Data Recovery Case File · Solid State & Flash · A Card Doing an Unfair Job
Heat After Power Means Current Went Somewhere It Should Not
His enquiry contains a detail that changes the priority. A small single-board computer that "refuses to boot up from the 32GB card, which is now quite hot to touch. I ran some basic diagnostics" on two different systems. A card that is warm to the touch after being powered has been drawing current abnormally — and behind that lies a broader point about what these cards are being asked to do, because running an operating system is not the job they were built for.
| Media | 32GB SD card used as the system and boot medium for a single-board computer — no longer booting; abnormal thermal behaviour on power |
| Reported situation | Single-board computer disconnected from power · refusing to boot from its card afterwards · card noticeably warm following power application · diagnostics attempted on two separate systems without result · content required |
| Fault class | Controller fault with abnormal current draw — wear failure consistent with sustained small-write workload; further powering contraindicated |
| Equipment used | Card removed from use and not powered further · current draw measured on a controlled bench supply before any read attempt · read at chip level past the controller where draw permitted · translation layer reconstructed in software · files validated by opening |
The decode: the heat first, then the underlying reason
Why warmth matters here: a memory card working normally does not become noticeably warm. It has no moving parts and draws very little. Heat is current, and current going somewhere it should not means either a short within the package or a controller stuck in a fault state drawing continuously. Neither improves with further attempts, and both are reasons to stop powering it.
So the immediate instruction: take the card out and leave it out. Not into a reader to check, not back into the machine to try once more. Each power application on a card in this state is more current through whatever is wrong.
Now the larger point, because it explains why this happened. Memory cards are designed for a specific workload: comparatively large, sequential writes, at intervals — a camera writing photographs, a recorder writing video. Their controllers, their wear-levelling and their endurance ratings all assume that pattern.
What an operating system does instead: constant small writes, continuously, in scattered locations. Logs updating every few seconds. Databases rewriting index pages. Temporary files created and removed. Swap activity when memory runs short. That is the opposite of the workload the card was built for, and it wears one out far faster than anybody expects — a card rated for years of camera use can fail within months as a system drive.
Why this is the commonest failure in that whole category of device: not because the computers are unreliable, but because the storage medium is being asked to do a job it was never specified for. It is a design compromise rather than a defect, and it is entirely avoidable once known.
What the diagnostics on two systems established: that the fault travels with the card rather than belonging to the host. That is worth having and there is nothing further to try at that level.
What to do differently afterwards: boot from a solid-state drive over the machine's own interface where the design allows it. Where a card must be used, keep a written image of it — a complete copy taken while it works, from which a replacement card can be restored in minutes. That converts a card failure from a loss into an inconvenience.
On the bench
The card was removed from use and not powered further, warmth indicating current through a short or a controller drawing continuously in a fault state. Current draw was measured on a controlled bench supply before any read attempt, an abnormal draw requiring the fault to be characterised before power is applied through a reader. The memory was read at chip level past the controller where draw permitted, the translation layer reconstructed in software, and files validated by opening.
The outcome
The card taken out of use, draw measured before any read attempt and the memory recovered past the controller. Free assessment, one fixed written figure including VAT, 50% of parts and labour upfront with the balance only on successful recovery. The decode: a card working normally does not get warm, so heat means current going where it should not. And the underlying reason is workload — cards are built for large sequential writes at intervals, while an operating system does constant small scattered ones, which wears them out far faster than their rating suggests.
Memory card running a small computer that has failed
Take it out and stop powering it — a card that gets warm has been drawing current abnormally, which means either a short inside the package or a controller stuck drawing continuously, and neither improves with more attempts. Worth knowing why it failed, because it's avoidable. Cards are designed for large sequential writes at intervals, the way a camera writes photographs, and their controllers and endurance ratings all assume that. An operating system does the opposite: constant small writes in scattered locations, logs updating every few seconds, databases rewriting, swap activity. A card rated for years of camera use can fail in months doing that. Going forward, boot from a proper drive, or keep a written image of the card.
Stop powering it — call Guildford Data Recovery on 01483 901310; draw measured on a controlled bench supply before any read attempt, memory read at chip level past the controller, files checked by opening.
Request a quote online →
Our case files are drawn from genuine enquiries received by our laboratory over the past ten years, anonymised to protect client confidentiality. Each one describes the diagnostic and recovery procedure our engineers apply to that fault, using the equipment listed.