It depends on the drive, the machine and the work: the later PCI Express generations run warmer and step back when they get too hot. Both the drive maker and the machine maker state what they expect, and clearance decides what can be fitted.
The answer is a clearance question as much as a thermal one, and both halves are published.
Why the question arises at all
The later PCI Express generations run faster and warmer, and drives on them step back when they get too hot rather than continuing at full rate. That behaviour is deliberate and protective.
It shows up as a long transfer that starts quickly and settles lower, and it recovers as the drive cools. Whether it happens depends on how long the work lasts, which is why the same drive can be untroubled in one machine and throttling in another.
What each maker says
The drive maker publishes an operating temperature range and states whether the drive ships with a heatsink or expects one. Some models are sold in both forms.
The machine or board maker states whether the socket has a heatsink over it, whether a thermal pad is fitted to the cover above, and what card thickness the socket accommodates. On a laptop this is the constraining answer.
Where both makers say a heatsink is expected, fit one. Where the machine already provides cooling over the socket, adding another one is a clearance problem rather than an improvement.
Clearance, which stops most of these conversations
A drive with a factory heatsink is still a card of its stated length and width, and it is now considerably taller. That extra height has to clear whatever is above the socket: a graphics card, a laptop lid, a drive tray in an enclosure, or the machine’s own cooling plate.
Single-sided and double-sided cards are the related question. A double-sided card carries flash on both faces and is thicker, and some laptop sockets have room only for the single-sided kind.
Neither of those is visible from a photograph. Both are published.
Where an M.2 heatsink genuinely matters
Sustained writing. Video work, large image sets, a machine used as a scratch volume, a drive receiving backups for hours.
Enclosed spaces with no moving air. A compact desktop, a machine in a cabinet, a portable enclosure sitting in the sun.
A socket under a graphics card, which is both hot and airless, and which several boards use for their fastest socket.
Where the drive stays cool on its own
Ordinary desktop use. Starting a machine, opening applications and working on documents produces bursts too short to warm a drive appreciably.
A drive behind a slower link. A card in an external enclosure or in a narrower socket is not working hard enough to reach the temperatures that prompt the question.
And in a portable enclosure
The enclosure’s own housing is the cooling, and makers vary enormously in how seriously they take it — some use the metal shell as a heatsink with a thermal pad onto the drive, others are a plastic box. A drive that throttles in an enclosure and not in a machine is telling you about the enclosure.
The enclosure’s host link is usually the ceiling anyway, so a drive that never reaches its own rate never produces its own heat either. That is the case where the question answers itself.
The questions that come up before an order
What does thermal throttling actually feel like?
A long copy that starts at one rate and settles at a lower one after a while, then recovers when the drive cools. It is the drive protecting itself, and it is not a fault. Short transfers and ordinary desktop use rarely reach the point where it happens at all.
Can I add a heatsink to a drive in a laptop?
Usually not, and this is where clearance decides. Many laptops have no room above the socket and some place a thermal pad on the underside of the cover instead, which is the machine maker’s own cooling arrangement. Fitting something the machine was not designed around risks the cover, the socket and the drive.
Last reviewed 10 September 2026