TBW Analysis: How Much Write Life Does Your SSD Have Left? JEDEC Standard and Calculation

When buying an SSD we look at capacity, speed, and maybe the warranty. But the real endurance indicator, the TBW value, is something most users have never opened. Yet exactly how much more writing an SSD can withstand is measured by this metric. In this article we explain what TBW is, the logic behind the JEDEC standard, the differences between manufacturers, how to read the real value via SMART, and the chances of data recovery on a drive whose TBW has run out, all from a field perspective.

What Is TBW? (Total Bytes Written)

TBW expresses the total amount of data that can be written to an SSD within its warranty period, measured in terabytes. NAND flash cells can withstand a limited number of program/erase (P/E) cycles: TLC NAND tolerates roughly 1000-3000, while QLC NAND is around 500-1000 cycles. The manufacturer publishes a guaranteed write budget by factoring in this cell limit, the efficiency of wear leveling, and the write amplification factor.

An important distinction: TBW does not count read operations. Even if you read 5 TB of data per week from an SSD, the cells do not wear out. Only writing (and indirectly erasing) consumes cell life. That is why write-heavy workloads such as video editing, compilation, virtual machines, and database logs burn through TBW quickly.

JEDEC JESD218 Standard

JEDEC (Joint Electron Device Engineering Council), the standards body of the semiconductor industry, defines SSD endurance through JESD218 and post-mortem data retention through the JESD219 document. Manufacturers do not write down TBW figures arbitrarily; they test against the acceptance criteria set by these two documents.

JESD218 defines two usage classes:

  • Client (Consumer) class: Office and gaming load, 8 hours of operation. After power loss, the drive must retain data for 1 year at 30 °C, and the guaranteed write load should typically be around 1 DWPD.
  • Enterprise class: 24/7 heavy writing. Data retention is 3 months at 40 °C, and the guaranteed load is generally at the 3-10 DWPD level.

For UBER (Uncorrectable Bit Error Rate) there is a worst-case limit of 10^-15 for client SSDs and 10^-16 for enterprise SSDs. When a manufacturer declares TBW, it accepts how much can be written without exceeding this UBER target.

DWPD and TBW Conversion

DWPD (Drive Writes Per Day) tells you how many times per day you can write the entire drive. The formula is simple:

TBW = Capacity (TB) × DWPD × Warranty period (days)

Example: a 1 TB consumer SSD, 1 DWPD, 5-year warranty, gives 1 × 1 × 1825 = 1825 TBW. In reality, however, manufacturers do not assign 1 DWPD to client drives, because consumers never write that much. That is why the TBW of 1 TB consumer SSDs is usually in the 300-700 band. On the enterprise side, a 1.92 TB drive with 3 DWPD reaches roughly 10500 TBW over 5 years, meaning it is 10 times more durable.

TBW Comparison Table for 6 Brands

The figures below are taken from the manufacturers' official datasheet pages.

1 TB models

Model NAND TBW Warranty
Samsung 980 Pro 1 TB TLC 600 TBW 5 years
Samsung 990 Pro 1 TB TLC 600 TBW 5 years
Crucial MX500 1 TB TLC 360 TBW 5 years
Crucial P5 Plus 1 TB TLC 600 TBW 5 years
WD Black SN850X 1 TB TLC 600 TBW 5 years
Kingston KC3000 1 TB TLC 800 TBW 5 years
Seagate FireCuda 530 1 TB TLC 1275 TBW 5 years
SK hynix Platinum P41 1 TB TLC 750 TBW 5 years

2 TB models

Model NAND TBW Warranty
Samsung 980 Pro 2 TB TLC 1200 TBW 5 years
Samsung 990 Pro 2 TB TLC 1200 TBW 5 years
Crucial MX500 2 TB TLC 700 TBW 5 years
Crucial P5 Plus 2 TB TLC 1200 TBW 5 years
WD Black SN850X 2 TB TLC 1200 TBW 5 years
Kingston KC3000 2 TB TLC 1600 TBW 5 years
Seagate FireCuda 530 2 TB TLC 2550 TBW 5 years
SK hynix Platinum P41 2 TB TLC 1200 TBW 5 years

The striking point: at the same capacity, there is a 3-4 fold TBW difference between the Crucial MX500 (mainstream SATA) and the Seagate FireCuda 530 (PCIe 4.0 flagship). This difference comes from the NAND choice, the over-provisioning ratio, and the quality of the controller's wear leveling.

