What TBW and DWPD Actually Mean
Quick Answer: Terabytes Written is the cumulative host write volume a drive is warranted to accept. Drive Writes Per Day is how many times its usable capacity can be overwritten daily across the warranty term. Both describe one endurance guarantee, so a drive never has independent TBW and DWPD limits.
Endurance exists because NAND flash cells tolerate a finite number of program and erase cycles before they can no longer hold charge reliably. The controller spreads writes across the array through wear levelling, so the practical limit becomes a whole-drive figure rather than a per-cell one. Manufacturers publish that limit as a warranty commitment in one of two units.
- TBW is absolute. One cumulative number, independent of how long you take to consume it. Easier for comparing drives of different capacities.
- DWPD is a rate. Normalised to capacity and time, which makes it easy to check against a workload — but only once you know the term it assumes.
- Both count host writes. The figure describes data written by the server, not the larger volume the controller actually commits to NAND.
Converting Between TBW and DWPD
Quick Answer: Multiply DWPD by usable capacity in terabytes, by 365, and by warranty years to reach TBW. Reverse the arithmetic to go the other way. The warranty term appears in both directions, which is exactly why it must be stated on any quotation you intend to compare.
The two conversions
- DWPD to TBW: TBW = DWPD × capacity (TB) × 365 × warranty years
- TBW to DWPD: DWPD = TBW ÷ capacity (TB) ÷ 365 ÷ warranty years
- Worked example: a 3.84 TB drive at 1 DWPD over five years equals 3.84 × 365 × 5, or roughly 7,008 TBW
- Same drive, three-year term: the identical 7,008 TBW now reads as approximately 1.67 DWPD
The worked example is the whole point. Two vendors can quote 1 DWPD and 1.67 DWPD for drives with exactly the same endurance, and the higher number wins a naive comparison. Normalising both to TBW removes the ambiguity in one step.
Why the Warranty Term Changes the Number
Quick Answer: DWPD is endurance divided by time, so shortening the warranty term inflates the rating without changing the hardware. Endurance warranties also expire on whichever limit arrives first — the TBW ceiling or the term in years — so a generous TBW on a short term may never be reachable.
Two consequences follow, and both matter at the point of award rather than at installation.
- Comparison requires normalisation. Convert every candidate to TBW at a common term before ranking them, or the shortest warranty appears to be the strongest drive.
- Whichever comes first. A five-year drive with a TBW you would need eight years to consume is effectively a five-year drive. Conversely a write-heavy workload can exhaust TBW in year two, ending cover while the term still has years left.
Read Intensive, Mixed Use and Write Intensive
Quick Answer: Vendors group drives into tiers by endurance: read intensive around or below 1 DWPD, mixed use around 3 DWPD, and write intensive around 10 DWPD. The labels are marketing shorthand for over-provisioning decisions, and the tier names are not standardised across manufacturers.
| Tier | Typical DWPD band | Fits | Watch for |
|---|---|---|---|
| Read intensive | Around 1 or below | Content repositories, VDI images, analytics reads, general virtualisation | Adequate for most workloads; the default starting point |
| Mixed use | Around 3 | Transactional databases, virtualisation with heavy churn, mid-tier arrays | Justify with measured writes, not with caution |
| Write intensive | Around 10 | Write-through caches, logging tiers, metadata devices, message queues | Substantial price premium and lower usable capacity |
| Capacity / QLC | Well below 1 | Bulk and archival storage, read-dominant object tiers | Endurance is the constraint, not the bonus |
Because the bands are conventions rather than standards, one vendor's mixed use can be another's read intensive. Compare the DWPD or TBW figure, never the tier name.
NAND Type and Over-Provisioning
Quick Answer: Endurance falls as bits per cell rise, from SLC through MLC and TLC to QLC, because each additional bit narrows the voltage margin. Over-provisioning then sets the tier: reserving more spare area gives the controller room for wear levelling, turning identical NAND into different endurance SKUs.
This explains an otherwise puzzling catalogue. Two drives share a generation of NAND and a controller, yet one is rated several times the other and offers less usable capacity. The higher-endurance model simply hides more of the raw flash as spare area.
- Bits per cell set the ceiling. Fewer bits per cell means wider voltage margins and more tolerable program-erase cycles.
