How do I work out what capacity I actually need for my project?

In short

Start from the data the device will hold at its worst moment, not the average, then add headroom for the file system and for growth. Calculate the realistic payload, multiply by the number of files the device must retain, and add roughly 10–15% for file system overhead and fragmentation. Then stop: choose the next standard band up, not the biggest available. Over-specifying capacity is the most common and most expensive buying error in flash, because it moves you into denser memory where counterfeit product is more common, and it costs more per unit for space nobody uses.

The three inputs

Every capacity decision reduces to three numbers, and buyers usually skip two of them.

InputQuestion to answerCommon mistake
Payload per itemHow big is one file, session or record?Using average instead of worst case
Retention requirementHow many items must remain before overwrite?Assuming users delete old files
OverheadFile system and fragmentation allowanceForgetting it entirely

Worked example (illustrative)

Suppose a device logs 200KB per record, retains 90 days of records, and writes one record every five minutes. That is 288 records a day, so about 26,000 records retained, at 200KB each — roughly 5GB of payload.

Add 15% overhead and you need about 6GB. The right answer is an 8GB part, not a 32GB one. The larger part costs more, sits higher on the density curve, and provides nothing the workflow will ever touch.

The arithmetic is trivial; the discipline of doing it is what most orders lack.

Why over-specifying is a real cost, not just a lazy one

It looks like safety margin. It behaves like three separate risks.

  • Higher density means tighter supply and longer lead times, which is precisely wrong for a product you need reliably.
  • Counterfeit and under-sorted product concentrates at the high-capacity end, where the price gap tempts more people.
  • Unit cost rises for capacity that stays empty for the life of the product.
  • Usable capacity does not scale linearly with cost — the price curve steepens at the top.

When to deliberately over-specify

There are honest reasons to buy more than the arithmetic requires: if the device must hold a full export in one go, if field replacement is expensive, or if the product has a firmware roadmap that will add data volume.

Those are specific and arguable. "It was the next size up and barely more expensive" is neither — it usually means the price curve was misread.

Sources

Figures quoted on this page are from the published reporting listed above. Where a number is an estimate rather than a published figure, it is described as approximate.

Frequently asked questions

Why does a 32GB card show only about 29GB usable?

Because nominal capacity is quoted in decimal units (1GB = 1,000,000,000 bytes) while operating systems report in binary units (1GiB = 1,073,741,824 bytes). The same physical capacity therefore displays as a smaller number, and the file system reserves some space for its own structures. A healthy 32GB device showing roughly 29–30GB is normal; one showing much less warrants a capacity test.

Should I choose capacity by price per gigabyte?

No, because price per gigabyte falls as capacity rises, which would always push you upward. The relevant number is cost per unit for the capacity your workflow needs. Optimising price per gigabyte leads directly to over-specification, and to the supply and counterfeit risks that come with the top of the range.

Does capacity affect speed?

Often yes, in the direction people do not expect. Higher-capacity parts usually have more memory dies working in parallel, so sustained write speed can be better. But high density also produces more heat in the same small body, so a large-capacity drive may throttle harder during long writes. Capacity, speed and thermal behaviour are linked, which is another reason to specify the capacity the workflow needs rather than the largest available.

Working on a specification?

Send us the product type, capacity, quantity and target date. We reply to business enquiries within 1 business day.

Request a quote
WhatsApp / WeChat Request a quote