How Often Do SD Cards Really Fail?

There's been enough dissections on how Canon and Nikon shot themselves in their respective feet by releasing mirrorless systems with only single card slots. Trust Tony and Chelsea Northrup, though, to spice things up a bit with some scientific analysis.

I know that this has been debated ad-nauseam, but I haven't actually seen any proper data on just how often the unspeakable happens. If anyone is a little tight of purse or just on the fence about it all, it helps to have solid figures to guide one in the right direction.   

The Northrups conducted a poll among their followers asking them their experiences with memory card failure rates. The poll, consisting of 4,344 people, while not a giant sample size, still seems like a size which one can pull some relevant data out of. And that's one of the things that I think this channel is best at. I always enjoy the Northrup's videos, but I find the more analytical ones to be almost always enlightening with regard to the more technical aspects of photography.

The arguments have gone back and forth over this, even on this website, but it's this kind of analysis that might actually sway people instead of one-off horror stories. I won't spoil the results by repeating them here, so please watch the video and share your thoughts below.

Log in or register to post comments

59 Comments

Previous comments
user-55486's picture

As a professional wedding photographer/videographer I will NEVER buy another camera unless it has 2 card slots. There are NO do overs in the wedding biz. My reputation is on the line here. Oops, my card failed is not an option. I love being able to tell my brides that both of my a7iii's have 2 card slots. This gives her peace of mind and me too!

And you use only one camera? I hope you do not shoot a wedding with only one camera.

Daniel Medley's picture

You do realize that two cameras is no where near the same thing as having redundancy in card slots, right? Especially for a wedding photographer. Do you think they tote a couple of cameras around to take the same pics on each one?

You do realize that if you have only one camera and it fails you can’t take any photos, right?
Anything can fail; cards, cameras, lenses, flashes, batteries…
Highest risk: the photographer (human error), so you need at least a second shooter if “failure is not an option”.

Daniel Medley's picture

So now you're talking about a second shooter. OK. Generally, as I understand it, a second shooter is not necessarily for redundancy in the case of card failure (again, the strawman argumen), but to get different/simultaneous shots. Though I'm sure a second shooter would come in quite useful in a case of lost data on a primary camera.

Many wedding photographers have a second camera on them; not for redundancy, but usually equipped with a different lens. It's not a redundancy thing. And, like I pointed out, a second camera does not mitigate losing data. Lost data on a camera's card is still lost data.

At the end of the day, a redundant slot in a camera is vastly better than not having it.

No, I’m talking about failure not being an option.
You have to look at all the risks. If card failure is the highest risk of not getting the shot(s), it makes sense to address it first, but I don’t think it is the highest risk. To have redundancy for small risks is fine, but if you haven’t addressed higher risks, failure will still happen.

Daniel Medley's picture

You're obviously either not reading or unable to keep up. It's not the risk of of failure. It's the consequences of loss; either through failure, or more likely, human error. And the effort/cost of meaningful mitigation. Losing a card with a thousand images on it can be substantially consequential. That consequence can be greatly mitigated by a simple secondary slot. Not saying it's a guarantee, but it's greatly mitigated. For most people it will be enough. For others, they may want to take extra steps depending on their circumstances. The ratio of cost for a secondary slot to possible mitigation is such that not having a secondary slot is just stupid.

Kind of like having a spare tire. When's the last time you had a flat? I've NEVER had a flat tire while driving. But it's common sense to have a spare tire because the consequences of one of your tires going flat can be greatly mitigated by having a spare. No, it's not a 100 percent guarantee that you'll not still be stranded, but combined with the fact that tires rarely go flat, it greatly mitigates it.

I’m not reading or unable to keep up…
That’s what people say if they lack real arguments.

One spare tire will not reduce the risk of getting a flat, it will reduce the risk of getting stuck in the middle of nowhere (the consequence in your example). More than one tire can fail, so failure is still an option, but the chance of double failure Is way less, so we take the risk. Why not have two spare tires, or four, to reduce the risk even more?