Reading Your Real TBW via SMART

You can read how much of your SSD's TBW you have spent from the SMART (Self-Monitoring, Analysis and Reporting Technology) table. Manufacturers use different SMART IDs:

  • Samsung: 0xF1 (241) "Total LBAs Written". If we multiply the total LBA count by 512 and divide by 1024^4, we get TB. Samsung Magician shows it directly in TB under the "Drive Health" tab.
  • Crucial / Micron: 0xF6 (Total Host Sector Writes) and 0xF7 (Total Host Sector Reads). The Crucial Storage Executive interface gives a percentage via "Percentage Lifetime Used".
  • Kingston: 0xF1 Lifetime Writes. Kingston SSD Manager shows it under "SSD Wear Indicator".
  • WD / SanDisk: 0xFB (Total LBAs Written) and 0xF1-like variants. WD Dashboard presents a "Life Remaining" percentage.
  • General: 0xE9 "Media Wearout Indicator" is standard on Intel/Micron, 0xE7 "SSD Life Left" on Crucial, and 0xCA "Percentage Used Endurance" on enterprise NVMe.

If you do not have the manufacturer's tool, CrystalDiskInfo is an open-source alternative; it gives raw SMART values in hex and automatically converts the "Total Host Writes" line. For NVMe drives, the command nvme smart-log /dev/nvme0 with nvme-cli dumps the "Data Units Written" field; this value is in 1000 × 512 byte units, so you need to multiply by 1000 and divide by 1024^4.

A practical calculation: if the SMART 0xF1 raw value on your Samsung 980 Pro 1 TB drive shows 300000000000 sectors, then 300 billion × 512 bytes is approximately 140 TB written. This means about 23% of the 600 TBW limit has been consumed.

What Happens When TBW Runs Out?

Manufacturers apply different policies on this:

  1. Read-only mode: On some models, Samsung and Crucial put the drive into permanent read-only mode when TBW or the internal endurance counter drops below a critical threshold. Data can be read, but not a single additional byte can be written.
  2. Controller throttle: Some WD and Kingston series severely reduce write speed, and the warning is conveyed to the user via SMART flags.
  3. Spare block exhaustion: Wear leveling uses a pool of spare blocks to distribute the load across blocks. When this pool is exhausted, the controller starts incrementing "Reallocated Sector Count" (0x05), and the user experiences slowdowns and I/O errors.
  4. Catastrophic failure: On drives that also have high power-on hours, TBW exhaustion can combine with a capacitor fault and render the drive completely inaccessible; we covered these cases in detail in our article on the SSD data recovery process.

Important: TBW being exhausted means falling outside the warranty, but it does not mean the drive will immediately fail. In field observation, we have seen cases where a 600 TBW drive still works despite 900-1000 TB written; conversely, there are drives that die from a controller fault without even spending 200 TBW. In our article on the most frequently failing SSD models in Turkey we listed which series tend to "die before TBW".

