Every drive tracked here is listed with its recording technology. CMR is safe for RAID, ZFS, and NAS; SMR is not. We currently track 313 CMR drives, 8 SMR, and 30 unverified. Search your exact model below to check it.
Once you have picked a CMR drive, the price forecast covers whether to buy it now.
Seagate's BarraCuda line is the clearest case for reading this list by model rather than by brand. Seagate's own published recording-technology list names the 3.5-inch BarraCuda 4TB and 8TB as SMR and the 12TB, 16TB, 20TB and 24TB as CMR, so the largest drives in the budget line are conventional recording, which is the opposite of the blanket “BarraCuda is SMR” claim most coverage makes. The 1TB, 2TB, 3TB and 6TB capacities appear under neither heading, so those models sit in the Unknown group below rather than being assigned a technology by inference. Every in-stock BarraCuda with its per-model badge and live price is on the BarraCuda hub.
WD Red is the line this entire subject is named after, and it is also the line where the shorthand people have learned is wrong. The rule circulating in forums is that the EFAX suffix means shingled recording. Western Digital's own recording-technology list names only the 2TB, 3TB, 4TB and 6TB WD Red drives as shingled and states that all other capacity points use conventional recording, which makes WD80EFAX, WD100EFAX and WD120EFAX conventional drives carrying the suffix everyone treats as a warning.
The tiers above plain Red are cleaner. WD's current Red Plus data sheet states conventional recording across its range, and so does the current Red Pro data sheet, from the 2TB WD2002FFSX to the 26TB WD260KFGX. The shingled models live in the plain WD Red tier, without Plus or Pro in the name, and they are printed as SMR in WD's own current product brief with a footnote naming the implementation as device-managed.
Western Digital published that model-level list in 2020 alongside a blog post in which it stated that sustained random writes during a ZFS resilver leave device-managed shingled drives no idle time to execute internal data management tasks, resulting in significantly lower performance, and recommended its conventionally-recorded Red drives instead. That is the manufacturer describing the failure mode in its own words, which is why we cite it rather than a community write-up.
| Suffix | Recording | The rule |
|---|---|---|
| EFRX | CMR | Conventional at every capacity we have verified with this suffix. |
| EFAX | SMRCMR | Shingled at 2TB, 3TB, 4TB, 6TB; conventional at every other capacity carrying this suffix. |
| EFZX | CMR | Conventional at every capacity we have verified with this suffix. |
| EFZZ | CMR | Conventional at every capacity we have verified with this suffix. |
| EFPX | CMR | Conventional at every capacity we have verified with this suffix. |
| EFGX | CMR | Conventional at every capacity we have verified with this suffix. |
| EFBX | CMR | Conventional at every capacity we have verified with this suffix. |
| FFSX | CMR | Conventional at every capacity we have verified with this suffix. |
| FFWX | CMR | Conventional at every capacity we have verified with this suffix. |
| FFBX | CMR | Conventional at every capacity we have verified with this suffix. |
| KFBX | CMR | Conventional at every capacity we have verified with this suffix. |
| KFGX | CMR | Conventional at every capacity we have verified with this suffix. |
Built from the models we have verified against WD documentation. The full decoder, with capacities, examples and the source on each row, is on the WD Red hub.
The reason the suffix cannot be the final answer is that the marketing name is not reliable either: we have seen retail listings label a plain-Red part number as Red Plus. The full model number on the drive label is the ground truth. Every tier, every in-stock listing and the full decoder with sources sit on the WD Red hub, and Red Plus against Red Pro computes what the tier upgrade costs at every capacity where both are in stock.
Both surveillance lines are in this list, and it is the place the per-model answer matters most. A recorder writes continuously by definition and, once the disk fills, overwrites the oldest footage in a permanent loop. That is close to the worst-case pattern for shingled recording, and unlike a desktop drive it never gets an idle period to catch up.
Seagate's own SkyHawk datasheet states SMR against ST4000VX013 in the same column block where the other 4TB part, ST4000VX007, is stated as CMR, and Seagate's SkyHawk SATA product manual repeats it for that generation including ST2000VX015. Seagate also publishes a recording-technology list stated by capacity rather than by part number, and that list shows no SMR anywhere in the SkyHawk line. We do not resolve the disagreement between two of Seagate's own publications; we badge the part numbers the per-model documents name and link all three sources on the SkyHawk hub.
Western Digital is cleaner but incomplete. The current WD Purple data sheet and the Purple Pro brief state CMR explicitly for the models they cover. The previous-generation WD Purple data sheet prints no recording-technology row at all, so those part numbers carry no technology claim from us and sit in the Unknown group below rather than being assigned one by inference from their siblings.
This list answers by drive. Our exact-model resolver answers by part number and names the manufacturer document behind every answer, which matters because within one family at one capacity the answer frequently goes both ways. A Seagate BarraCuda 1TB is conventionally recorded as an ST1000DM010 and shingled as an ST1000DM014, under the same product name.
Filter by recording technology and search your exact model. This is the "is my drive CMR?" lookup. CMR drives are safe for arrays; the list defaults to CMR, in-stock. Out-of-stock drives show "Check availability" rather than a price, and are never ranked on $/TB.
Conventional Magnetic Recording writes data tracks side by side on the platter without overlapping them. Because each track stands alone, the drive can rewrite any single track without disturbing its neighbours, which gives CMR its defining property: predictable, consistent random-write performance under any workload. Every enterprise nearline drive and every NAS-rated drive uses CMR (or ePMR, an enhanced but still conventional, non-shingled variant), because datacenter and array workloads cannot tolerate the write penalties of shingling. When you buy an enterprise Exos, Ultrastar, or a NAS IronWolf or Red Plus, you are buying CMR by design.
