Published, independent testing found Gen5 SSDs drawing meaningfully more power than Gen4 (6.74W vs 3.97-4.87W in one measured test) and throttling hard - in one case to roughly a third of heatsinked write speed - without adequate cooling. Manufacturers including Corsair document a heatsink requirement for their Gen5 drives. Whether a pre-fitted drive or a bare drive plus separate heatsink costs less changes with the market - we compute both live below rather than assume either.
Updated · measurements cited from publications dated 2023, checked 16 September 2026
"Numbers taken by Tom's Hardware show the T700 averaging 6.74 W during a 50GB file folder copy, with the PCIe 4.0 drives tested averaging 3.97 to 4.87 W, depending on the model."
Citing Tom's Hardware's own DiskBench 50GB-copy power measurements on Crucial's T700 Gen5 engineering sample against several Gen4 drives. Same piece: "Reviewers agreed that the T700 requires some sort of heatsink for sustained workloads," citing Linus Tech Tips measuring the drive reaching 67°C with its heatsink fitted after 15 minutes at 100% disk usage.
Read the source →"Crucial's PCIe 5.0 SSD Throttles to HDD Speeds Without Cooler." Recorded average write performance after 15 minutes on the bare (uncooled) drive: 1,027 MB/s, down from 3,681 MB/s on the same drive with its heatsink fitted - dipping as low as roughly 100 MB/s at points during the test.
Crucial T700 Gen5 SSD tested bare versus with its factory heatsink, sustained write workload. Corroborated by German publication ComputerBase, cited in the same piece, which found the same drive thermal-throttling in CrystalDiskMark without a heatsink.
Read the source →"According to testing data from both Phoronix and TechPowerUp, if you don't take measures to properly cool down the [Corsair] MP700, it can start generating file system errors or completely shut down your PC."
Reporting on independent testing by Phoronix (file-system errors within under three minutes uncooled) and TechPowerUp (thermal shutdown within under two minutes uncooled) on a Corsair MP700 Gen5 SSD. Same piece notes a heatsink-equipped Gigabyte Aorus Gen5 1000 "never once threatened to throttle during multiple benchmark runs" in PCWorld's own review.
Read the source →"Consumers [should] heed Corsair's listed requirements and put an M.2 heatsink or cooler on the MP700, as the drive can experience uncontrolled thermal shutdowns if pushed hard enough without one."
Reporting Phison's own statement acknowledging the issue on its PS5026-E26 controller (used in the Corsair MP700 and other Gen5 drives): "After carefully reviewing the recent reports from TechPowerUp and Phoronix, Phison would like to acknowledge the issue found in the reviews of products using the new Phison PS5026-E26 controller. We take this matter seriously and are committed to resolving it promptly."
Read the source →"This is why Gen5 SSDs require a heatsink for proper operation, at least for the time being... Gen5 drives are designed to use their own heatsink, high-end third-party heatsinks, or motherboard M.2 heatsinks."
Sabrent's own manufacturer guidance, also citing Corsair's documented MP700 heatsink requirement ("Corsair lists a heatsink as a requirement to avoid 'thermal shutdowns' when 'pushed hard enough'") via Tom's Hardware's reporting above.
Read the source →Tom's Hardware's reporting above names Corsair's own documented requirement directly: a heatsink is listed as necessary for the MP700 to avoid uncontrolled thermal shutdown under sustained load. Sabrent's own guidance, cited above, describes the general pattern across the category: "Gen5 drives are designed to use their own heatsink, high-end third-party heatsinks, or motherboard M.2 heatsinks."
Whether a pre-fitted drive costs more or less than assembling the same thing yourself - a bare drive plus a separate heatsink - is a live price question with no fixed answer, so we compute it rather than assert it. This is the same machinery our PS5 SSD heatsink page uses, applied to the same tracked bare-drive-and-standalone-heatsink population - the mechanism is identical for any M.2 drive, PS5-bound or not.
| Capacity | Bare drive + heatsink (assembled) | Pre-fitted drive | Cheaper path |
|---|---|---|---|
| 1TB | $217.71 (Samsung + heatsink) | $152.99 (Fikwot) | Pre-fitted saves $64.72 |
| 2TB | $336.59 (Crucial + heatsink) | $219.50 (Corsair) | Pre-fitted saves $117.09 |
| 4TB | $534.08 (Samsung + heatsink) | $476.98 (Fikwot) | Pre-fitted saves $57.10 |
| 8TB | $1504.08 (Lexar + heatsink) | $1337.65 (WD) | Pre-fitted saves $166.43 |
Computed live, capacity by capacity, from the cheapest in-stock bare drive plus the cheapest in-stock standalone heatsink, against the cheapest in-stock pre-fitted drive at the same capacity.
Whether your motherboard's own M.2 heatsink shroud can be used instead of a drive-specific one is a physical-fit question, not a thermal-performance question we can answer generally - board M.2 heatsink designs vary in size, thickness allowance and contact-pad placement between models. Your board's own manual documents its M.2 heatsink's dimensions and which drive lengths and thicknesses it accepts; that document is the authority here, the same way it is the authority for lane sharing on the Gen4-vs-Gen5 page.
Every figure on this page - the 6.74W measurement, the throttled write speeds, the reported errors and shutdowns - is a named publication's own test result, attributed with its date and test context. We have not run a thermal test of our own on any drive this cluster covers, and we make no claim about how hot a specific drive will run in your specific case, with your specific airflow, under your specific workload. What we can do is point at the pattern these independent tests found - Gen5 draws more power, and without adequate cooling that power has documented consequences - and let the manufacturer requirements and the live pricing above inform the actual purchase.
If you buy a Gen5 drive, budget for cooling it properly - either a factory-fitted heatsink or a compatible motherboard M.2 shroud confirmed against your board's own manual. The published testing above found real consequences for skipping that step, from throttled speeds to reported shutdowns, and none of those risks are ones this cluster is willing to wave away with a "probably fine" on a component a game library depends on staying intact.
Gen4 vs Gen5, the full case · DirectStorage explained · PS5 SSD heatsinks, live prices · Back to the gaming SSD hub
Published testing says the risk is real without one. PCWorld, citing independent testing from Phoronix and TechPowerUp, reported uncooled Gen5 drives generating file-system errors or shutting down entirely. Tom's Hardware separately measured a Gen5 drive's sustained write speed dropping to roughly a third of its heatsinked figure without cooling.
Ars Technica, citing Tom's Hardware's own measurements, reported a Gen5 drive averaging 6.74W during a sustained 50GB copy test, against 3.97-4.87W for the Gen4 drives tested alongside it - a real, attributed, measured difference, not a maker spec.
It varies and flips - see the live comparison below. A pre-fitted drive is sometimes cheaper than buying a bare drive plus a separate heatsink; sometimes the reverse is true. We compute both live from current stock rather than assume either answer.
Whether your board's M.2 heatsink shroud fits your specific drive length and thickness is a physical-fit question your motherboard's own manual answers, not something we can generalize - board M.2 heatsink designs vary considerably between models. Check your board manual before assuming compatibility.
We make no performance claim either way - see the doctrine note above. What is documented is the opposite risk: testing cited on this page found uncooled Gen5 drives throttling or erroring under sustained load, which a heatsink is documented to prevent.
The published testing on this page is specific to Gen5 power and throttling behavior - we have not cited an equivalent Gen4 measurement, and are not extending the Gen5 findings to Gen4 by assumption. Manufacturer heatsink recommendations, where stated, apply per-model regardless of generation.