SSD Life Calculation

If you know your daily writing habits, you can roughly estimate your remaining life. The formula:

Remaining Years = (TBW limit − Written so far) / (Daily write × 365)

Example scenarios:

  • Office user: Writes 10 GB per day. Starting from scratch with a Samsung 980 Pro 1 TB (600 TBW): 600000 / (10 × 365) is approximately 164 years. So TBW will never be a problem; the controller or capacitor will fail much earlier.
  • Software developer: With compilation, Docker images, and logs, writes 80 GB per day. A 600 TBW drive: 600000 / (80 × 365) is approximately 20 years. Still comfortable.
  • Video editor / 3D artist: Writes 300 GB of cache and proxy per day. A 600 TBW drive: 600000 / (300 × 365) is approximately 5.5 years. This matches the exact warranty period; it is not suitable for the workload, and a higher-TBW model (e.g. FireCuda 530) is needed.
  • Chia mining: 1.5 TB of plot writing per day. A 600 TBW drive lasts 13 months. A consumer SSD is absolutely not suitable for this workload.

Add the write amplification (WA) factor to your calculation. 1 GB written by the host lands on the NAND as 1.2-2.0 GB at the controller level. Real consumption can run 20-100% above your calculation.

Wear Leveling and Over-Provisioning Effect

Wear leveling comes in two types: dynamic (only balances rewritten blocks) and static (also periodically moves static data to spread the load across every cell). Good static wear leveling tries to keep the cycle difference between the most worn block and the least used one below 10%.

Over-provisioning (OP) is the extra NAND capacity that the manufacturer does not show the user. On enterprise SSDs it can go up to 28% (a drive sold as 1.92 TB actually has 2.5 TB of NAND), and on consumer drives it is around 7-12%. As OP increases, wear leveling becomes easier, garbage collection efficiency rises, and the TBW limit can extend by 3-4 fold. The "Momentum Cache" and OP setting of Crucial Storage Executive allows the user to add manual OP. If you leave 10-20% of your drive unpartitioned and empty, the wear leveling pool expands.

Backblaze Data: SSD Life in the Real World

Backblaze SSD statistics share the real-world Annualized Failure Rate (AFR) data of tens of thousands of SSDs. According to the 2024 reports, the overall AFR of boot SSDs is at the 0.9-1.0% level, lower than the 1.4% of HDDs, but far from the claim of "never failing". Backblaze also emphasizes that failed SSDs die from a controller or firmware issue before reaching the TBW limit; the field data says the dominant culprit is electronic failure, not NAND wear. This finding matches DSET's laboratory statistics exactly; you can read the 12 lessons we drew from our 20 years of field experience.

Enterprise vs Consumer Life Difference

Let us make a comparison: the Samsung PM1733 (enterprise NVMe) 1.92 TB model comes in a 3504 TBW (1 DWPD) or 10512 TBW (3 DWPD) version. At the same size, the Samsung 990 Pro 2 TB consumer model has a TBW of 1200. So the enterprise model offers 9 times more write capacity. The difference is not only from NAND quality; the enterprise drive has a power-loss capacitor (PLP), end-to-end data path protection, wider OP, and more aggressive wear leveling. If your workload is database, log aggregation, or VM, buying a consumer SSD just to save a little money causes TBW to fill up within 6-12 months.

Data Recovery on a TBW-Exhausted Drive

Field perspective: when a drive that has exceeded its TBW limit arrives at the lab, the first thing we do is immediately take a read-only copy. Because in this state every additional write can turn into catastrophic data loss.

The workflow we use for these types of cases in the DSET lab:

  1. A read-only clone with DeepSpar Disk Imager or a similar hardware imager. Bad sectors and uncorrectable ECC blocks are isolated.
  2. If the controller is in a "frozen" state, we use PC-3000 to put the drive into technological mode via vendor-specific commands and take a raw NAND dump.
  3. If a NAND chip-off is required, the chips are removed at the BGA station and the controller's XOR/LDPC algorithm is solved by reverse engineering.
  4. If there are inconsistencies at the file system level (NTFS $MFT corruption, exFAT FAT table loss), the metadata is rebuilt. This process is similar to that in recovering photos from a formatted drive cases, but on a TBW-exhausted drive the raw bit error rate is 10 times higher due to cell aging.

