Quick Answer

  • According to Backblaze data center reports, the annualized failure rate (AFR) of HDDs averages in the 1-2 percent band, with mechanical wear being the dominant factor.
  • For SSDs, AFR starts around 0.5-1 percent in the first year and rises over time with NAND wear, but it plateaus for a long stretch.
  • The JEDEC JESD218 standard defines a 1-year data retention guarantee for consumer SSDs and 3 months for enterprise SSDs.
  • For the Samsung 870 EVO 1TB, the manufacturer TBW value is 600 TBW, with a DWPD of roughly 0.3, meaning 300 GB of daily writes lasts 5 years.
  • In real use, a service life of 5-10 years is expected for both HDDs and SSDs, and the 3-2-1 backup rule is essential for critical data.

The answer to the question "is the HDD more durable, or the SSD" depends on the situation, but we can now speak with real field data rather than guesswork. In its quarterly Drive Stats reports, Backblaze shares the real failure rates of more than 250,000 drives. JEDEC, meanwhile, sets official endurance metrics for SSDs with the JESD218 standard. In this guide we compare both technologies with lab data, manufacturer datasheets and DSET's 20 years of field experience.

If you have already experienced a failure on your computer, first check our main guide, Data Recovery Guide 2026 Turkey.

The Fundamental Difference Between HDD and SSD

An HDD, that is a Hard Disk Drive, is a mechanical device consisting of spinning magnetic platters and read-write heads that fly 5 to 9 nanometers above those platters. The platter spins at 5400, 7200, or up to 15000 RPM in enterprise models. The head motor (VCM) changes position hundreds of times per second. Therefore an HDD is highly exposed to physical wear. Drops, vibration, voltage fluctuations and temperature changes disrupt that nanometer gap between the head and the platter.

An SSD, that is a Solid State Drive, contains no moving parts. Data is trapped as electrons in NAND flash cells. Each write applies high voltage to the cell, so NAND has a limited write cycle. As cell density increases from SLC, MLC, TLC, QLC and in recent years PLC, the lifespan per unit drops. In return, capacity and cost improve.

So the simple summary: an HDD dies from mechanical wear, an SSD dies from electrical fatigue. One with time, the other with the amount written.

What Are MTBF and AFR, and Why We Trust Manufacturer Numbers Less

MTBF (Mean Time Between Failures), which we often see on manufacturer datasheets, is a statistical estimate. For example, 1,500,000 hours MTBF means "if we run 1000 of this drive model, on average one will fail every 1500 hours." It does not mean a single drive lasts 171 years.

In the real world, Backblaze uses AFR (Annualized Failure Rate), that is the annual failure percentage. This metric is not a lab simulation, it is the real statistic of a drive running in a data center. There is often a serious gap between the manufacturer MTBF and the Backblaze AFR. That is why field data is always more valuable to us.

Table 1, HDD Annualized Failure Rates by Brand (Backblaze Drive Stats)

Taken from the Backblaze 2023 year-end Drive Stats report, the annualized failure rate of the most common enterprise HDD models:

Manufacturer Model Capacity Drive Count AFR (annual %)
HGST HUH721212ALN604 12 TB 10,685 0.40
HGST HUH721212ALE600 12 TB 2,601 0.79
Toshiba MG07ACA14TA 14 TB 38,214 1.02
Toshiba MG08ACA16TEY 16 TB 5,345 0.69
Seagate ST12000NM0008 12 TB 19,879 1.38
Seagate ST14000NM001G 14 TB 11,840 0.73
Seagate ST16000NM001G 16 TB 21,687 0.69
WDC WUH721414ALE6L4 14 TB 8,422 0.46
WDC WUH721816ALE6L0 16 TB 4,072 0.11
Total fleet average 251,485 1.46

Comment: HGST (now part of WDC) and the new WDC Ultrastar series are at the bottom. The Seagate 12 TB class historically showed high AFR, while the 14 and 16 TB generation recovered. These figures are data center conditions. A home user's drive sits in worse conditions, so the field AFR can be 1.5 to 2 times these numbers.

Source, Backblaze Drive Stats.

Table 2, SSD AFR by Brand (Backblaze SSD Edition Reports)

Since 2018 Backblaze has also used SSDs heavily as boot drives and publishes SSD Edition Drive Stats. From the 2023 year-end report:

Manufacturer Model Capacity Drive Count AFR (annual %)
Seagate ZA250CM10003 250 GB 1,099 0.90
Seagate ZA250CM10002 250 GB 547 0.73
Seagate ZA2000CM10002 2 TB 158 0.00
Crucial CT250MX500SSD1 250 GB 506 0.99
Dell DELLBOSS VD 480 GB 376 0.00
Samsung MZ7LH240HAHQ 240 GB 152 0.00
WDC WDS250G2B0A 250 GB 1,212 0.75
Toshiba KXG60ZNV256G 256 GB 47 0.00
SSD fleet total average 3,144 0.90

Comment: In the Backblaze fleet the average annualized failure rate of SSDs comes out lower than HDDs (roughly 0.90 vs 1.46). But this figure can be misleading, because the SSDs have not yet run as long in the fleet as the HDDs. Backblaze's own comment: "SSDs are safer in the short term, but the failure curve rises in the long term as the TBW limit approaches." So the SSD death curve differs from the HDD, it starts later but hits harder.