Do you also have a spare car battery? Probably not. If your battery fails the consequences might be the same (getting stuck in the middle of nowhere). A car battery is not that big and not really expensive, so why do we all have a spare tire but no spare battery?
Talking about cars… Out of every 100,000 licensed drivers 16 die in a car accident. Low risk, high consequence. We all take the risk if we get into a car.

»The ratio of cost for a secondary slot to possible mitigation is such that not having a secondary slot is just stupid.«

Both Canon and Nikon have cameras (DSLR) with two card slots, so they know how to do it. The first two generations of Sony’s A7 series had only one card slot. If there was a simple low-cost solution, I think it would be there right from the start. It might not be that simple.

Daniel Medley's picture

I have real arguments. It's just that you're not cognitively able to understand them. Then you blame me for lacking real arguments.

That accusation speaks more of you than anyone else.

The last word seems to be important to you, so I'll let you have it and leave this circular nonsense behind.

Have a good day.

Saying you have real arguments without providing them followed by insults and childish assumptions…
Really funny how some people project their own inability to have a grownup conversation onto others.

user-55486's picture

Did you read my post? Here is what I said, "I love being able to tell my brides that BOTH of my a7iii's have 2 card slots." So to answer your question, no, I use two cameras. Yes, they are BOTH on me at all times during the wedding day. I have the Tamron 28-75 on my left and the Sony 70-200 on my right. In a matter of seconds I can get a wide, medium and tight shot. This leads to more images being sold.

dale clark's picture

I find it funny nobody read your original post all the way thru. It's so obvious you use 2 a7iii's. Amazing how trolls can't even read a small paragraph properly

Burt Johnson's picture

TOTAL NONSENSE! This is not even remotely useful "data."

First off, the participants self-selected themselves, and those (I'd bet a month's mortgage payment) a highly disproportionate of those who had failures responded, while an even larger disproportionately number of people who did not have failures did not bother to respond to such a nonsense survey.

Second -- Of those who say they had failures, I'd bet yet another mortgage payment that many NEVER actually had a failure. I know way too many people who have complained of card failures when it was really user error. (And one who complained that his card failed after the camera went into a fountain...)

I could probably come up with even more reasons this is pseudo-science, but those two already make the whole thing a joke, so I won't bother to continue.

SHAME ON FSTOPPERS for republishing something so misleading. You guys are better than this!

Allan Savage's picture

I disagree Burt. I used to work for the Australian Bureau of Statistics AND I have a double degree in Computer Science/Mathematics, so I'm qualified to speak on sample sizes. A few points:

1. 4300 is a pretty reasonable sample - if gathered and processed properly it would result in a highly reliable analysis.

2. Northrup himself is very careful to acknowledge that some of the stats may not be entirely supportable.

3. If you discard all else there is still one incontrovertible fact that comes from this sample: memory cards DO fail, just as disk drives do fail, and it has nothing to do with user error. That's not to say that sometimes it is attributable to user error, but it's totally illogical to assert that ALL card failures are attributable to that factor.

4. If it is possible for a card to spontaneously fail, even if the failure rate is < 1%, then it is madness not to take steps to mitigate the potential loss due to card failure.

It is a basic engineering probability problem. You should be able to compute the results. Two slots of a inferior SD memory device which fails at a higher rate than a XQD card, at some failure point, will be less reliable than a single higher reliability XQD. What is that number and did Nikon get it right?

»1. 4300 is a pretty reasonable sample - if gathered and processed properly it would result in a highly reliable analysis«

How about a “defined procedure” to determine the group?
Are Northrup followers responding to a pole an accurate representation?

»If you discard all else there is still one incontrovertible fact that comes from this sample: memory cards DO fail«

Everything will fail eventually. It’s stating the obvious.

»4. If it is possible for a card to spontaneously fail, even if the failure rate is < 1%, then it is madness not to take steps to mitigate the potential loss due to card failure. «

Depends on the real number.
1 in 200 is less than 1%
1 in 1,000,000,000 is also less than 1%
Do you think in both cases it’s “madness not to take steps to mitigate the potential loss due to card failure”?