- Over-provisioning sets the tier. Larger spare area reduces write amplification and extends usable life, at the cost of advertised capacity.
- Capacity affects the rating. Within a family, larger drives usually carry higher TBW because writes spread across more cells — which is why DWPD stays flat while TBW rises.
NAND supply conditions also move the price of these tiers independently of the specification. Current dynamics are covered in DRAM and NAND BOM risk and the enterprise SSD and HDD supply crisis.
Write Amplification: The Hidden Multiplier
Quick Answer: Write amplification is the ratio of data the controller writes to NAND versus data the host sent. Large sequential writes approach a factor of one; small random writes multiply it substantially, because a partial page update forces the controller to relocate and erase whole blocks.
Endurance warranties are stated in host writes, so amplification does not directly consume the TBW figure. It matters for a different reason: it determines whether the drive can sustain the throughput at all, and it explains why an identical daily write volume ages two drives differently depending on the access pattern.
- Access pattern dominates. A database performing small random updates amplifies far more than a backup target writing sequentially.
- Array features change the picture. Deduplication, compression and RAID parity all alter what actually reaches the drives, so drive-level and array-level write volumes are not the same number.
- Filesystem alignment helps. Aligning writes to the drive's page and block size reduces avoidable relocation.
Sizing Endurance from Real Workload
Quick Answer: Measure host writes over a representative period including month-end and backup windows, divide by days to get daily terabytes, then divide by the candidate drive's usable capacity. That quotient is the DWPD the workload demands. Add growth and headroom, then choose the lowest tier that covers it.
Deriving the requirement instead of guessing it
- Collect telemetry. Total host writes from drive attributes or array reporting, over at least one full business cycle.
- Normalise to a day. Divide by days measured for average daily host writes in terabytes.
- Express as DWPD. Divide daily writes by the usable capacity of the drive under consideration.
- Add growth. Apply the expected data growth rate across the intended service life, not today's volume.
- Add headroom, not a tier. Round up within reason; jumping a whole tier without measurement is how overspend enters a BOM.
- Re-check after consolidation. Fewer, larger drives concentrate the same writes onto less capacity, raising required DWPD.
What to Put on the RFQ
Quick Answer: Require TBW and DWPD together with the warranty term for every line, state usable rather than raw capacity, and ask whether the endurance warranty ends at the TBW ceiling or the term. Without those three fields, quotations from different vendors cannot be ranked.
Fields that make endurance comparable
- TBW per drive, and DWPD with the warranty term it assumes
- Usable capacity, distinguished from raw capacity
- Warranty duration in years, and which limit terminates cover first
- NAND generation and bits per cell, where the vendor discloses it
- Interface and form factor, so the drives fit the intended enclosure
- Whether the array applies deduplication or compression ahead of the drives
- Measured or estimated daily host writes for the target workload
- Expected data growth rate across the service life
Frequently Asked Questions
Quick Answer: TBW and DWPD are the same warranty in different units, related through capacity and warranty term. Convert everything to TBW to compare vendors, size the tier from measured writes rather than caution, and treat write intensive drives as the exception rather than the safe default.
Drive Writes Per Day: how many times the usable capacity can be overwritten daily across the warranty term. On a 3.84 TB drive, 2 DWPD permits roughly 7.68 TB of host writes per day.
Terabytes Written: the total host write volume the drive is warranted to accept over its lifetime. It states the same limit as DWPD as a single cumulative number.
TBW = DWPD × capacity in TB × 365 × warranty years. Reverse the arithmetic for the other direction. Both need the warranty term, which is why it must appear on the quotation.
Over-provisioning. The higher-endurance SKU reserves more raw flash as spare area, giving the controller room for wear levelling at the cost of advertised usable capacity.
No, it is usually overspend. Most enterprise workloads never reach 1 DWPD. Write intensive drives cost substantially more and give up usable capacity the workload will not consume.
Warranty cover ends, on whichever limit arrives first. Drives commonly run beyond their rating, but past that point monitoring reported wear becomes a requirement rather than good practice.
Key Takeaway: Normalise every candidate to TBW at a common warranty term, derive the required DWPD from measured host writes, and buy the lowest tier that covers it with headroom. BENQUAN GLOBAL LIMITED supplies enterprise SSDs and all-flash arrays with competitive pricing and BOM validation for global buyers.