Why a Dash Cam Kills Memory Cards in Summer

This page contains affiliate links. If you buy through one we may earn a commission, at no cost to you.

Summer is the clue and the high endurance part is the red herring. Endurance describes how many times a cell can be erased and rewritten, and at a realistic dash cam write rate that allowance takes years to exhaust. What fails in July and not in January is a temperature problem, and temperature attacks a different property entirely: how long a cell holds its charge once written.

There is also an arithmetic result worth having before any of that, because it changes what to buy. Program and erase cycles per year are simply the bytes written per year divided by the capacity of the card. So capacity and endurance are the same lever, one for one, and capacity is the number that is always printed while endurance never is. A 512 GB ordinary card outlasts a 64 GB high endurance card.

Every rating on a card is a speed floor

Start with what the marks mean, because none of them is about the failure in question. A Kingston Canvas Go Plus 256GB microSD card is described in its record as carrying "U3 and V30 ratings [which] ensure consistent performance", along with an A2 classification. Those are three statements about speed and nothing else.

U3 means a minimum sustained sequential write of 30 megabytes a second. V30 means a minimum sustained sequential write of 30 megabytes a second. They are the same promise in two notations, which is why cards print both and why nobody notices that one of the two marks is redundant. A2 is a random operations per second floor, aimed at running applications off the card. UHS-I and UHS-II are bus interfaces with theoretical ceilings of 104 and 312 megabytes a second. The Speed Class number is another sequential write minimum.

Now the demand. A 1080p dash cam running at 15 megabits a second writes 1.875 megabytes a second. Against a V30 guarantee of 30, that is 6.25 percent of the floor, a headroom of 16 times. A 4K camera at 60 megabits writes 7.5 megabytes a second, which is still only a quarter of the guarantee.

So every rating on the card describes a quantity the application never comes close to needing, which is worth saying because of what our own page recommends. Our dash cam coverage advises planning for "a high endurance card rated for continuous recording to avoid dropouts". The first half of that is the subject of the next section and the second half is a category error: dropouts are a write speed failure, and endurance has nothing to do with them. A card that drops frames is failing its V30 floor. A card that is out of endurance does not drop frames, it stops accepting writes. Our microSD card roundup gets this right, treating "a string of ratings like u3, v30 and a2" as the speed marks they are.

Endurance is capacity, and capacity is published

Here is the part that makes the purchase decision. Flash wears out by program and erase cycles on each block, and a card's controller spreads writes across all of it. So the number of times the whole card has been cycled in a year is the total written in that year divided by how big it is.

At 1.875 megabytes a second, a camera writes 6.75 gigabytes an hour. Two hours of driving a day is 13.5 gigabytes a day, which is 4.93 terabytes a year. Leave parking mode recording around the clock and it is 162 gigabytes a day, 59.1 terabytes a year.

Divide by capacity and the cycles follow. In parking mode, a 64 gigabyte card is cycled 924 times a year, a 256 is cycled 231 times, and a 512 is cycled 115.5 times. Consumer flash is generally good for somewhere between 500 and 1,000 cycles, and a part sold as high endurance is typically rated nearer 3,000.

So, with parking mode running: a 64 gigabyte ordinary card lasts between six and thirteen months. A 64 gigabyte high endurance card lasts 3.25 years. A 256 gigabyte ordinary card lasts 2.2 to 4.3 years. And a Samsung PRO Plus 512GB microSDXC lasts 4.3 to 8.7 years, which is longer than the high endurance card at a quarter the capacity, for the simple reason that there is eight times as much silicon to spread the same writes across.

That card's record has an irony in it. Its drawbacks note that the "Speed class (U3/V30) not specified", so the one card in our catalogue whose capacity makes it the strongest endurance choice is the one that does not print the only class it carries. And its 180 megabytes a second read with 130 write is far beyond anything a dash cam asks of it, which is the pattern of the whole category: generous on the published number, silent on the governing one.

The conclusion is worth stating plainly because it inverts the usual advice. Buying four times the capacity buys exactly four times the endurance. Buying a high endurance label buys perhaps three to six times the cycles, and the label is not a number, and no card on this site publishes one. Our memory storage guide suggests considering "cards with a higher endurance rating", which is sensible and unactionable, because there is no rating to compare.

The part that fails in summer

None of the above explains a seasonal failure, and seasonal is what the question describes. The mechanism is retention rather than endurance, and it is the same Arrhenius law that turns up all over this site in different disguises.

A flash cell stores a few hundred electrons on a floating gate behind an insulating oxide. They leak out, slowly, and the leak is thermally activated with an activation energy somewhere near 1.1 electron volts. Put that through the Arrhenius expression and the retention time shortens by a factor of 3.84 for every ten kelvin, by 136 times going from 30 degrees to 70, and by 391 times going from 30 to 80.

Three hundred and ninety one. A card that would hold its contents for a year on a shelf at 30 degrees holds them for less than a day at 80.

