Why a 128GB Drive Shows 119GB — Nominal vs Usable Capacity, and How to Prove a Fake

In short

A 128GB card shows about 119GB in your operating system, and that is arithmetic rather than a defect: capacity is sold in decimal units of 1,000,000,000 bytes while the operating system counts in binary units of 1,073,741,824 bytes, a ratio of about 0.93, before the file system takes its own share. A card that shows 7.4GB against a 128GB label is a different problem entirely — the controller has been programmed to report memory that is not there. The first is settled by a calculation. The second is settled only by filling the entire claimed capacity with known data and reading every byte back, because every faster check asks the controller, and the controller is the part doing the lying. H2testw on Windows, or f3write followed by f3read on macOS and Linux, does that. A folder copy does not, because it never reaches the boundary where a faked drive fails. WINTEL supplies this from its own plant in Shenzhen, China.

Two problems wearing one name

A buyer opens a shipment, mounts a card, sees 29.8GB where the label said 32GB, and files a complaint. Another buyer opens the same shipment, mounts a card, sees 7.4GB, and files the same complaint. The two situations have nothing in common, and a supply chain that treats them as one problem spends money on the wrong fix.

One is unit notation. The other is fraud. Sorting them out is the difference between a productive conversation with a supplier and eighteen months of arguing about a number that was never in dispute.

What you seeWhat it actually isWhat resolves it
128GB reports about 119GBDecimal capacity displayed in binary unitsNothing — it is arithmetic
32GB reports 7.4GBA controller reporting memory that is not thereA full write-and-read-back test
Passes at 10 per cent full, then corruptsThe same fake, reached further inThe same test, run to full capacity

Where the shortfall comes from

Nominal capacityBinary bytesRoughly usable in the OS
8GB7.45 GBRoughly 7.4 GB
32GB29.8 GBRoughly 29.8 GB
64GB59.6 GBRoughly 59.6 GB
128GB119.2 GBRoughly 119 GB
256GB238.4 GBRoughly 238 GB

Why the gap exists at all

Divide one unit definition by the other and you get 0.9313. Multiply 32GB by that and you land at 29.8GiB. The gap is about 7 per cent, it applies to every capacity sold the same way, and it is not a defect any supplier can fix by shipping more NAND. IEC 80000-13 is the standard that names both prefixes, which is why both numbers are technically correct.

A second, smaller slice goes somewhere else. Controller firmware occupies space. A bad-block reserve is held back so the device can substitute failing blocks over its service life. The file system adds its allocation tables and metadata. On a large card these are small; on a card near the bottom of its band they are a visible fraction.

Neither is recoverable and neither is a defect. You cannot claim the missing gigabytes back, so the useful response is to know the figure your devices will report before you size an order against it.

When the shortfall is normal

  • A 128GB card reporting around 119GB
  • A small spread between individual units in the same batch

When the shortfall is a defect

The line between the two is drawn by unit of measurement, not by size of gap. A shortfall counted in single-digit gigabytes on a large card is the conversion. A shortfall counted in tens of gigabytes is a device reporting capacity it does not have.

  • A 128GB card that reports 8GB — the classic misrepresentation
  • A device that stops accepting writes part-way through a full write
  • Reported capacity materially below the binary equivalent, for example a 32GB card reporting under 28GB

How a faked capacity actually works

A flash drive or card is a controller plus NAND. The operating system never talks to the NAND; it asks the controller how big the media is and trusts the answer. Faking edits that answer. The controller is configured to report, say, 128GB when the NAND behind it is 32GB, and three things then happen in order.

It looks fine — Explorer, Finder and Disk Utility all report the larger size, and a format completes successfully. It works for a while — the first few gigabytes land in real memory, which is why a folder of photos copies over without complaint. Then it fails silently at the boundary: once writes pass the real capacity the controller has nowhere to put them, so it wraps around and overwrites the oldest data. The new file looks saved. The old one is gone, usually without an error message.

The defining feature of a faked drive is not that it fails. It is that it fails late and quietly, which is why the damage lands in exactly the uses that hurt most: wedding photographs, a music library handed out as an album, a student archive, a shop backup that was never a backup.

Why a quick test always passes

Most buyers test a new drive by copying a folder onto it and opening a couple of files. That test is structurally incapable of detecting the problem. A folder of documents is a few hundred megabytes and lands entirely inside the first few gigabytes, which is real memory. Every file opens, the drive appears perfect, and the order goes in.

Testing 10 per cent of the nominal capacity tests the 10 per cent that almost always works.

The test that does work

A capacity test does three things a file copy does not: it writes until the media is full, it writes a known pattern, and it reads every byte back and compares.

