What microSD card does a dash cam actually need?

The endurance rating is in hours, and the hours depend on a bit rate nobody states.

Memory cards are the part of a dash cam setup that people buy once and stop thinking about, and it is the part most likely to fail quietly. What follows is what the endurance numbers actually measure, why the same rating covers a fourteen-year card and a seven-month card, and the mechanism behind the fault everyone solves by reformatting without knowing why it worked.

Our method

We started from the SanDisk High Endurance datasheet and its endurance white paper, and the SD Association’s own documentation of the speed class standards and how they are applied. We then worked the arithmetic that connects a rated number of hours to the way a dash cam is actually used.

We did not run a card to failure, and we have not measured the write volume of any dash cam. Where a figure comes from the SanDisk datasheet or from the SD Association, we name which. Where the conclusion is our arithmetic from a published rating, the text says so, and the assumption it rests on is stated rather than buried.

The rating is in hours, and the hours move

Cards built for continuous recording are rated differently from ordinary ones. Rather than a total-bytes figure, the endurance specification is given in hours of video recording, and it scales with capacity.

The rating is stated for Full HD (1920×1080) video only, and the datasheet adds that the total hours are less for 4K UHD (3840×2160). No bit rate is published for the rating.

The first thing to notice is that the rating is exactly proportional to capacity — roughly 78 hours per gigabyte at every point on the ladder. That tells you the underlying quantity is a total volume of data written, divided by an assumed recording rate to produce a number in hours. The rate is the part that is not published.

The second thing to notice is the footnote. The rating is stated for Full HD, and the datasheet explicitly says the hours are lower for 4K. So the headline number is already conditional on a resolution, and it does not tell you by how much it drops when you change it. A 4K dash cam writing at a higher bit rate consumes the same rated hours faster, and the datasheet gives you no factor to apply.

None of that makes the rating useless. It makes it a rating for an unspecified workload, which is only a problem if you treat it as a lifespan.

What your duty cycle does to that number

The rating is in hours of recording, so the conversion to a lifespan is just a division — and the divisor is how many hours a day your camera actually records. That number varies more than most people expect, because a dash cam does not record only when you drive.

Take the 64 GB card and its published 5,000 hours:

Same card, same rating, and a spread of roughly 24 times between the top and bottom rows. This is the calculation that decides whether a high-endurance card is a sensible purchase or an expensive one, and it is the calculation the rating on the packaging does not invite you to do.

It also cuts the other way. If your camera only records while you are driving, an ordinary card may outlive the car, and paying a premium for endurance is buying something you will never use. If you run continuous parking recording, the card is not a one-off purchase at all — it is an annual or twice-annual consumable, and it should be budgeted that way before you decide that parking mode is free.

One clarification, because it is a common misconception: loop recording does not spare the card. Overwriting the oldest footage writes just as many bytes as recording onto empty space. A card that has been running for a year has taken a year of writes regardless of how full it looks in the app.

Why you cannot check the card you own

On a computer you can read a drive’s health, see its total bytes written, and watch it decline. None of that exists here, and the reason is documented rather than accidental.

SanDisk’s own endurance white paper makes two statements that are worth quoting precisely. Consumer cards, it says, “typically do not include specified write endurance ratings” — the endurance number is a feature of the cards sold for recording, not of the ordinary ones. And consumer-grade products “do not include a defined health monitoring register”, because that functionality is neither specified by the SD Association nor commonly implemented as a way to report remaining endurance.

The consequence is stated plainly in the same document: remaining lifetime has to be estimated by tracking operating hours and written data, because the card will not tell you. There is no percentage-left figure to check, no warning before it starts dropping frames, and nothing on the card that records how much of its life you have spent.

There is a second trap in the same document. The bill of materials — the controller, the firmware and the NAND technology inside a card — can change over time without prior notice. So a model that performed well for you two years ago is not guaranteed to be the same product today. Brand and model line are a weaker signal here than they are in most categories, and the endurance rating on the current packaging is a stronger one.

The speed class trap

Speed ratings on cards are one of the few parts of this category that are actually standardised, and the SD Association publishes how they work. There are four separate systems — the original Speed Class, UHS Speed Class, Video Speed Class and the newer Express classes — and they are not interchangeable labels for the same thing.

The Association states that the symbols must be matched between host and card, and gives the reason explicitly: expected write speed will not be available from a combination of different class symbols, even where the two indicate the same nominal speed. Its own examples are a host marked Class 10 with a U1 card, and a host marked U1 with a V10 card — both pairs that look equivalent on paper and are not, under the standard, equivalent in practice.

For a dash cam buyer the takeaway is narrow but useful: match the class symbol the camera’s documentation asks for, rather than buying the highest number you can find. A card carrying a faster class in a different system is not a safer choice by definition. This is also the reason a card can be blamed for a camera fault when the camera is the half of the pair that set the expectation.

Why cards slow down, and why reformatting fixes it

There is a fault pattern specific to this application that owners know well: the camera runs fine for months, then starts reporting errors, stopping recordings or dropping frames — and reformatting the card restores it. The standard documentation explains the mechanism, and it is not that the card was full.

The Association defines speed class performance as sequential writes to a completely un-fragmented area. Repeated deletion and writing fragments the data area, and writing to a fragmented area is slower than sequential writing — a property of the flash memory itself, not a fault in the card. The documented consequence is that even when free space exists in total, the host may be unable to perform speed class recording at all. The documented remedy is to rearrange the data or reformat the card.

