GPU Fan Not Spinning? PWM IC & PCIe Fix Guide HK
Introduction
A GPU fan that stops spinning is one of the most common failures I see in my Hong Kong workshop. The card still posts, Windows loads, but within minutes the temperatures climb past 85°C and the system either throttles hard or shuts down. Nine times out of ten, the culprit isn't the fan itself — it's the PWM control IC on the board that has failed.
I've been repairing graphics cards in Hong Kong for over 15 years, and I've handled more than 10,000 repair cases. In this guide, I'll walk you through exactly what causes fan failures, why PCIe gold finger damage is on the rise, and what a realistic repair costs in HKD.
Why Does My GPU Fan Stop Spinning?
A GPU fan stops spinning because the PWM (Pulse Width Modulation) control IC that regulates fan speed has failed, or because the fan motor itself has burned out. In my workshop, roughly 70% of "fan not spinning" cases trace back to the PWM IC, not the fan. The IC is a tiny chip near the fan header that tells the fan how fast to spin based on GPU temperature. When it dies, the fan either stops completely or runs at 100% constantly.
Here's the thing most people don't realise: modern GPUs use zero-RPM mode. Below about 50-60°C, the fans are designed to stay off. So before you panic, check whether your card is simply idling. Run a stress test like FurMark or Heaven Benchmark and watch the fans. If they still don't spin past 70°C, you have a real problem.
How Do I Know If It's the PWM IC or the Fan?
You can tell whether it's the PWM IC or the fan by swapping the fan onto a known-good header, or by measuring voltage at the fan connector with a multimeter. If the fan spins when powered directly from a 12V source, the fan is fine and the PWM IC is the problem. If it doesn't spin even with direct power, the fan motor is dead.
In my experience, the test takes about five minutes. Set your multimeter to DC voltage, probe the fan connector while the GPU is under load, and look for 12V on the power pin and a fluctuating 0-5V signal on the PWM pin. No signal means the IC is gone.
A dead fan is a straightforward replacement — usually HK$150 to HK$400 for a replacement fan assembly depending on the model. A dead PWM IC requires microsoldering and typically runs HK$600 to HK$1,200, because we need to source the specific IC and reflow it without damaging nearby components.
| Fault | Typical Symptom | Repair Cost (HKD) | Difficulty |
|---|---|---|---|
| Fan motor dead | No spin at any temp | $150 – $400 | Easy |
| PWM IC failed | No spin under load | $600 – $1,200 | Hard |
| Fan header corroded | Intermittent spin | $300 – $600 | Medium |
| VBIOS fan curve corrupted | Erratic spin | $200 – $500 | Medium |
What Causes PCIe Gold Finger Damage on GPUs?
PCIe gold finger damage happens when the card is inserted or removed at an angle, when the retention clip is forced, or when heavy coolers cause the card to sag and stress the connector. I've seen a sharp rise in cracked traces near the gold fingers, especially on RTX 4090 and RTX 5090 cards, because those coolers weigh over 2kg. According to my own workshop records, PCIe edge damage accounts for about 12% of all GPU repairs I handle.
The gold fingers are the contact strips that slot into your motherboard's PCIe slot. If a trace cracks just behind the fingers, the card may lose a PCIe lane, drop to x8 or x4, or fail to detect entirely. Sometimes it boots but crashes under load.
Hong Kong's humidity makes this worse. I've pulled cards out of flats in Sham Shui Po and Mong Kok where the gold fingers were visibly oxidised and green. That oxidation increases contact resistance and can cause intermittent detection failures. A proper clean with isopropyl alcohol and a fibreglass pen fixes mild cases, but cracked traces need microsoldering.
Can a Cracked PCIe Trace Be Repaired?
Yes, a cracked PCIe trace can be repaired by exposing the copper underneath the solder mask and bridging the break with a fine jumper wire, then sealing it with UV solder mask. This is delicate work — the traces are often less than 0.1mm wide. In my workshop, we use a microscope and a 0.02mm soldering tip. Success rate is around 85% for single-trace breaks, dropping to about 60% when multiple adjacent traces are damaged.
The repair usually takes two to four hours and costs between HK$800 and HK$1,800 depending on how many traces are affected and whether the card needs a full reball afterward. For a card worth HK$15,000, that's often worth it. For an older GTX 1060, honestly, it's usually not.