MethodWhat it proves
Copy a folder and open the filesThat some memory works — nothing past the real capacity
Read the size in Explorer or FinderNothing. The reported size is the claim under suspicion
Format the driveNothing. A faked drive formats successfully
chkdsk, or disk repairFixes directory errors; cannot create missing NAND
Write to full, read back, compareThe real usable capacity — the only method that does

Reading the result

The tools are free and established: H2testw on Windows, f3write followed by f3read on macOS and Linux. Both fill the device to its nominal capacity and verify what comes back. A pass reports the full claimed capacity written and verified with no corrupt files. A fail stops early, reports a smaller usable size, or lists corrupt files starting at a specific point.

That starting point is diagnostic. Corruption beginning about a quarter of the way into a 128GB drive points to roughly 32GB of real NAND, which tells you what the drive actually is rather than only that it is bad. H2testw labels the three failure modes Data LOST, Corrupted and Overwritten; none of them is a percentage shortfall, because the drive did not lose a little capacity — it destroyed data the host believed was safely written.

It is slow because it is real I/O, and hours on a large drive at USB 2.0 speeds is normal. That has a procurement consequence buyers rarely think about: a supplier claiming to test every unit at volume must have the equipment and the hours to do it, and you can ask them to describe both.

Running it at volume: sampling that holds up

Filling a 256GB card takes hours. Doing it across 5,000 units is not a plan, it is a month of one bench doing nothing else. This is what acceptance sampling exists for, and the standards are public: ISO 2859-1 and its American twin ANSI-ASQ Z1.4.

Lot sizeSample size at General Inspection Level II
1,201 – 3,200125
3,201 – 10,000200
10,001 – 35,000315

What AQL does and does not mean

At a sample size of 125 the acceptance and rejection numbers are Ac=3 / Re=4 at AQL 1.0, and Ac=7 / Re=8 at AQL 2.5. Common defect-class assignments run critical 0, major 2.5, minor 4.0.

The point about AQL that gets misread most often is that it is a planning parameter, not a defect allowance. AQL 2.5 does not mean 2.5 per cent of the shipment may be bad. It means a lot whose true defect rate sits near that level will usually be accepted — a statement about the risk you are choosing to carry, not permission for a supplier to ship it.

Capacity is a classic outgoing-stage item, because it is the last thing that can be verified before a carton is sealed.

The five lines to write into an inspection plan

  • Nominal capacity in decimal bytes, then the binary equivalent, stated so nobody has to re-derive it
  • The calculated usable floor, after firmware, bad-block reserve and file system overhead
  • The test method and tool, named — full-capacity write-and-read-back with H2testw, or f3write plus f3read
  • Sample size and where it comes from, anchored to ISO 2859-1 Level II rather than to habit
  • What happens when a unit fails, written into the purchase order before the goods move

Why the fifth line decides the other four

A test result recorded in an inspection email has no standing once the invoice is settled. The same figure written into the order turns a measurement into an acceptance line with a remedy attached. Any supplier can be wrong once; the question that matters is what they have already agreed to do when your test disagrees with their label.

How large does flash actually go — and why a 2TB drive does not exist

The largest genuine USB flash drive in production sits around 1TB, and those products are physically large metal-bodied things. A 2TB stick is not hard to make — it is a product that does not exist, which is why the ones on sale are lying about capacity rather than exceeding a limit. Three constraints set the ceiling.

  • Die density. NAND is made in dies of a fixed capacity per generation. To reach a total you stack dies, and each die generates heat when written.
  • Thermal limits in a small body. A USB-A stick is passively cooled with very little surface area. Stack enough dies and the sustained write generates more heat than the body can dissipate, so throughput throttles and in the worst case the drive turns unreliable.
  • Controller addressing. The controller has to address the whole array and manage bad blocks, wear levelling and error correction. Past a point the management overhead becomes significant against the usable space.

The capacities that cannot be built

Buyers also ask for numbers that fall between the standard steps, and the answer is arithmetic rather than price or effort. Flash is addressed in binary: an address line doubles the addressable range, and nothing multiplies it by 1.125. The buildable set is powers of two and small whole-number multiples of them — 4, 8, 16, 32, 64, 128, 256, 512 gigabytes.

RequestedNearest standard stepsWhy
36GB32GB or 64GBNo die and addressing combination produces 36GB
48GB32GB or 64GBSame — the grid has no step there
96GB64GB or 128GBSame
200GB128GB or 256GBSame

But someone online sells a 36GB one

They do, and it is one of two things. Either a larger array with the controller instructed to expose only part of it — a real product, priced off the larger array, which is why a 36GB part costs more than a 64GB one rather than less — or a firmware change with no array behind it, which is the fake described above.

