Used enterprise drive reliability report
Across 172 tested units of 26 models since Jun 2026, 163 passed our full test (94.8%), median power-on hours 25,170.
Updated continuously from our test database; last test Nov 1, 2026.
Download the data as JSON · CC BY 4.0
Pass rate by model
| Model | Brand | Capacity | Interface | Tested | Passed | Pass % | Median hours | Hours range |
|---|---|---|---|---|---|---|---|---|
| ST18000NM000JExos X18 | Seagate | 18TB | SATA | 28 | 28 | 100% | 25,568 | 10,033 to 36,505 |
| ST10000NM002GExos X16 | Seagate | 10TB | SAS | 21 | 21 | 100% | 2,356 | 2,353 to 3,464 |
| WUH721818AL4200Ultrastar DC HC550 | Western Digital | 18TB | SAS | 19 | 19 | 100% | 12,784 | 12,762 to 12,882 |
| WUH721414AL5204Ultrastar DC HC530 | Western Digital | 14TB | SAS | 18 | 18 | 100% | 25,170 | 22,235 to 44,137 |
| ST16000NM002GExos X16 | Seagate | 16TB | SAS | 16 | 16 | 100% | 27,178 | 51 to 27,284 |
| ST14000NM001GExos X16 | Seagate | 14TB | SATA | 10 | 10 | 100% | 17,542 | 16,789 to 18,319 |
| ST8000DM004BarraCuda | Seagate | 8TB | SATA | 7 | 7 | 100% | 3,040 | 2,754 to 3,097 |
| DL2400MM0159Exos | Seagate | 2.4TB | SAS | 6 | 6 | 100% | 40,748 | 39,926 to 40,785 |
| ST12000NM0127Exos | Seagate | 12TB | SATA | 6 | 6 | 100% | 52,310 | 43,679 to 55,885 |
| MTFDDAK1T9TDC5200 ECO | Micron | 1.92TB | SATA | 4 | 4 | 100% | 14,380 | 14,341 to 20,524 |
| MTFDDAK1T9TDS5300 PRO | Micron | 1.92TB | SATA | 4 | 4 | 100% | 20,628 | 18,681 to 22,028 |
| ST12000VN0008IronWolf | Seagate | 12TB | SATA | 4 | 2 | 50% | 10,740 | 10,491 to 10,988 |
| WUH721818AL5204Ultrastar DC HC550 | Western Digital | n/a | n/a | 4 | 0 | 0% | n/a | n/a |
| AL15SEB24EQYAL15SE | Toshiba | 2.4TB | SAS | 3 | 3 | 100% | 41,772 | 41,772 to 41,949 |
| HUC101818CS4204Ultrastar C10K1800 | HGST | 1.8TB | SAS | 3 | 2 | 66.7% | 21,238 | 19,047 to 21,316 |
| HUH721010AL42C0Ultrastar He10 | HGST | 10TB | SAS | 3 | 3 | 100% | 40,444 | 40,444 to 41,281 |
| ST24000DM001BarraCuda | Seagate | 24TB | SATA | 3 | 3 | 100% | 5,927 | 5,924 to 6,478 |
| WD181KRYZWD Gold | Western Digital | 18TB | SATA | 3 | 3 | 100% | 48,415 | 47,965 to 48,723 |
Small samples: 8 models with fewer than 3 tested units (10 units)+
With one or two units, a single result swings the percentage between 0% and 100%, so these rows are kept apart from the main table. They are included in every total on this page.
| Model | Brand | Capacity | Interface | Tested | Passed | Pass % | Median hours | Hours range |
|---|---|---|---|---|---|---|---|---|
| H7240AS60SUN4.0TUltrastar 7K4000 (Sun/Oracle OEM) | HGST | 4TB | SAS | 2 | 2 | 100% | 19,098 | 17,799 to 20,397 |
| ST4000NM0023 | Seagate | 4TB | SAS | 2 | 0 | 0% | 2,610 | 2,610 to 2,611 |
| CS900CS900 | PNY | 1TB | SATA | 1 | 1 | 100% | 4,474 | 4,474 to 4,474 |
| MK001920GWUGKRead Intensive SATA SSD | HPE | 1.92TB | SATA | 1 | 1 | 100% | 20,673 | 20,673 to 20,673 |
| ST1000NM0008Exos 7E2 | Seagate | 1TB | SATA | 1 | 1 | 100% | 19,901 | 19,901 to 19,901 |
| ST12000NM0027Exos X12 | Seagate | 12TB | SAS | 1 | 1 | 100% | 33,802 | 33,802 to 33,802 |
| ST16000NE000-2RW103 | Seagate | 16TB | SATA | 1 | 1 | 100% | 188 | 188 to 188 |
| ST16000NM000JExos X18 | Seagate | 16TB | SATA | 1 | 1 | 100% | 35,746 | 35,746 to 35,746 |
Pass % is passed units over tested units. Hours are SMART power-on hours at intake. Each model links to its page with the published unit reports.
