How to Choose a GPU Repair Shop in Hong Kong: Chip-Level Repair vs Board Swap & How to Judge Real Expertise
Key Takeaways
Chip-level repair keeps your original board and fixes only the faulty component; a board swap replaces the card. Neither is always better. Judge a shop by verifiable behaviour — diagnosis before quoting, a written itemised quote, written warranty terms, a receipt, and answers about parts — not by slogans or by a price quoted before anyone has seen the card.
First, What Board Swap Means for a Graphics Card
In graphics card repair, a board swap means replacing the whole graphics card, or the whole printed circuit board assembly, rather than repairing the faulty part on your existing card. It does not mean swapping the computer's motherboard. That distinction matters, because readers coming from laptop or phone repair will have met board swap in its motherboard sense, and the two ideas are easy to confuse.
Put plainly, there are two broad ways to deal with a faulty graphics card. Chip-level repair, also called component-level repair, keeps your original board and fixes the specific component that failed. A board swap removes the whole card and puts a different card in its place — a new, second-hand, or refurbished one, depending on what is available and what you agree to. Everything in this article follows from that difference: one approach repairs your card, the other replaces it.
How Graphics Cards Fail: Why the Repair Method Matters
A graphics card is an assembly of parts that fail in different ways, and the right response depends on which part failed. Three layers cover most cases.
The first layer is the small components around the GPU core: memory chips, power delivery components such as MOSFETs and supply ICs, capacitors and resistors, and connectors such as the 12VHPWR socket. These are individual parts on the board. When one of them fails, the fault is often local: one memory chip with a bad channel, one phase of the power delivery system, one damaged connector. A local fault can often be repaired locally.
The second layer is the solder joints that connect those components to the board. Graphics cards run hot, cycle through heating and cooling, and are sometimes carried, knocked, or flexed. Over time, solder joints can crack or fatigue, and connections become intermittent. A joint problem can mimic a component problem — the card works, then does not, then works again — which is why diagnosis has to separate the two before anyone decides on a method.
The third layer is the board and the core itself. Severe board damage, such as burned or lifted traces and multi-point short circuits, and a damaged GPU core, are a different category. These are not local faults with a clean fix. As one repair guide puts it, a dead GPU core is generally not economically repairable, and replacing the card is the sensible outcome. Knowing which layer your fault sits in is the whole point of diagnosis, and it is what should drive the choice between repair and replacement.
Chip-Level Repair: Fix the Faulty Part, Keep the Board
Chip-level repair is board-level component repair. The process runs in a chain: measure, locate, desolder, replace, resolder, test. First, the technician measures the board — checking resistance values across the power rails and other test points — to narrow down where the fault sits. Then the faulty component is located, desoldered under a microscope, and replaced with a component of the correct specification. The new part is soldered back, and the card is tested, including memory testing and load testing, to confirm the fault is gone.
That chain is worth understanding because it explains both the strength and the cost of chip-level work. Its strength is precision: the repair targets the root cause, and the rest of your card — its board, its other components, its firmware — stays as it was. Its cost is that it needs real diagnostic work and fine soldering skill, and it is not suitable for every fault. A shop that is honest about those limits is more trustworthy than one that promises to fix everything.
One part of chip-level work confuses owners most, because three similar-sounding procedures are often blurred together. They are not the same, and the difference matters.
| Procedure | What actually happens | What it can and cannot do |
|---|---|---|
| Reflow | The chip is heated so the existing solder joints re-melt and rejoin; no new solder is added | Often only temporary: heating old joints can make the card work again for a while, but the underlying joint or component problem remains |
| Reballing | The chip is removed, the old solder is cleaned off, and a fresh grid of new solder balls is applied before the chip is reinstalled | Addresses fatigued or cracked solder joints properly, by replacing the solder rather than reheating it |
| Chip replacement | The faulty chip is removed and a new component of the correct specification is installed | The right answer when the chip itself is faulty; but if the chip's internal die is damaged, neither reflow nor reballing can save it |
Two rules follow from that table. First, a bare reflow offered for visual artefacts is a warning sign: one repair guide warns against accepting a bare reflow for artifacting, describing heating the board until it temporarily works as a rental, not a repair, that usually fails again in weeks. Second, reballing cannot repair a damaged die, a defective internal connection, or an unrelated board fault. An intermittent symptom does not by itself prove that the solder joints are the cause; the failure evidence has to be reviewed first. If a shop's entire answer to every fault is to heat the board, it is not diagnosing.