For the most frequently received models and typical failure patterns in the DSET lab, you can read our article Frequently Failing SSD Models in Turkey.

What Is TBW, the JEDEC JESD218 Standard

TBW (Terabytes Written) is the manufacturer's declaration of "I warrant this SSD until this many terabytes are written." This figure is not arbitrary, it is set according to the JEDEC JESD218 standard. The standard says, "A consumer SSD must operate under the JESD219 workload, at 30°C ambient and 40°C active temperature, with a guaranteed 1-year shelf life." For enterprise SSDs the shelf life is 3 months but the write intensity is tested more aggressively.

The important consequence of JESD218 for us: when the TBW warranty is exhausted, the SSD does not die suddenly. But the data retention period shortens. If you power off an SSD whose write capacity is exhausted and leave it on the shelf for 6 months, the chance of seeing bit errors increases significantly. It is the natural result of NAND's electrical structure.

For a detailed practical guide on TBW calculation, see TBW Analysis, SSD Write Endurance, JEDEC Standard and Calculation.

Table 3, TBW Declaration of Popular SSD Models (Manufacturer Datasheet)

Brand Model Type Capacity TBW (manufacturer) Warranty
Samsung 870 EVO SATA TLC 1 TB 600 TBW 5 years
Samsung 870 EVO SATA TLC 2 TB 1,200 TBW 5 years
Samsung 980 Pro NVMe TLC 1 TB 600 TBW 5 years
Samsung 980 Pro NVMe TLC 2 TB 1,200 TBW 5 years
Samsung 990 Pro NVMe TLC 1 TB 600 TBW 5 years
Samsung 990 Pro NVMe TLC 2 TB 1,200 TBW 5 years
Crucial MX500 SATA TLC 1 TB 360 TBW 5 years
Crucial MX500 SATA TLC 2 TB 700 TBW 5 years
Crucial P5 Plus NVMe TLC 1 TB 600 TBW 5 years
Crucial P5 Plus NVMe TLC 2 TB 1,200 TBW 5 years
Kingston KC3000 NVMe TLC 1 TB 800 TBW 5 years
Kingston KC3000 NVMe TLC 2 TB 1,600 TBW 5 years
WD SN850X NVMe TLC 1 TB 600 TBW 5 years
WD SN850X NVMe TLC 2 TB 1,200 TBW 5 years

Comment: The industry average is roughly 600 TBW per TB. The Kingston KC3000 stands out here (800 TBW per TB). The Crucial MX500 SATA side is at the bottom with 360 TBW per TB, due to the older 64-layer NAND generation. The newer 176 and 232-layer NAND (Micron B47R, Samsung V7) withstands more program-erase cycles per cell.

For the Samsung-specific warranty process and failure patterns, we have our article Samsung SSD Data Recovery Process and Risks.

DWPD Calculation, How Many Years an SSD Lasts

DWPD (Drive Writes Per Day) is the manufacturer saying "you can rewrite the entire SSD this many times every day." The calculation is simple, DWPD = TBW / (Capacity × Warranty × 365).

Table 4, Average Expected Lifespan Scenarios

The table below shows how many years an SSD will last before the warranty expires, based on the daily write amount. A 1 TB Samsung 980 Pro (600 TBW) is used as the reference.

Daily Writes User Profile Expected Lifespan (years)
10 GB / day Light office, invoicing, email 164 years (NAND retention limit comes first)
30 GB / day Normal desktop, gaming 54 years
50 GB / day Developer, IDE, compilation 32 years
100 GB / day Content creator, 4K editing cache 16 years
300 GB / day Video studio, local AI training 5.4 years
600 GB / day Data center-like workload 2.7 years
Enterprise 3 DWPD Samsung PM9A3 class 5-year warranty
Enterprise 10 DWPD Optane / SLC cache class 5-year warranty

Conclusion: The average home user (20-30 GB daily writes) will practically never reach the TBW limit of a consumer SSD. The SSD most likely says goodbye first because it ages out or due to a controller failure. In the enterprise class, the 3-10 DWPD range is designed to handle 5 years of heavy 24/7 writing.

Cases Where the HDD Is Still Advantageous

Do SSDs win in every scenario? No. There are three areas where the HDD is unquestionably superior:

1) Cost per TB. As of 2026, a 14 TB Toshiba MG09 is around 250 USD, that is 18 USD per TB. An SSD of the same capacity (16 TB enterprise NVMe) is over 1,500 USD, about 95 USD per TB. For a backup archive, the HDD is still economically overwhelming.