Pass rate by brand
| Brand | Models | Tested | Passed | Pass % |
|---|---|---|---|---|
| Seagate | 14 | 107 | 103 | 96.3% |
| Western Digital | 4 | 44 | 40 | 90.9% |
| HGST | 3 | 8 | 7 | 87.5% |
| Micron | 2 | 8 | 8 | 100% |
| Toshiba | 1 | 3 | 3 | 100% |
| HPE | 1 | 1 | 1 | 100% |
| PNY | 1 | 1 | 1 | 100% |
Pass rate by interface
| Interface | Models | Tested | Passed | Pass % |
|---|---|---|---|---|
| SAS | 11 | 94 | 91 | 96.8% |
| SATA | 14 | 74 | 72 | 97.3% |
| Not recorded | 1 | 4 | 0 | 0% |
Power-on hours distribution
- Open Box under 100 hrs4 units · 2.5%
- Like New 100-500 hrs1 unit · 0.6%
- Light Use under 20,000 hrs79 units · 50%
- Standard Use 20,000-40,000 hrs53 units · 33.5%
- Value over 40,000 hrs21 units · 13.3%
Units by usage tier, from their SMART power-on hours. How tiers work
Failure reasons
No failure reasons recorded yet.
Why failed units failed, from their SMART and self-test results. A unit can have more than one reason.
Sanitization methods used
- Sanitize overwrite70 units · 42.7%
- ATA secure erase46 units · 28%
- SCSI sanitize22 units · 13.4%
- Crypto erase20 units · 12.2%
- sg_format5 units · 3%
- Block erase1 unit · 0.6%
Units by the erase method recorded with them. Every method is followed by a read-back check. Data destruction
Throughput by model
- CS900read 526 MB/s · write n/a · 1 unit
- MK001920GWUGKread 516 MB/s · write 480 MB/s · 1 unit
- MTFDDAK1T9TDSread 494 MB/s · write 427 MB/s · 4 units
- MTFDDAK1T9TDCread 491 MB/s · write 457 MB/s · 4 units
- ST12000VN0008read 283 MB/s · write 284 MB/s · 4 units
- ST16000NE000-2RW103read 276 MB/s · write 275 MB/s · 1 unit
- ST18000NM000Jread 275 MB/s · write 276 MB/s · 28 units
- DL2400MM0159read 274 MB/s · write 274 MB/s · 6 units
- ST16000NM000Jread 269 MB/s · write 271 MB/s · 1 unit
- WUH721818AL4200read 269 MB/s · write 143 MB/s · 19 units
- ST16000NM002Gread 267 MB/s · write 264 MB/s · 16 units
- AL15SEB24EQYread 266 MB/s · write 265 MB/s · 3 units
- ST14000NM001Gread 262 MB/s · write 263 MB/s · 10 units
- WD181KRYZread 261 MB/s · write 274 MB/s · 3 units
- WUH721414AL5204read 260 MB/s · write 263 MB/s · 18 units
- ST24000DM001read 256 MB/s · write n/a · 3 units
- ST12000NM0127read 254 MB/s · write 251 MB/s · 6 units
- ST10000NM002Gread 253 MB/s · write 254 MB/s · 21 units
- ST12000NM0027read 251 MB/s · write 245 MB/s · 1 unit
- HUC101818CS4204read 235 MB/s · write 236 MB/s · 3 units
- ST1000NM0008read 203 MB/s · write 204 MB/s · 1 unit
- ST4000NM0023read 193 MB/s · write 189 MB/s · 2 units
- ST8000DM004read 185 MB/s · write 182 MB/s · 7 units
- H7240AS60SUN4.0Tread 182 MB/s · write 175 MB/s · 2 units
Median sequential read (dark bar) and write (light bar) from fio, per model, across the units benchmarked.
Methodology
A unit counts as tested once it has been through our bench: a SMART health read, the drive's extended (long) self-test, and an fio throughput benchmark where applicable. It counts as passed when its overall result is pass, using the same criteria for every usage tier, listed on how we test.
The units are datacenter pulls that we buy in lots, so the numbers describe used drives at intake, not new drives and not drives in service. Hours are SMART power-on hours, grouped into usage tiers. The overall median is the unit-weighted median of each model's median. Erase methods are covered on data destruction.
Many models have only a few units, so read a single model's percentage with care; models with fewer than 3 tested units are listed separately. Each unit's report is public at /r/<unit id>, the address on its tamper seal, and is linked from its model page.
Cite this report
Atomic Racks, Used enterprise drive reliability report, atomicracks.com/reliability, retrieved Nov 13, 2026
The data is licensed CC BY 4.0: reuse it with attribution to Atomic Racks and a link to this page. Machine-readable numbers: /reliability.json.
Questions about this report
What counts as a pass in this report?+
A unit passes when its overall result in our test database is pass: SMART health reports PASSED, there are no pending or uncorrectable sectors and no medium errors, the extended self-test completes, and the fio benchmark lands in the expected range for the model. Units that fail are pulled from sale.
Is the pass rate the same as an annualized failure rate?+
No. An annualized failure rate counts drives that fail while in service over a year. This report counts used drives that pass or fail our intake test after they leave a datacenter. It describes the condition of pulled drives on arrival, not how long a drive will last in your system.
How current are the numbers?+
Each model page carries statistics synced from our test database, and this page reads them on every request with at most a five-minute cache. The last test date at the top of the page is the most recent unit test across all models.
Can I use these numbers in my own article or tool?+
Yes. The data is published under CC BY 4.0: you can copy, chart and republish it with attribution to Atomic Racks and a link to this page. The same aggregates are available as JSON at /reliability.json for embedding.