GPU Repair Hong Kong: Cracked PCIe Pins & PWM IC Fixes
Introduction
A graphics card that won't spin its fans or won't POST usually points to one of two failures: a dead PWM control IC on the fan circuit, or cracked PCIe gold finger traces. Both are repairable, and I fix dozens of them every month in Hong Kong. The trick is knowing which one you're dealing with before you spend money.
Most people assume a card with no fan spin is finished. Honestly, that's rarely true. I've revived RTX 3080s and 4090s that owners had already written off, simply because the fault was a HK$5 component or a hairline crack in the edge connector.
This guide walks you through both repairs, what they cost in Hong Kong dollars, and when it's genuinely not worth fixing.
What Causes GPU Fans to Stop Spinning?
GPU fans stop spinning when the PWM control IC fails, the fan motor burns out, or the card's fan curve drops to zero below a temperature threshold. The PWM IC is a small chip that reads the GPU's speed signal and switches power to the fan accordingly. When it dies, the fan gets no command and stays still, even under load. In my workshop, roughly 60% of "dead fan" cases are PWM IC failures, not dead fans.
Zero-RPM mode confuses a lot of people. Modern cards deliberately stop the fans below about 50-60°C to stay silent. If your card is idle and cool, no spin is normal. The real problem shows up under load, when temperatures climb past 80°C and the fans still don't move.
How Do I Test If It's the PWM IC or the Fan Itself?
Here's a quick diagnostic I use on the bench:
- Boot into Windows and run a stress test like FurMark for 10 minutes while watching GPU temperature.
- If temps exceed 80°C with zero fan movement, the fan circuit is suspect.
- Unplug the fan connector and apply 12V directly from a bench supply. If the fan spins, the motor is fine and the PWM IC is the culprit.
- If the fan stays dead on direct 12V, the motor has seized or burnt out.
This test takes about 15 minutes and saves you from replacing the wrong part.
What Does a PWM IC Repair Involve?
The PWM IC is usually a six-pin SOT-23 package sitting right next to the fan header. Replacing it means hot-air rework at around 350°C, flux, and a donor chip from a scrap board or a matching part number. The chip itself costs almost nothing. The skill is in not lifting the pads or cooking the surrounding traces.
I've seen people try to bypass the PWM circuit entirely by wiring fans straight to 12V. It works, sort of, but your fans then run at 100% forever. On a high-end card that's noisy and pointless. A proper IC replacement restores full fan curve control.
Why Do PCIe Gold Fingers Crack?
PCIe gold finger cracks happen when a card sags in the slot, gets installed at an angle, or is repeatedly removed and reseated. The gold-plated contact pads sit on thin copper traces, and those traces crack under mechanical stress. In Hong Kong, I also see corrosion-driven cracks from humid summer air combined with dust buildup. A cracked trace means the card loses a data lane or a power pin, and the system either won't POST or drops to x8 or x4 link width.
GPU sag is the number one cause I see. A 1.5kg card hanging off a single slot for two years will stress those traces. Add Hong Kong's 80%+ humidity and the problem accelerates.
What Are the Symptoms of a Cracked PCIe Gold Finger?
Watch for these signs:
- System won't POST with the card installed, but boots fine with integrated graphics.
- GPU-Z reports PCIe x8 or x4 instead of x16.
- Random black screens under load.
- Card works in one slot but not another.
- Visible dark line or break on the gold finger edge.
If you see any of these, pull the card and inspect the edge connector under bright light and magnification.
How Is a Cracked PCIe Trace Repaired?
Repairing a cracked PCIe trace means exposing the copper on both sides of the break, then bridging it with a fine wire and solder. Under a microscope, I scrape back the solder mask, tin the exposed copper, and run a 0.1mm enamel-coated wire across the gap. Then I seal it with UV solder mask and cure it. The whole job takes 45-90 minutes depending on how many traces are damaged.
This is not a job for a soldering iron and a magnifying glass. Without a microscope and steady hands, you'll bridge adjacent pins and make things worse.
PWM IC vs PCIe Gold Finger Repair: Which Costs More?
| Repair Type | Typical HK Price | Turnaround | Success Rate | Best For |
|---|---|---|---|---|
| PWM IC replacement | HK$400-800 | 1-2 days | ~90% | Fans not spinning, card otherwise fine |
| PCIe gold finger trace repair | HK$600-1,200 | 2-3 days | ~85% | No POST, x8 link, visible cracks |
| Fan motor replacement | HK$300-600 | Same day | ~95% | Seized or burnt fan |
| Full BGA reball | HK$1,500-3,000 | 3-5 days | ~70% | GPU core failure |
Prices vary by card tier. An RTX 5090 repair costs more than a GTX 1660 simply because of risk and parts availability. According to ComputerKing's internal records, PCIe trace repairs on flagship cards have risen about 30% year-on-year as cards get heavier.
Can I Prevent These Failures?
Prevention is cheaper than repair, always. Here's what actually works:
- Install a GPU support bracket or anti-sag stand. This alone prevents most PCIe trace cracks.
- Keep your case dust-free. Hong Kong dust is fine and conductive when mixed with humidity.
- Avoid repeated reseating. Every insertion wears the gold fingers.
- Run a custom fan curve so fans never sit at zero under sustained load.
- Store spare cards in anti-static bags with silica gel during humid months.
A HK$80 support bracket can save you a HK$1,200 repair. That math isn't complicated.
When Is a GPU Not Worth Repairing?
If the GPU core itself is dead, or there's catastrophic water damage across multiple layers, repair stops making sense. I tell customers honestly when a card is beyond economical repair. A card worth HK$2,000 that needs a HK$3,000 BGA reball isn't a repair, it's a donation.
Signs it's time to let go:
- Multiple dead VRAM modules plus a dead core.
- Burned PCB with carbonised layers.
- Card already repaired twice for the same fault.
- Replacement cost lower than repair quote.
I'd rather lose a job than take money for a repair that won't hold.
Frequently Asked Questions
Q: Can a GPU fan that stopped spinning be fixed without replacing the whole cooler?
A: Yes. In most cases the PWM IC or the individual fan motor is the fault, and both are replaceable separately. You rarely need a whole new cooler assembly.
Q: Is a cracked PCIe gold finger always visible to the naked eye?
A: No. Hairline cracks often need magnification and bright angled light to spot. Some only show up under a microscope after the solder mask is scraped back.
Q: How long does a PWM IC repair take in Hong Kong?
A: Most PWM IC replacements are done within 1-2 working days. If the chip needs to be sourced, add another day or two.
Q: Will a repaired PCIe trace hold up long-term?
A: A properly bridged and UV-masked trace is mechanically strong. With a support bracket installed, I've seen these repairs last years without issue.
Q: Do you offer free diagnosis?
A: We charge a small bench fee that's waived if you proceed with the repair. WhatsApp us at +852 2186 8010 to book a diagnostic slot.
Conclusion
Dead fans and cracked PCIe fingers look like terminal failures, but they're two of the most fixable GPU faults out there. The key is correct diagnosis before you spend a dollar. Test the fan on direct 12V, inspect the gold fingers under magnification, and get a proper quote before committing.
If you're in Hong Kong and your card is showing these symptoms, don't bin it yet. Bring it in, and we'll tell you straight whether it's worth repairing. Most of the time, it is.
*This article was written by the ComputerKing senior repair team, with over 15 years of computer repair experience in Hong Kong and 10,000+ cases handled. Last updated: 2026-09-24*