Shingled Magnetic Recording overlaps adjacent tracks like the shingles on a roof, packing more capacity onto the same platter. The catch is that overlapping tracks cannot be rewritten independently: changing data on one track forces the drive to read and rewrite every track in the overlapping band. This is write amplification, and it is where SMR's problems begin.
Under the sustained random-write load of a RAID rebuild or a ZFS resilver, that amplification is catastrophic. Rebuilds that should take hours stretch into days, dangerously extending the window during which a second drive failure would destroy the entire array. ZFS resilvers can slow to a crawl or fail to complete at all. Drives can even drop out of an array under load when the controller times out waiting for the drive to finish its internal rewrites. This is why SMR is disqualifying for RAID and ZFS: it is not a performance preference you can tune around, it is a reliability failure mode. For any array, CMR only.
To verify a drive you already own or are about to buy, look up its exact model number, the full part number, not the marketing family name, since recording technology varies within a family. The manufacturer's spec sheet lists the recording method definitively. If you cannot find documentation, there is a behavioural tell: write a large file (bigger than the drive's CMR cache, typically tens of gigabytes) and watch the sustained write speed. A CMR drive holds a steady rate; an SMR drive's throughput collapses once the cache fills and it must start shingling in place. The searchable list above already carries the recording method for every model we track.
The most notorious incident is the 2020 WD Red controversy. WD shipped SMR in some plain WD Red (non-Plus) drives without disclosing it, and ZFS users discovered the problem the hard way when resilvers failed. WD's resolution clarified the lineup: WD Red Plus and WD Red Pro are CMR, while plain WD Red is SMR at 2TB, 3TB, 4TB and 6TB per WD's own current product brief. The WD Red section above carries the suffix decoder and the sources. Seagate has similarly shipped SMR in some Barracuda desktop drives at several capacities. We state these facts as documented by the manufacturers and the community reporting at the time, and we deliberately do not assert specific capacities we have not verified on a current spec sheet or in our own database. Recording technology can change between revisions, so always confirm the exact model before you buy.
For NAS and RAID, use CMR only, with no exceptions. See the NAS HDD tracker, the TrueNAS picks, and the Plex server guide, all of which are CMR-focused. For a single-drive backup or archive target where writes are sequential and there is no array to rebuild, SMR is acceptable and its extra capacity per dollar can even be an advantage. The distinction is entirely about whether the drive lives in an array: in an array, shingling is a liability; standing alone doing sequential writes, it is fine.
Glossary · NAS HDD · TrueNAS drives · Plex drives · SATA HDD · Refurbished enterprise
Look up the exact model number, not the marketing family name, on the manufacturer's spec sheet, which lists the recording technology. On this page, every tracked drive is labeled with its recording method, so you can search your model directly. A behavioural tell also works: SMR drives show sustained write speed collapsing once their CMR cache fills during a long write, while CMR drives hold a steady rate.
For RAID, ZFS, and NAS use, CMR is unambiguously better: it delivers predictable random-write performance and rebuilds reliably. SMR trades that away for slightly higher capacity per platter, which only makes sense for single-drive, mostly-sequential workloads like an archive or a backup target. For any array, CMR is the correct and safe choice.
SMR overlaps data tracks like roof shingles, so rewriting one track forces the drive to rewrite the whole overlapping band, which is massive write amplification. Under the sustained random writes of a RAID rebuild or ZFS resilver, that causes rebuilds to take days instead of hours (extending the window in which a second failure destroys the array), resilvers that may never complete, and drives dropping out of the array under load. It is a reliability problem, not a performance preference.
The 2TB, 3TB, 4TB and 6TB plain WD Red drives, per Western Digital's own recording-technology list and its current WD Red product brief, which prints SMR against those capacities with a footnote naming the implementation as device-managed. WD Red Plus and WD Red Pro are conventionally recorded across their current lines per WD's data sheets. The EFAX suffix on its own does not settle it: WD's list names only those smaller capacities, so WD80EFAX, WD100EFAX and WD120EFAX are CMR. Read the full model number, and see the WD Red section above for the decoder.
Effectively yes. Every enterprise nearline and datacenter drive family, including Seagate Exos, WD Ultrastar and Toshiba MG, uses CMR (or ePMR, a conventional non-shingled variant), because datacenter workloads demand predictable performance. This is a major reason refurbished enterprise drives are a safe RAID choice: they are CMR by design. Every enterprise HDD tracked here is CMR/ePMR.
Yes. Single-drive backup and archive is the one use case where SMR is acceptable, because those workloads are largely sequential writes that avoid SMR's worst-case rewrite penalty. Just never place an SMR drive in a RAID array, ZFS pool, or NAS parity role. Keep SMR to standalone, sequential, cold-storage duty.
Severely. ZFS resilver and scrub operations generate exactly the sustained random-write pattern that triggers SMR's worst-case behaviour, and SMR drives can make a resilver crawl for days or fail to complete, putting the pool at risk. The 2020 WD Red incident surfaced precisely because ZFS users hit resilver failures. Use CMR-only in any ZFS vdev.
No, they are orthogonal. HAMR (Heat-Assisted Magnetic Recording) is a method for writing bits more densely by briefly heating the platter, and it can be used with conventional non-shingled tracks. SMR (Shingled Magnetic Recording) is about overlapping tracks regardless of how each bit is written. A HAMR drive is not inherently SMR, and current high-capacity HAMR enterprise drives are CMR-class for RAID purposes.