Both look identical in an operating system. Only a full-capacity test tells them apart, which is why the test is the answer to this question as well as to the previous one.

Why anyone bothers to make one

Every fake capacity in the market is one of three mechanisms, and they are not equally expensive. That asymmetry is the whole story.

MechanismWhat the buyer sees
Relabelled capacity — a smaller die marked, packed and labelled as a larger oneA card that fails as soon as writes pass the real capacity
Rejected die — parts that failed wafer-level screening, bought on the open marketUnstable data even before capacity is reached
Firmware over-report — the die is honest, the controller is told to report a bigger numberPerfect behaviour until the last few gigabytes

The one that costs nothing

The third mechanism needs nothing but a decision and a copy of a tool that is already in circulation. There is no material to trace, no purchase order to compare, no batch that looks wrong on paper. That is why the problem never goes away in a market where the detection tools are free: the gain is cheap, the check is not, and the check has to be paid for by someone who would rather not.

A buyer cannot audit a die cost breakdown, and asking for one is not a productive opening. What works is a sanity check on the number itself. Take the quoted unit price, subtract what you can see and verify — housing, printing, packaging, freight, the certificate, the sample — and ask whether what remains can pay for screened die, a controller and assembly. If it cannot, the difference is being found somewhere, and there are only three places to find it. That is a filter rather than a proof, but it holds because the honest answer is bounded. Our own fixed items are published: a 50-piece floor, USD 200 for a tool only if the shape is new, USD 350 per order for certification, about USD 20 for a sample.

When a capacity test is the wrong tool

Four cases where this is not what you should be doing. If the shipment arrives with a certificate from a testing house your own quality system already recognises, re-running a full sweep on every carton is duplicated cost — spot-check, file the certificate, move on. If the card goes into a product where the data is disposable, such as a promotional giveaway or a two-week trade show loop, a capacity failure is an inconvenience rather than a liability; test the sample, not the shipment.

If your real risk is counterfeit branding rather than fake capacity, the method is different: a write-and-read-back confirms how much the part holds, and says nothing about whether the silicon inside is the grade the certificate claims. And if a supplier says their units are individually capacity-tested but will not share the sample size or the method, the number on offer is not evidence. Ask for the method, or buy the test yourself.

What to specify on an order

  • State the nominal capacity you require and accept the usable shortfall as normal — it always exists
  • Do not specify in usable gigabytes unless you have a reason to; the market convention is nominal
  • If capacity is critical, work backwards from your content size to the nominal figure that comfortably covers it, with headroom
  • State the usable floor you will inspect against, so the acceptance line is agreed before the goods move
  • For a recording workload, capacity that looks implausibly generous is usually a sign of a misrepresented device rather than a bargain

Not sure which applies to your product?

Send your workload — what writes, how often, at what temperature — and we will propose a configuration with the reasoning. We reply to business enquiries within 1 business day.

Request a quote

Frequently asked questions

Is a 128GB card showing 119GB fake?

No. That is the normal decimal-versus-binary gap plus file system overhead, about 7 per cent. A card showing 8GB when the label says 128GB is the real problem.

Can a card reader detect fake capacity?

No. A reader reports what the controller says, and a controller advertising false capacity is the component under suspicion. Only a full write-and-read-back compares the claim against the NAND.

Can I get the missing capacity back by reformatting?

No. The difference is arithmetic and file system overhead, both unavoidable. A reformat can only recover space taken by a poorly chosen file system, not the decimal-versus-binary gap.

How much should a 256GB card show?

Roughly 238GB. If it shows materially less than that, test it properly before you rely on it.

How many units should I pull from a shipment?

ISO 2859-1 and ANSI-ASQ Z1.4 set the framework, and General Inspection Level II is the default when the contract is silent. That gives 125 units for a lot of 1,201 to 3,200, 200 units for 3,201 to 10,000, and 315 units for 10,001 to 35,000.

Why can I not order a 36GB card?

Because flash is addressed in binary. Capacity is built from whole numbers of dies of fixed size, so the buildable set is powers of two and small multiples of them. There is no die and addressing combination that lands on 36GB, which is why the honest answer is 32GB or 64GB rather than a longer lead time or a higher price.

What is the largest genuine USB flash drive?

Around 1TB. A 2TB USB-A stick would need more NAND dies than the passively cooled body can keep within thermal limits, so drives advertised at 2TB are capacity-faked rather than ahead of the market.

Will your products pass a capacity test?

Yes. Every unit is capacity-verified by a full write-and-read before packing, and the method is published so you can repeat it on anything you buy from us or from anyone else.

Sources

External sources explain industry standards (accessed 2026-11-10); they describe the standard, not WINTEL product parameters.

WhatsApp / WeChat Request a quote