The same law, with different numbers, drives our page on what a 100,000 hour LED rating actually promises, where the degrading thing is a phosphor, and it drives the cooking rate arithmetic in our page on why steam cannot get hotter than 100 degrees, where the same exponential is the reason a sealed pot is quick. Three different products, one relationship, and in this case it is working against you.

How hot the card actually gets

So the question is whether a dash cam reaches those temperatures, and it does, computably.

A summer cabin behind closed glass sits in the fifties Celsius, and the air near the windscreen is the hottest part of it. Call it 55 degrees. The camera adds its own heat: a body of roughly 70 by 50 by 30 millimetres has 142 square centimetres of surface, and in the still air behind a windscreen the effective coefficient for convection plus radiation is something like 12 watts per square metre kelvin. Two and a half watts of electronics then sits 14.7 kelvin above ambient, which puts the card at about 70.

Then put the sun on it. Direct solar radiation is around 800 watts per square metre, the camera presents perhaps 35 square centimetres to it, and a dark plastic case absorbs most of what lands. That is 2.24 watts, which is comparable to the entire power consumption of the camera, so it very nearly doubles the self heating and takes the total rise to 27.8 kelvin. Eighty three degrees at the card.

The usual operating ceiling for a microSD card is 85 degrees. So a camera in full sun on a hot day is operating its card at the edge of its specification, and a dark dashboard surface or a car parked facing the afternoon sun comfortably exceeds it.

And here is the gap. Not one card listing in our catalogue publishes an operating temperature range. One of them says so itself, which is more useful than the rest: a SanDisk 128GB Extreme PRO SD card lists among its drawbacks "No specific temperature or waterproof ratings provided". That is precisely the figure this question needs, identified as missing in our own record, on a card otherwise described in terms of 250 megabytes a second of read speed that no dash cam will ever use.

The durability language that does appear is worth reading carefully too. A SanDisk 256GB Extreme microSD card with adapter is described as robust enough "to withstand harsh conditions", and its own drawbacks are straight about what that means: "Durability claims are based on SanDisk testing, not independent." Those claims are generally about water, shock, X-rays and magnetic fields, which are the hazards a card in a pocket faces. None of them is sustained operation at 80 degrees, which is the hazard a card glued to a windscreen faces.

Why the card dies rather than just losing clips

One more piece, because it explains the severity. If retention were the whole story you would expect corrupted video and nothing worse, and in fact the video is the one thing that does not care. Loop recording overwrites every file within hours, so a clip only has to survive until the next pass. A retention time of a day is ample for the footage.

What does not get rewritten is the card's own bookkeeping. A flash controller keeps a flash translation layer, a set of tables mapping logical blocks to physical ones along with wear counts and bad block lists, and those are written rarely and must stay readable. Lose enough of that and the controller cannot find anything, which is why a heat damaged card does not come back as a card with some bad files. It comes back as read only, or as zero bytes, or not at all.

There is also an awkward asymmetry in flash physics that makes a hot camera the worst case on both counts. Cells program more reliably when hot and retain worse when hot, so the controller writes the tables with high confidence into cells that will lose them fastest, and it has no way of knowing the difference.

The honest limit

Three soft numbers here and it is worth being clear which. The activation energy of 1.1 electron volts is representative for NAND charge loss rather than measured on any particular part, and it varies with process and with how worn the cell already is, so treat the 391 as an order of magnitude and not a figure. The 500 to 1,000 and 3,000 cycle endurances are typical values for consumer and high endurance flash, published by nobody. And the 15 megabit bitrate is an ordinary dash cam figure that no camera listing we carry states.

What is firm is the structure. Cycles per year is bytes divided by capacity, which is arithmetic, so the capacity for endurance substitution holds exactly whatever the cycle rating turns out to be. The temperature calculation uses the camera's own power and the solar constant and nothing proprietary. And the distinction between a speed mark and an endurance property is a fact about the specifications rather than an estimate.

What to do, and the three free moves

First and largest: turn parking mode off, or set it to trigger on impact or motion rather than recording continuously. That is a twelvefold cut in the write load, and it takes a 64 gigabyte ordinary card from six to thirteen months up to six and a half to thirteen years. No card on the market offers an improvement of that size. If you need parking surveillance, accept that you have chosen to consume a card and plan on replacing it rather than on finding one that survives.

Second: buy capacity rather than a label. It is the same lever, it is always published, and it comes with the side benefit of a longer loop before footage is overwritten.

Third: get the camera out of direct sun. The solar gain was very nearly as large as the camera's own dissipation, so simply mounting behind the rear view mirror, in the shadow it casts, removes roughly half the temperature rise. In Arrhenius terms, 13 kelvin is a factor of about five on the retention clock, for a mount position that is free.

And one diagnostic, also free, which settles the question the asker was really asking. Endurance failure is gradual and aseasonal: writes slow down, the camera starts reporting errors on particular files, and eventually the card goes read only, at any time of year. Thermal failure is abrupt and seasonal, and it happens to cards that were working perfectly in April. Two cards killed in one summer is the second pattern, and no amount of endurance rating addresses it.