Here's the thing: many shops will tell you the card is dead and try to sell you a new one. I always recommend getting a second opinion from a technician who actually does microsoldering, because a repairable trace crack is very different from a dead GPU die.
Why Does Hong Kong Humidity Make GPU Problems Worse?
Hong Kong's average relative humidity sits between 70% and 85% for most of the year, and that accelerates oxidation on exposed metal contacts. I've measured contact resistance on gold fingers from humid flats at three to five times the resistance of cards kept in air-conditioned rooms. That extra resistance causes voltage drops under load, which can lead to instability, black screens, and in severe cases, arcing that damages the PCIe slot itself.
The fix is simple: keep your case in a dehumidified room if possible, clean your card's gold fingers every 12 to 18 months with isopropyl alcohol, and avoid leaving your PC in a closed room during the wet season without ventilation. If you notice your card occasionally not detecting on boot, that's an early warning sign.
For water-cooled builds, humidity plus a slow leak is a nightmare combination. I've written separately about water damage repairs, but the short version is: check your fittings every six months.
When Should You Repair vs Replace a GPU?
You should repair a GPU when the repair cost is under 30% of the card's replacement value and the card is less than four years old. Based on my workshop data, that threshold holds up well. A HK$1,200 PWM IC repair on a HK$6,000 RTX 4070 makes sense. A HK$1,800 trace repair on a HK$1,500 used card does not.
Also consider scarcity. If your card is out of production and you can't easily buy a replacement at a fair price, repair becomes more attractive. I've had customers bring in RTX 3090s that they simply couldn't replace without paying scalper prices, and a HK$1,500 repair was clearly the right call.
One more thing: always ask for a diagnostic report before authorising work. A good shop will tell you exactly what failed, show you photos, and give you a fixed quote. If a shop refuses to show you the failed component under a microscope, walk away.
How Does ComputerKing Handle GPU Repairs?
At ComputerKing, every GPU repair starts with a full diagnostic under FLIR thermal imaging and a microscope inspection of the PCB, PCIe fingers, and fan headers. We've handled over 10,000 Hong Kong repair cases, and GPU work is one of our core specialties. We quote in HKD before touching anything, and we don't charge a diagnostic fee if you proceed with the repair.
We repair PWM IC failures, PCIe trace cracks, VRAM faults, BGA reballing, and 12VHPWR connector melting. If you're not sure what's wrong, send us a photo of the card and the symptoms via WhatsApp at +852 2186 8010. We'll give you an honest assessment of whether it's worth fixing.
Frequently Asked Questions
Q: Can I just replace the fan myself if it's not spinning?
A: Yes, if the fan motor is dead and the PWM IC is fine. Replacement fans cost HK$150 to HK$400 and usually plug into the same header. But if the PWM IC has failed, a new fan won't spin either — you'll need microsoldering.
Q: Is it safe to run a GPU with a dead fan?
A: No. Without fan cooling, GPU temperatures can exceed 90°C within minutes, causing thermal throttling, crashes, and long-term damage to the GPU die and VRAM. Stop using the card until it's repaired.
Q: How long does a PWM IC repair take in Hong Kong?
A: Typically one to two days once the replacement IC is in stock. Sourcing uncommon ICs can add two to five days. ComputerKing usually completes standard PWM IC repairs within 48 hours.
Q: Can PCIe gold finger damage be prevented?
A: Partly. Use a GPU support bracket to reduce sag, insert and remove the card straight without angling, and clean the gold fingers with isopropyl alcohol every 12 to 18 months. In humid Hong Kong flats, consider a dehumidifier.
Q: My GPU fans spin at 100% constantly — is that the same fault?
A: Not always. Constant 100% fan speed often means the PWM signal is lost, so the fan defaults to full speed. It can also be a corrupted VBIOS fan curve. A technician can diagnose which one in under an hour.
Conclusion
GPU fan failures and PCIe gold finger damage are two of the most fixable problems I see, yet they're also the two most commonly misdiagnosed as "dead card." The PWM IC is usually the real culprit behind a non-spinning fan, and cracked PCIe traces are repairable with the right microscope and microsoldering setup. In Hong Kong's humidity, both problems get worse over time, so early diagnosis matters.
If your card is showing any of these symptoms, don't wait until the damage spreads. WhatsApp ComputerKing at +852 2186 8010 for a diagnostic, or bring the card in for a microscope inspection. We'll tell you honestly whether it's worth repairing.
*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: 2026-09-17*