reuse, or repurposing into AI / ML workstations
Thermal & Cooler Service
- Full strip-down, thermal paste replacement, and thermal-pad replacement with correct thickness
- Bearing replacement or fan swap on common shroud designs
- Heatsink unwarping and contact-pressure restoration where the cooler has bowed
- Aftermarket cooler fitting where the original is past saving
- Post-service thermal stress test before the card leaves the workshop
What We Don't Pretend To Do
- If your card has a cracked die we'll say so — that's an end-of-life call, not a repair
- If parts (specific VRAM chip, specific GPU package) are unsourceable we tell you up front
- If economic repair is below the replacement cost we recommend replacement — with a sensible spec, not the most expensive option
- If the card's been "fixed" before by someone with a hot-air gun and prayer, we triage what's recoverable from what's been killed
Before you bring it in
If the card was already pulled, don't keep poking at it. Repeated reflows with a hot-air gun and no profile control are the most common way a recoverable card becomes a paperweight. Bring it in and we'll do it properly.
Free initial consultation by phone or at our Winnellie workshop. Clear scope and cost before any chargeable work begins.