Board Swap: Replace the Whole Card
A board swap takes the other path: instead of finding and fixing the faulty component, the whole card is exchanged for another one. The replacement might be a new card of the same model, a second-hand card, or a refurbished one. Which of those you are getting, and what warranty it carries, are questions to settle before agreeing, because the words new, refurbished, and second-hand describe very different things.
The main advantage of a board swap is speed. If a suitable replacement card is sitting on the shelf, the exchange can be quick, with no slow hunt through individual components. For a card with severe board damage, multi-point faults, or a dead core, replacement may also be the only realistic option: there is no component to replace that would make the card whole again.
The main disadvantage is that a swap does not necessarily find the root cause. Published guidance on component-level versus board-level repair makes the point directly: a board swap is quick when a spare board is on the shelf, but without diagnosis, the new board can fail the same way. The logic applies to graphics cards with full force. If the real cause of the failure sits outside the card — a poor-quality power supply, a damaged 16-pin cable, or a case with badly restricted airflow — then a replacement card is exposed to the same cause, and can fail the same way. A swap that skips diagnosis can therefore be an expensive way of repeating the original fault.
There is also a compatibility question. A replacement card needs to be the right card for your system: the correct model and memory configuration, with firmware that matches the board revision. Firmware and board revisions are covered in more detail below. And the replacement card's own warranty — its length, its scope, its exclusions — needs to be in writing, exactly as a repair warranty would. A swapped-in card with no written warranty is a risk, not a solution.
One reassurance is worth stating plainly. A graphics card holds no user files: no photos, documents, game saves, or work files live on the card. What it holds is firmware on its own chip. Your drivers, settings, and game profiles live on the PC's storage drive. Swapping the card does not move your files, because there are no files on the card to move. After a swap, the usual step is to remove the old driver cleanly and install the latest official driver, then test the card.
Which One Fits? A Decision Framework
Neither method is always better. The right choice depends on what the inspection finds, and the table below gives the shape of that decision. It is a framework for thinking, not a verdict on any individual card: the verdict belongs to the inspection results and the written quote.
| What the inspection finds | Which way it usually leans | Why |
|---|---|---|
| A single faulty component, clearly located, with the rest of the board healthy | Chip-level repair is usually worth considering | The fault is local, the fix is targeted, and the original board is preserved |
| Intermittent fault traced to fatigued solder joints | Reballing, within chip-level repair | The joints are the problem, and new solder addresses them properly; a bare reflow is usually only temporary |
| Severe board damage, burned traces, or multi-point short circuits | Replacement is often the more realistic option | The board itself is the problem, and component-level fixes cannot restore it |
| Damaged GPU core | Replacement is usually the sensible outcome | A damaged core cannot be fixed by reflow or reballing |
| Repair cost approaching the cost of a replacement card | Replacement deserves serious consideration | A repair that costs nearly as much as a replacement leaves little margin for the risk that further faults appear |
| Fault cause sits outside the card, such as a poor-quality power supply or damaged cable | Fix the outside cause first, whichever route you take | A replacement card exposed to the same cause can fail the same way |
Notice what is missing from the table: any mention of the shop's size, branding, or marketing. The choice between repair and replacement is a technical and economic decision about your card. It should be made on inspection evidence and a written quote, not on a slogan.
VBIOS, Board Revisions, and the Line Between Repair and Modification
Graphics cards do not have the phone-style pairing locks that some consumer electronics use, where a replaced part must be digitally paired to the device before it works. That is not how graphics cards behave. What does matter is matching: the VBIOS firmware and the board revision have to correspond, and a VBIOS written for a different board revision or a different memory configuration cannot be used casually.
