Your 32GB Drive Formats to 29.8GB: Real Capacity, Fake Capacity and the Full-Write Test

Two drives arrive in the same carton, with the same branding, both labelled 32GB. Plug either one into a laptop and Windows reports 29.8GB. One of them is completely normal. The other will accept your files, show them in the folder listing, and quietly destroy everything written past the eighth gigabyte — and nobody will find out until a recipient tries to open the video three weeks after your event.

Those are two entirely different problems that look identical on screen. The first is arithmetic. The second is fraud. This article separates them, and gives you the incoming-goods acceptance procedure we run on custom USB flash drives Malaysia buyers order for events, press kits and staff gifts — including the exact test, the sample size, and how long it takes.

29.8GB is not a short measure — it is arithmetic

Storage is sold in decimal gigabytes. One gigabyte, to a memory manufacturer, is 1,000,000,000 bytes. Operating systems have historically counted in binary: what Windows calls a “GB” is actually 1,073,741,824 bytes — a gibibyte. Divide one by the other and about 6.87% of the labelled capacity disappears before anything else happens.

The controller then sets aside a slice of the flash for spare blocks and wear levelling, and the file system takes a small allocation table. Nothing has gone wrong; every honest drive behaves this way.

Labelled capacity	Bytes on the datasheet	What the laptop shows	Typical free space once formatted
8GB	8,000,000,000	about 7.45GB	about 7.4GB
16GB	16,000,000,000	about 14.9GB	about 14.8GB
32GB	32,000,000,000	about 29.8GB	about 29.7GB
64GB	64,000,000,000	about 59.6GB	about 59.4GB
128GB	128,000,000,000	about 119.2GB	about 118.9GB

Figures are the decimal-to-binary conversion plus a small allowance for controller reserve and file-system tables. Exact free space varies slightly by model and by whether the drive ships formatted as exFAT or FAT32.

Put one line from that table on the insert card and the most common post-delivery query — why the 16GB drive “only has 14.9GB” — never reaches you. It costs nothing and removes any impression that the gift was short-changed.

The other kind of shortfall: a controller that lies

A flash drive has two parts that matter here: the memory itself, and a small controller chip that tells the computer how big the memory is. Those two numbers are supposed to match. Counterfeiters reprogram the controller so that a drive holding, say, 8GB of real memory announces itself as 32GB.

Every operating system believes the controller. It has no independent way to count the memory cells. So the drive mounts as 32GB, the properties dialog says 32GB, and any capacity-reading utility says 32GB, because they are all asking the same lying chip the same question.

The failure only appears once you write past the real limit. Two things happen, depending on how the fake was built:

• Wrap-around. Data written past the real capacity loops back to the beginning of the memory and overwrites what is already there. The file listing still shows every file, because the directory entry survives — but the earlier files now contain the later files’ data. Everything opens as corrupt.
• Silent discard. Data past the limit is accepted and thrown away. The file appears in the folder at full size. Opening it produces zeros, a broken video, or an error.

Note what both have in common: the copy operation completes without an error message. That is the whole reason this survives ordinary quality control.

Why the usual spot check cannot find it

The standard incoming check on a promotional drive is: open the carton, plug in three or four drives, confirm the logo is straight, copy a small file across, open it, tick the sheet. That check passes a counterfeit every single time, because the first few megabytes of a fake drive are perfectly real. You have to fill it to find out.

The same logic applies to preloading. If your production run loads a 40MB company deck onto a fake 32GB drive, the deck lands in the first fraction of real memory and opens perfectly. The drive ships, gets handed out, and the failure belongs to the recipient the first time they use it for their own files. Which is exactly the moment your logo is on the casing.

The full-write test, step by step

The test is conceptually simple: fill the drive completely with data whose contents you can predict, then read every byte back and confirm it matches. If the drive wrapped around or discarded, the read-back will not match and the tool will tell you exactly at which gigabyte the memory ran out.

Step 1 — Use an empty drive, formatted

Test before branding and before preloading. A drive that fails should never have absorbed decoration cost, and a drive with a preloaded folder on it cannot be filled to 100%.

Step 2 — Run a write-and-verify tool

On Windows, H2testw is the long-standing choice: it writes 1GB test files until the drive is full, then verifies each one. On macOS and Linux the equivalent is the F3 pair — f3write followed by f3read. Both report the verified capacity in plain language. A genuine drive verifies everything it wrote; a fake reports a much smaller figure and lists the point of failure.

Step 3 — Budget the time honestly

This is not a five-minute check, and pretending otherwise is why it gets skipped. A 32GB drive that writes at 20MB per second takes about 25 minutes to fill and another 12 to 13 minutes to verify. The same drive on slower memory, writing at 6MB per second, takes closer to an hour and a half just to fill. Plan a working session, not a coffee break — and run several drives at once on a powered hub, since the bottleneck is the drives, not the laptop.