2) Cold storage (archival). When an HDD is powered off, the data lasts 10 years or more magnetically. The SSD powered-off shelf life is stated by JEDEC as 1 year for consumer, 3 months for enterprise. So leaving archival data on an SSD is risky.

3) Forensic recoverability. Even if an HDD breaks down, reading data from the platter is possible, the sectors physically remain there. When an SSD controller dies, the NAND chips must be desoldered (chip-off) and logically reassembled, which is far more expensive and process-heavy. Our article Chip-off NAND Extraction focuses on this topic.

Scenarios Where the SSD Is Unquestionably Superior

1) Performance. NVMe Gen4 SSDs can do 7000 MB/s sequential reads, while even the fastest HDD cannot exceed 280 MB/s. The random IOPS difference is over 1000x.

2) Mechanical shock resistance. A laptop HDD dropped while running has over an 80% chance of head-platter contact, while this risk is zero in an SSD. In roughly two-thirds of the laptop cases that come to DSET, when the drive is an HDD, head damage is found.

3) Silence and energy. An SSD draws 2-3 W when active, an HDD 7-9 W. At data center scale this difference means thousands of dollars in annual energy savings.

4) TRIM and garbage collection. Modern SSDs perform cell balancing (wear leveling) in the background, so that a single block dying early is prevented. There is no such "self-protection" mechanism in an HDD.

The Hybrid Approach, the Right Architecture

In professional setups today, an HDD alone or an SSD alone is rarely preferred. The hybrid architecture we recommend:

  • System and applications: 1-2 TB NVMe SSD (e.g. Samsung 990 Pro). Speed comes from here.
  • Frequently accessed data (projects, cache, virtual machines): SATA SSD or a second NVMe.
  • Cold archive, photo backups, video masters: 8-16 TB enterprise HDD (Toshiba MG, WD Ultrastar). 7200 RPM, helium-filled.
  • NAS / central backup: SHR or RAID 6 on a Synology or QNAP, SSD cache + HDD data layer.

For Synology-specific crash scenarios and SHR behavior, the article Synology NAS Crashed, Data Recovery SHR contains practical information. If a RAID 5 has crashed, in our article RAID 5 Crashed, Recovery Process and Cost we explained what should and should not be done.

The 3-2-1 Backup Rule, a Mix of HDD and SSD

The classic 3-2-1 strategy is 3 copies, 2 different media, 1 offsite. Our recommendation:

  • Copy 1 (active work): On an NVMe SSD.
  • Copy 2 (local backup): HDD in a NAS, RAID 6 or SHR-2.
  • Copy 3 (offsite): Cloud (Backblaze B2, AWS S3 Glacier) or an external HDD kept on a shelf. Do not leave an external SSD on the shelf for more than 12 months, NAND data retention does not allow it.

If you want to make a local external backup an SSD, plug it into the computer every 6 months and power it on for a few minutes, the controller will refresh the cells.

Backblaze Field Data vs Manufacturer Declaration, the Real Gap

While the manufacturer says 1,500,000 hours MTBF, Backblaze reports an annual 0.4-1.5 % AFR for the same drives in a real data center. The gap between the theoretical AFR ~0.58 % calculated from the MTBF formula and Backblaze's 0.69-1.38 % real AFR figures shows that manufacturer tests are done in a controlled environment. In the real world, the following three factors disrupt MTBF:

  1. Temperature fluctuation. In the Turkish summer, when the server room hits 28°C, HDD AFR rises linearly.
  2. Voltage quality. Sudden spikes in the city grid hit the SSD controller, while brownouts hit the HDD motor driver.
  3. Mechanical shock. A workstation being moved, a laptop dropped on a desk, an external drive thrown around in cargo.

At DSET we read the WD Reliability Report and Seagate FARM (Field Accessible Reliability Metrics) logs before SMART in every recovery case. FARM data holds start and stop counts and head load/unload cycles, so we see the drive's real fatigue level far more accurately than SMART.

DSET Field Experience, Behavior Under Turkish Conditions

Qualitative observations drawn from our 20 years of laboratory experience:

  • Seagate Barracuda 2 TB ST2000DM006, the home user's most common HDD. Head preamp failure is common, it arrives with the "click of death" after 3-4 years.
  • WD Blue 4 TB, more stable but experiences firmware lock, PCB transfer is not enough, ROM swap is required.
  • Toshiba MG enterprise series, the least failing drive in Turkish data centers. Helium leak field cases are almost nonexistent.
  • Samsung 870 EVO SATA, sudden disappearance cases due to controller firmware bugs (2021-2022 batch). In these cases the SSD does not show in the BIOS, while the NAND is healthy, and data is read via chip-off.
  • Crucial MX500, the "uncorrectable error count" SMART error is common, but data is mostly readable.
  • Kingston KC3000 and WD SN850X, we have not yet seen a large failure wave, these models are relatively new.