The point is easy to see in an extreme example. Fitting a GPU core onto a different printed circuit board only produces a working card once a matching BIOS is flashed — the core, the board, and the firmware have to agree. Ordinary repair work does not go anywhere near that territory, but the principle is the same at smaller scale: replacing a component with one of the correct specification keeps the card's firmware valid, while mixing board revisions or memory configurations and patching it with the wrong VBIOS invites new problems.
The same line separates repair from modification. Expanding a card's memory capacity — fitting higher-capacity memory chips, changing strapping resistors, and pairing the result with a modified VBIOS — is a modification, not a routine repair. It changes the card's design parameters rather than restoring them. A shop that presents memory expansion as an ordinary repair option is blurring a line that should stay sharp. Repair restores the card to its original working state; modification changes what the card is. Owners should understand which of the two they are being offered, and decide accordingly.
After any repair or swap, the standard finishing step is to remove the old graphics driver cleanly — using a driver removal tool — and install the latest official driver, then test the card for the original symptom: visual artefacts, black screens, or error codes such as Code 43. A repair that is not tested after collection is not yet a confirmed repair.
How to Judge Real Expertise: Verifiable Standards
Expertise in repair is not something you can judge from a logo, a slogan, or a promise. It shows up in verifiable behaviour: things the shop does, writes down, or shows you, before and after the work. The checklist below is drawn from published consumer guidance for repair shops generally, and it applies to graphics card repair just as well.
Diagnosis before quoting. A reliable shop diagnoses before it quotes. A firm price given over the phone, before anyone has looked at the card, is a guess dressed up as a price. Ask what the inspection includes: resistance measurements, thermal imaging or oscilloscope checks, memory testing. A shop should be able to describe its diagnostic steps in plain language.
A written quote that names the fault and the part. A written quote should say what is wrong, what will be done, and what it costs. Vague wording is a warning sign: a quote that names the part or the fault — replacing a specific memory chip, for example — is an answer, while a quote that says only board problem is not. If the price changes during the repair, the shop should explain why before doing anything more.
Itemised pricing. The quote should separate parts, labour, and any inspection fee, so you can see what you are paying for. Ask what happens if the repair does not fix the problem, and what you owe if you decide not to proceed after inspection. The shop's published price range is HK$400–2,800, with the actual fee set by the post-inspection quote; any figure you are given should be tied to an inspection, not to a phone call.
Written warranty terms. A shop that offers a warranty but cannot tell you its duration or its terms when asked is unlikely to honour it. Ask how long the cover lasts, which repaired area it covers, what it excludes, and how the start and end dates are counted — and get the answers in writing on the receipt or work order. The warranty questions are covered in detail in the companion article on how GPU repair warranty works, including the point that this shop's public pages show both 90 days and 90 to 180 days, so the receipt or work order is the binding document.
A proper receipt or invoice. The receipt should carry the shop's name, address, and contact details, plus the repair item details. The shop's published FAQ states that its receipts carry the Hong Kong Business Registration number, the repair item details, part batch numbers, and an engineer stamp. Without paperwork, a later warranty claim has no foundation.
Before-and-after records. Before sending the card, record the fault yourself: a short video, a photo, a note of the symptom and when it appears. At collection, ask what the shop's own testing showed — memory test results, load testing, or whatever its process produces — rather than accepting a bare statement that the card is fixed.
Parts provenance, stated plainly. Ask what kind of replacement parts are used: new, original-specification, refurbished, or aftermarket. A legitimate shop answers that question transparently. Refusal to discuss parts provenance is a warning sign, as is a price so low that it can only mean cut corners or inferior parts.
No pressure. Be cautious of urgency tactics: a price that only exists if you decide on the spot, or warnings that the card will get worse if you do not commit immediately. A normal repair decision deserves time to compare and think. A legitimate shop welcomes questions and answers them transparently; pressure is a substitute for answers.
Relevant experience, demonstrated rather than claimed. Ask whether the shop has handled your card's model and your type of fault — a burned 12VHPWR connector, memory artefacts, a power delivery failure — and what test equipment it uses. Doing a bit of everything is not the same as being equipped for your fault. The answer you are looking for is specific, not a slogan.