There is a chance of recovering data from a TBW-exhausted drive, but every new boot and write attempt lowers that chance. The "let me try one more time" behavior is the most common cause of data loss.

Extending SSD Life: 7 Field Recommendations

  1. Keep TRIM on. On Windows, fsutil behavior query DisableDeleteNotify should return 0. TRIM increases garbage collection efficiency by informing the controller of deleted areas, and lowers write amplification.
  2. Leave 10-20% free space. Filling the entire drive destroys the wear leveling pool and forces the same cells to be written over and over.
  3. Move the pagefile/swap to an HDD or a separate drive (especially on systems with less than 16 GB of RAM). Constant swapping means an extra 30-50 GB of writing per day.
  4. Redirect constantly written logs and compiler cache to a RAM disk or HDD. Chrome cache, Docker overlay, and Webpack output can easily write 100+ GB/day.
  5. Prefer sleep over hibernate. Hibernate writes all of RAM (16 GB) to the SSD every time.
  6. Keep firmware up to date. Manufacturers release wear leveling and garbage collection improvements via firmware. Samsung Magician, Crucial Storage Executive, WD Dashboard, and Kingston SSD Manager manage this update automatically.
  7. Have a backup plan. Despite the best care, an SSD can die suddenly. The 3-2-1 rule: 3 copies, 2 different media, 1 off-site. Tracking TBW does not reduce the probability of death, it only makes it predictable.

Frequently Asked Questions (FAQ)

Is TBW counted annually or over the lifetime? Over the lifetime. Every write counts from the day the drive is manufactured; there is no reset.

Does the warranty continue when TBW runs out? No. The manufacturer's warranty works on the logic of "the TBW limit or the warranty period, whichever runs out first". Even if 5 years have not passed, the warranty ends once TBW is exhausted.

Will the drive explode if the SMART "Wear Level" hits 0%? No, there is no instant explosion. Usually the counter first drops to 1%, then switches to read-only mode. At this stage the drive can still be read, your data can be extracted, but new writes are rejected.

Why do some SSDs have twice the TBW at the same capacity? The NAND type (TLC/QLC), the over-provisioning ratio, the controller's wear leveling quality, and the manufacturer's warranty policy create the difference. Flagship models (Samsung 990 Pro, FireCuda 530) provide 2-4 times higher TBW than entry-level ones.

Should I buy a QLC SSD? For low write loads (media library, game storage), QLC is economical. But as an OS drive or work drive, TLC should be preferred; QLC's TBW is half that of TLC or lower.

Should I defrag my SSD? No. On an SSD, defragmentation consumes TBW by generating unnecessary writes, even though fragmentation has no impact. Windows already runs retrim instead of defrag on SSDs.

Will data come out of a TBW-exhausted drive even when plugged into a computer without backing it up? Usually yes, but this behavior is risky. Every boot process writes to the SSD (USN journal, file system metadata). DSET's advice: when a drive is suspect, rather than plugging it in and trying, work on it after taking a professional clone.

Will data be lost if I keep an SSD in cold storage for a long time? Yes. JESD218 guarantees retention for 1 year at 30 °C for a client SSD. In a hotter environment, or on a TBW-exhausted drive, this period can drop to a few months. An SSD is not an ideal medium for archiving.

DSET Data Recovery, Professional SSD Support

Every kind of SSD case, whether TBW-exhausted, switched to read-only mode, controller-frozen, or requiring a NAND chip-off, is solved in the DSET lab. We work with PC-3000, DeepSpar Disk Imager, a BGA rework station, and brand-specific vendor command sets. Across all manufacturers' NVMe and SATA SSDs, Samsung, Crucial, WD, Kingston, SK hynix, Seagate, Intel, Micron, our success rate is above the industry average.

Address: Hacettepe Teknokent, Ankara SSD Line: +90 536 662 38 09 No fee if no data is recovered.