A dash cam is close to a worst case for this: continuous writing, continuous deleting of the oldest files, and never being formatted in normal use. Which is why the reformat that appears to fix a dying card is usually restoring the condition the speed class was defined against, and why the same card will degrade again on the same timescale.

The distinction matters because the two causes look identical and have opposite responses. If reformatting restores the card, you have fragmentation — a maintenance issue. If reformatting does not restore it, you have consumed the write endurance — a replacement issue. Doing the first when the second is true is how people lose footage they thought they were recording.

What to check before you buy

• Match the speed class symbol your camera documents. Class 10, U3 and V30 are the ratings commonly published for this application. A higher class in a different system is not automatically the safer match.
• Look for an endurance rating in hours, and note the resolution it assumes. If a card has no endurance figure at all, it is a consumer card and it carries no stated write endurance — that is SanDisk’s own characterisation, not ours.
• Work out your duty cycle before you choose a capacity. Hours of driving a day, plus hours of parking recording a day. That divisor is what turns a rated figure into a lifespan, and it is the number that varies by a factor of twenty.
• Check the camera’s maximum supported capacity. Not every device accepts the larger microSDXC capacities, and the limit is a property of the camera, not the card.
• Plan a reformat interval. The fragmentation mechanism is documented, and the remedy is documented. Reformatting in the camera, on a schedule, addresses the failure that a reformat can actually fix.
• Budget the card as a consumable. On a continuous parking-recording duty cycle, the rated hours can be consumed inside a year. That cost belongs in the decision about whether to enable parking mode at all.

The honest answer

The card question has a straightforward half and a hard half. The straightforward half is the speed class: match the symbol the camera asks for, and buy for the write endurance rather than the read speed on the front of the pack. The hard half is that no published figure tells you how long the card in your car has left, because the rating is expressed against an unstated bit rate and the card has no way to report its own remaining life.

What you can do is put the two published numbers together — a rated number of hours, and your own recording hours per day — and treat the result as a replacement schedule rather than a hope. For whether parking mode is worth running at all, and what it does to the car’s battery rather than to the card, see dash cam hardwire kits. For whether the camera itself earns its place, see are dash cams worth it and do you need a rear dash cam.

Questions this page answers

What microSD card does a dash cam actually need?

One rated in hours of recording, not in bytes. SanDisk rates its 64 GB High Endurance card at 5,000 hours and its 512 GB card at 40,000 hours, and states those figures are for Full HD, with fewer hours for 4K.

How long does a dash cam memory card last?

It depends on your duty cycle. At one hour of driving a day, 5,000 hours is about 14 years; recording continuously while parked, the same card is about 7 months.

Can you check how much life is left on a dash cam card?

No. Consumer cards carry no endurance rating at all, and SanDisk's own white paper states the standard has no health-monitoring register for remaining life.

Is a dash cam memory card a consumable?

Yes. It is the only part of a dash cam installation that is expected to wear out from normal use, and it should be budgeted like one.

Notes on sources

Every figure above comes from one of the four below, read on 1 October 2026. Two are SanDisk’s own documents — the High Endurance datasheet and the endurance white paper — a manufacturer source with a commercial interest in card sales, used for its stated ratings and conditions. The speed-class rules come from the SD Association’s own documentation of its standards. The duty-cycle table is our arithmetic from the published 5,000-hour rating: a comparison of duty cycles, not a prediction of when a particular card fails.

• The 2,500 / 5,000 / 10,000 / 20,000 / 40,000 hour endurance ratings by capacity; that the rating is stated for Full HD only and that total hours are less for 4K UHD; the Class 10, U3 and V30 ratings; the stated 30 MB/s sustained video capture rate; the up to 100 MB/s read and up to 40 MB/s write figures; and the operating temperature range of −25 °C to 85 °C: the SanDisk High Endurance datasheet. This is a manufacturer source with a commercial interest in card sales; we have used its stated ratings and conditions, not its product claims.
• That consumer cards typically do not include specified write endurance ratings; that endurance for recording cards is expressed in hours at a specific bit rate rather than in total bytes written; that consumer-grade products do not include a defined health monitoring register, because this is neither specified by the SD Association nor commonly implemented; that remaining lifetime is estimated by tracking operating hours and written data; and that the bill of materials including controller, firmware and NAND can change without prior notice: SanDisk’s published endurance white paper. These are SanDisk’s own statements about the limits of its consumer products.
• That there are four separate speed class systems; that the class symbols must be matched between host and card; that expected write speed will not be available from a combination of different class symbols even where they indicate the same nominal speed; the Class 10 with U1 and U1 with V10 examples; that speed class write performance is defined for sequential writes to a completely un-fragmented area; that fragmentation slows writing as a property of flash memory; that the host may be unable to perform speed class recording even when free space exists; and that reformatting is a documented remedy: the SD Association’s published documentation of its speed class standards. This is the standards body for the format.
• The years and months in the duty-cycle table, and the roughly 78 hours per gigabyte figure: our arithmetic from the published 5,000-hour rating for a 64 GB card. The table assumes the rating is consumed linearly and that the recording rate matches whatever rate the rating was based on. It is a comparison of duty cycles, not a prediction of a specific card’s date of failure.

Three things we could not establish and are not going to imply: no published bit rate behind the endurance rating, so the hours cannot be converted to bytes; no published factor for how much the rating drops at 4K, only the statement that it does; and no way to read a card’s remaining endurance, which SanDisk attributes to the absence of a health-monitoring register in the standard.