Red Flags: When to Slow Down
No single item on this list proves a shop is dishonest. But when two or three appear together, slow down and compare elsewhere.
- No written quote, or a quote that does not name the fault or the part.
- A firm price quoted over the phone before anyone has seen the card.
- A warranty described only as we offer a warranty, with no duration, scope, or terms the shop can state.
- No receipt or invoice, or paperwork without the shop's name, address, and contact details.
- A bare reflow presented as the whole repair for visual artefacts.
- Refusal to say what kind of parts will be used, or what happens if the repair does not work.
- Pressure to decide immediately, or a price that evaporates if you leave to think.
- Guarantees that sound too good to be true: claims of fixing everything, fixing anything permanently, or never failing. Real repair work has limits, and honest shops say so.
The Questions to Ask Before You Commit
Take these questions with you, in this order, and judge the shop by the quality of its answers:
- What does your inspection include, and will I see the inspection result?
- What exactly is wrong with my card, and which part will you replace or repair?
- Can I have that in a written, itemised quote — parts, labour, and any inspection fee, separately?
- What kind of replacement parts do you use, and can I see the part before it is installed?
- What are the warranty terms, in writing: how long, what is covered, what is excluded, and how are the start and end dates counted?
- What happens if the repair does not fix the problem, or if I decide not to proceed after inspection?
- What did your testing show after the repair — and can I see the result?
A shop that answers these plainly, in writing, has given you something verifiable. A shop that waves them away has given you something else: information about itself.
Frequently Asked Questions
Is chip-level repair always better than a board swap?
No. Chip-level repair is the better fit when the fault is local, the faulty component can be located, and the rest of the board is healthy. A board swap or whole-card replacement is often the more realistic option for severe board damage, a damaged GPU core, or a repair whose cost approaches the cost of a replacement card. The choice should follow the inspection evidence, not a preference for one method over the other.
Is a reflow a real repair?
Usually not a lasting one. Reflow reheats the existing solder joints without adding new solder, and the card often fails again in weeks. Reballing replaces the solder balls with fresh ones and addresses fatigued joints properly. If the chip itself is faulty, the chip needs replacing. A shop whose answer to every fault is to heat the board is not diagnosing.
Should I accept a firm price quoted over the phone?
No. A firm price given before anyone has inspected the card is an estimate, not a quote. A reliable quote comes after diagnosis, names the fault and the part, itemises parts and labour, and is put in writing. The shop's published price range is HK$400–2,800, with the actual fee set by the post-inspection quote.
If the card is swapped, do I lose my files or settings?
No files live on the card, so there is nothing to lose from the card itself. Your drivers, settings, and game profiles live on the PC's storage drive. After a swap, remove the old driver cleanly and install the latest official driver, then test the card.
What if the shop says my card cannot be repaired?
Ask what the inspection found, and whether the conclusion is a damaged core, severe board damage, or unavailability of a part. Then decide with that evidence in hand: a second opinion is legitimate, and a written inspection finding makes one possible. A card that cannot be repaired is not a failure of your judgement in seeking repair; it is information.
How do I compare two shops' quotes?
Compare like with like: the same named fault, the same named part, itemised pricing, and written warranty terms with the same scope. A cheaper quote that omits the warranty terms, or names no part, is not cheaper — it is less specified. The comparison that matters is the total of the repair, the warranty, and the paperwork, not the headline figure alone.
Final Checklist
Before committing a card to any shop, run through this list: inspection described and offered before quoting; written itemised quote naming the fault and the part; parts provenance stated; written warranty terms with duration, scope, exclusions, and start and end dates; receipt or invoice with the shop's details and the repair items; a clear answer on what happens if the repair fails or you decline after inspection; no pressure to decide on the spot; and prompt testing after collection, with the old driver removed and the latest official driver installed. If a shop passes that checklist, the choice between chip-level repair and a board swap becomes what it should be: a technical decision about your card, made on evidence, with the terms in writing.