Our incoming-goods acceptance procedure for a 300-piece run

Testing all 300 drives is not realistic — that is why sampling exists. What matters is that the sample is drawn across cartons, not out of the top of the first box, and that a single failure escalates rather than being written off as bad luck. Fake memory arrives in production lots, so one bad drive is evidence about the lot, not about the drive.

Stage	What is done	Pass rule	Rough time
1 — Carton sweep	Open every carton. Mount three drives per carton, confirm reported capacity and that the volume label matches the specification.	All mount, all report the correct capacity band	Under an hour
2 — Full-write sample	Draw eight drives spread evenly across all cartons. Run write-and-verify to 100% on each, in parallel.	Verified capacity matches reported capacity on all eight	One working session
3 — Escalation	Any single failure at stage 2 triggers a second, larger draw from the same cartons.	A second failure rejects the whole lot — no partial acceptance	A further session
4 — Post-preload verify	After loading, open the largest preloaded file from a sample of drives on a second machine.	File opens and plays or renders to the last page	Minutes

Stage 4 is deliberately the largest file, not the first one. On a wrap-around fake, the last file written is the one that survives; on a discard fake, it is the one that fails. Opening the biggest file catches both patterns.

If you are also handing over a large file on the drive, the write speed of the memory decides how long stages 2 and 4 take — which is why we normally settle capacity and speed together rather than in sequence. We have set out how thumb drive read and write speed changes a production schedule in a separate piece.

What a fake-capacity drive actually does at the event

Nothing. That is the problem. The drives are handed out, delegates plug them in that evening, your 40MB deck opens flawlessly, and the campaign is a success. The damage is deferred by weeks or months to the point where the recipient fills the drive — a holiday photo dump, design files, an offline backup — and loses the lot.

There is no recall mechanism for a giveaway. Once 300 drives have left the room you cannot get them back and you have no way to warn anyone. That is the strongest argument for testing before branding: it is the last moment at which the problem is still yours to fix.

Writing capacity into the purchase order

Most corporate purchase orders for promotional drives say nothing more than “32GB”. That is not a specification, it is a label. Here is wording that any competent supplier can answer, and that gives a procurement team something to hold on to if the batch is wrong. It works equally well inside an ePerolehan submission or an ordinary purchase order with an SST invoice.

Capacity clause

1. Nominal capacity: 32GB decimal (32,000,000,000 bytes minimum
   addressable), shipped formatted, exFAT unless otherwise agreed.

2. Verified capacity: the full addressable capacity shall pass a
   write-and-verify test to 100% of reported capacity with zero
   mismatched sectors.

3. Sampling: buyer may draw a sample across all cartons on arrival
   and run the test above. One failure triggers an enlarged sample;
   two failures reject the lot in full.

4. Remedy on rejection: replacement of the full lot at supplier's
   cost, including re-branding and re-preloading.

5. Testing occurs before decoration and before data preloading.

Clause 4 is the one that changes behaviour: a supplier happy to replace a rejected lot including decoration is a supplier confident in their memory source. Clause 5 removes the argument that branding voided the claim.

One footnote: a rejected lot restarts the calendar. Production runs 20 to 30 working days after artwork approval plus roughly 15 days shipping, express 7 working days on limited models and quantities. Build the acceptance test in at the front — and if the body is still open, our notes on which drive housing survives a laptop bag will save you a second round of samples.

What buyers ask us about drive capacity

Is a 32GB drive that shows 29.8GB defective?

No. That is the normal decimal-to-binary conversion plus controller reserve, and every manufacturer’s drive behaves the same way. A defect would be a drive that shows 29.8GB but cannot store 29.8GB of verified data.

Can a capacity-reading utility detect a fake drive?

Not reliably. Utilities that simply read the reported capacity are asking the same controller that has been reprogrammed to lie. Only a tool that writes to the full capacity and reads it back can prove the memory exists. Some utilities also read the controller and memory chip identifiers, which can raise suspicion, but the write-and-verify test is the one that settles it.

Does a thumb drive need battery shipping paperwork?

No. A flash drive draws its power from the port and contains no cell, so it travels as ordinary goods with no battery documentation attached. It is one of the few branded electronics items where the paperwork is genuinely simple — the scrutiny belongs on the memory instead.

What is the minimum order for branded drives?

Our minimum order is from 300 pcs per design. That quantity is also large enough for meaningful acceptance sampling — testing eight drives drawn across the cartons gives you a real signal about the lot, which a handful of pieces never would.

When in the process should the capacity test happen?

Before decoration and before anything is written to the drives. A branded drive that fails has already absorbed cost, and a drive carrying a folder cannot be filled to 100 per cent, so the test has to come first — blank stock, then test, then branding, then any file loading, which we cover alongside the folder rules in our guide to preloading a company deck onto the drives.