In all of these cases we apply technological intervention with PC-3000 SSD edition, in most of them PCB repair under the microscope, and a NAND chip-off procedure when needed. The 12 lessons we learned over 20 years are gathered neatly in our article Data Recovery Anatomy, 20 Years of Field Experience.

When data destruction is required, we comply with the NIST SP 800-88 Rev. 1 standard, for SSDs cryptographic erase + ATA Sanitize as the "purge" procedure, for HDDs Secure Erase + degauss. For security reasons, we do not deliver drives to be resold without passing them through this procedure.

Frequently Asked Questions

1) Which lasts longer, HDD or SSD? It depends on the use. In a low write volume + frequent power on/off scenario, the SSD lives longer. In a high write + continuous operation scenario, the enterprise HDD offers a more predictable lifespan. In the Backblaze fleet data, the average AFR over the last 3 years favors the SSD.

2) What happens when an SSD's TBW is exhausted? It does not die immediately. Write performance drops, the cell error rate increases, and the data retention period shortens significantly. If you leave a full SSD whose TBW is exhausted on the shelf for 6 months, the chance of loss is high.

3) Should I buy an SSD or HDD external drive? If you will carry it, take it around in a bag, plug/unplug frequently, choose an SSD. If you will back up and put it in a cabinet, choose an HDD. An SSD is not suitable for data that will stay on the shelf for a long time.

4) Should I put an HDD or SSD in a NAS? An HDD is economical for the data layer. If performance is needed, systems like Synology and QNAP support SSD cache. An SSD-only NAS rarely makes sense outside of video editing or heavy databases.

5) How do I use an HDD for a long life? Stable temperature (between 25-35 °C), a quality power supply, no physical shaking, sequential write patterns are preferred. Do not let it constantly enter and exit sleep, and keep idle spin-down from being aggressive.

6) How do I use an SSD for a long life? Do not keep it more than 80% full (leave room for over-provisioning), keep TRIM on, do not defrag frequently (an HDD habit harms the SSD), and keep firmware up to date.

7) What does Backblaze data mean for a home user? It does not apply directly because data center conditions are ideal. In home conditions you can assume the AFR figures are 1.5-2 times higher. Still, the relative ranking between brand and model remains valid.

8) Should I avoid throwing out my old HDD without destroying it? Do not throw it out. Without secure erase + physical destruction per the NIST SP 800-88 procedure, it is possible to read data from the drive. DSET performs certified data destruction.

9) Is the Samsung 870 EVO firmware bug still valid? Samsung released a firmware update for the 2021 batches, but the number of users who have not applied the update is very high. Cases of this model still come to DSET weekly.

10) Which lasts longer, NVMe SSD or SATA SSD? If the NAND is the same type, the lifespan is similar. NVMe runs hotter and requires a heatsink. An overheating controller creates premature failure.

Conclusion and DSET's Recommendation

An HDD dies, by mechanical wear. An SSD dies, by electrical fatigue. Neither is "infinite." The right strategy is to place both in the same system, divide their duties, and strictly apply the 3-2-1 backup rule. Look at Backblaze field data rather than the manufacturer declaration, and compare TBW and DWPD numbers with your daily write habits.

If there is currently a sign of failure on your device (noise, freezing, not showing in BIOS, SMART error), do not keep writing. Every additional write reduces the recoverable area. Power off the device and get expert support.

DSET Hacettepe Teknokent Ankara Data Recovery Laboratory, free discovery and diagnostics service. Phone, +90 536 662 38 09

In our article Main guide: Data Recovery Guide 2026 Turkey, the recovery process, pricing, and what should be done in which situation are explained from start to finish.

Sources

  1. Backblaze Drive Stats, https://www.backblaze.com/cloud-storage/resources/hard-drive-test-data
  2. Backblaze SSD Edition Quarterly Reports, https://www.backblaze.com/blog/ssd-edition-2023-mid-year-drive-stats-review/
  3. JEDEC JESD218 Solid-State Drive Requirements and Endurance Test Method, https://www.jedec.org
  4. JEDEC JESD219 SSD Endurance Workloads
  5. Samsung Semiconductor SSD Product Specifications, https://semiconductor.samsung.com/
  6. Crucial MX500 and P5 Plus Datasheet, https://www.crucial.com/
  7. Western Digital Reliability Report and Ultrastar Datasheet
  8. Seagate FARM (Field Accessible Reliability Metrics) Technical Brief
  9. Kingston KC3000 Product Datasheet
  10. NIST Special Publication 800-88 Rev. 1, Guidelines for Media Sanitization