I would like to see someone kill an SD card in "minutes." My camera shoots HD video at some stupidly high bit rate and stores it onto my SD card at damn near as fast as the card will take it, and I do this a lot. There are thousands of people who do it even more than I do, and it can take months or years to go through a card.
> (April 17th) We are currently testing a sample of 40 4GB microSD cards from Sandisk to failure
> Its May now and still no cards have failed.
> we were testing a card from a customer yesterday on a TS-7553 and it failed the DoubleStore stress tests within the day. [...] The test failed at approximately 13 GB of raw data written
> The ATP card failed a 3rd time after 65GB of data written.
> Another, separate ATP card failed after 877 GB of write activity
If the card is MLC and there's no or limited wear leveling, repeated local writes will shut the card dead quickly ("limited" as in wear leveling being applied only among e.g a 4MB group of cells, Sandisk happens to do that). I've also heard of some controllers implementing wear leveling, muddling the test results.
Some cards are truly reliable, others are craptastic, and everything in between is possible. The problem is you can't trust a brand and go for it, and getting the technical details, if at all possible, involves digging into cryptic and hard to find documentations.
I can't find the link again to a test of someone who tested various card brands and had them die semi-reliably within 15 minutes.
I'm sure longevity is good for taking photos. I bet it wouldn't be so great for compiles on a dev box. Of course, no one would ever do that. It would be just too slow.