Allan Savage's picture

"Do you think in both cases it’s “madness not to take steps to mitigate the potential loss due to card failure”?"

Yes. I'm also a paraglider and was once a rockclimber. I am risk-averse to a single point of failure.

So a single point of failure with a 1:1,000,000,000 change of failure is a risk you don’t want to take, but paragliding and rock-climbing are ok?
http://www.besthealthdegrees.com/health-risks/

Motti Bembaron's picture

You would lose two monthly mortgage payment.

I lost two cards in three years, no user error, they just stopped working. Data was lost and could not reformat them either. Gone.

Both were from Lexar.

JetCity Ninja's picture

fortunately the Nikons don't use SD cards.

more importantly, polls, in and of themselves, are not scientific. anyone with the faintest grasp on STEM and has viewed a single Northrups video is well aware that facts and science have little to no importance to them.

Roberto Pavan's picture

Why is he (or anyone) lumping the Nikon Zs in with the Canon R here? The Nikon Zs use XQD cards, which are far more robust, both physically and in the bus protocol, and are known to fail far, far, FAR less often than SD cards. In fact, Thom Hogan did an informal survey among his colleagues and couldn't find a single instance of an XQD card failure.

Sorry,, that is not the way it works. Two of an inferior SD cards, which have a higher failure rate, if that failure rate is significantly higher than say a XQD card, at some failure rate, will be less reliable than a single XQD card. It is a basic engineering probability problem. An example is airplanes. It used to be that 4 engines were required to fly across the Atlantic. Now, 2 engine airplanes fly across because their 2 engines are just as or more reliable than the previous 4 engine planes.

Mark James's picture

They can use two now because they can still fly the plane with one, which is the requirement (that it stays in the air). In the old days with the old tech the planes couldn't fly with only one engine due to power and design limitations. Jet engines have always been reliable, because they have to be.

Daniel Medley's picture

Absolutely correct. People make it far more complicated than it is; failure rates, some cards are more reliable than others, etc. It's not relevant.

The fact is that having a backup is better than no backup. Period. Card failure isn't the ONLY thing that can result in loss of data.

Jerome Brill's picture

General probability, regardless of your opinion or experience will state having a backup is better than not having one. Use a single card slot, I don't care. Just don't argue against it. Your reason doesn't matter.

I remember my very first wedding (an event that takes around 1-2 hours in total with a peak of critical events lasting for maximum 10 minutes) shooting with a 6D as main camera and the card inside it got corrupted. Oh boy, the chills... Fortunately, my heavy electronics & computer science background kicked in and after 2 long and dizzying days I was able to recover all of the pictures (and more) from the SD card, but that event taught me a very good lesson.
It's not a novelty to get your card corrupted because a filthy data stream was unexpectedly still active at disconnect (the same reason why USB sticks get corrupted sometimes when you don't "safely remove hardware") or god knows what other reasons. But if you expect to shoot events that cannot simply be re-shot, that additional slot is something to consider if you expect to build up some reputation.

Emmet Adriaans's picture

Lexar cards are pure rubbish.

Mike Stern's picture

It’s amazing. Tony recognizes hateful commenters about his SD card failure opinions and 1 or 2 slot cameras, then quickly makes a new video addressing the issue yet shutting up those who claim the otherwise.
He is efficient, quick and very good on his job. Congratulations to Tony.

Mike Stern's picture

Here is some obvious precautions we all can take:
1) Do not take in and out your SD card often. Prefer to use usb cable and keep the card in as much as you can.
2) If you must take the card out, do not use it on multiple devices. Multiple laptops, iPads through adapters.
3) Dedicate SD card per camera. Try not to use the same card in multiple cameras. Especially multiple brands and type of cameras.
4) Do buy 2 card slot cameras. At least for the important shoots configure the slots so you have a back up.
5) Do demand the camera manufacturers to install inbody memory. Either to use for primery or secondary recording. Anybody has heard of a disk failure on iPhones? Less likely.