SSD Endurance Explained: DWPD, TBW and Choosing the Right Class

Enterprise SSDs are quoted with an endurance rating, and it is the specification that separates a drive that lasts five years from one that fails in eighteen months in the same slot. It is also the one most often ignored in favour of capacity and price.

This guide covers what DWPD and TBW actually mean, how to work out which class you need, and why the wrong choice is expensive in both directions.

Why flash wears out and mechanical drives do not

The mechanisms are fundamentally different, and this is the root of everything else.

A mechanical drive has moving parts that eventually fail — a bearing, a head, a motor. It fails over time, and writing to it more does not meaningfully accelerate that. A drive used lightly and one used heavily have broadly similar lifespans.

Flash is different. Each cell can be written and erased a finite number of times before it can no longer hold charge reliably. An SSD wears out through use, specifically through writes, and a drive under heavy write load reaches end of life far sooner than one that is mostly read.

That is why endurance is quoted at all. It is not a reliability figure, it is a consumption budget.

DWPD: drive writes per day

The most common enterprise measure. It states how many times the drive’s full capacity can be written every day, across the warranty period, before endurance is exhausted.

A 1 DWPD drive of 1TB can absorb 1TB of writes daily. A 3 DWPD drive of the same capacity can absorb 3TB daily.

Two things about this trip people up.

It scales with capacity. A 4TB drive at 1 DWPD absorbs 4TB of writes a day — four times what a 1TB drive at the same rating absorbs. So a larger drive at a lower DWPD can have more total endurance than a smaller drive at a higher one.

It is an average, not a limit. A day of heavy writes followed by a quiet week is fine. The figure is a budget over the warranty period, not a daily cap.

TBW: terabytes written

The same information expressed as a total. Instead of a daily rate, TBW states the cumulative volume of writes the drive is rated to absorb over its life.

TBW is more common on consumer and prosumer drives; DWPD is standard in enterprise. They convert into each other given the capacity and warranty period, and neither is more accurate — they are the same budget expressed two ways.

TBW is arguably easier to reason about when comparing drives of different capacities, because it is an absolute number rather than a ratio.

The three endurance classes

Enterprise SSDs are broadly sold in three tiers, and the price gap between them is substantial.

Read-intensive sits around 1 DWPD or below. Built for workloads that are predominantly read: web servers, content delivery, boot volumes, reference data, analytics on data written once and queried often. Cheapest per gigabyte.

Mixed-use sits around 3 DWPD. The general-purpose tier for virtualisation hosts, general databases and application servers where reads dominate but writes are steady and significant.

Write-intensive sits at 10 DWPD or above. For sustained heavy writes: transactional database logs, caching tiers, write-back buffers, high-frequency ingest. Considerably more expensive per gigabyte.

The pricing gap between read-intensive and write-intensive at the same capacity is large enough that getting the class wrong is a meaningful budget error in either direction.

Working out which class you need

The calculation is straightforward and most people never do it.

Measure the actual write volume of the workload over a representative period — a week covering both busy and quiet days. Most operating systems and hypervisors report bytes written per volume.

Divide daily writes by the capacity of the drive you are considering. That gives the DWPD the workload actually demands.

Then add margin, because workloads grow and estimates are optimistic. If the measurement says 0.8 DWPD, a 1 DWPD drive is cutting it fine over five years.

What this exercise usually reveals is that the workload needs less than assumed. Write-intensive drives get specified defensively far more often than the write volume justifies, and read-intensive drives at higher capacity frequently deliver more total endurance for less money.

Write amplification: why you write more than you think

The factor that makes measured figures optimistic.

Flash cannot overwrite in place. To modify data it must erase a whole block and rewrite it, which means changing a small amount of data can cause a larger physical write internally.

The ratio between what the host writes and what the flash actually writes is write amplification, and it is always above one. It rises when the drive is nearly full, because the controller has less free space to work with when consolidating blocks.

Two practical consequences. Real endurance consumption exceeds host-measured writes, so build in margin. And keeping a drive very full accelerates wear — leaving headroom is not wasted capacity, it extends the drive’s life.

Enterprise drives include over-provisioning — spare capacity reserved by the manufacturer specifically for this, which is part of why an enterprise drive of a given usable capacity contains more flash than a consumer drive of the same rating.

Power loss protection

A specification that separates enterprise from consumer SSDs and is invisible until it matters.

SSDs buffer writes before committing them to flash. If power is lost between the acknowledgement and the commit, that data is gone — and worse, the internal mapping tables can be left inconsistent, which can corrupt data unrelated to the write in progress.

Enterprise drives include capacitors holding enough charge to flush the buffer on power loss. Consumer drives generally do not.

This is one of the main reasons consumer SSDs are inappropriate in servers regardless of endurance rating. A server with redundant power supplies and a UPS is less exposed, but power loss protection is protecting against controller-level and supply-level events too.

Where mechanical drives still win

Being clear about this matters, because flash is not the answer to everything.

Capacity tiers. Cost per terabyte on nearline drives remains far below flash and will for the foreseeable future. Backup, archive, surveillance and bulk storage are sequential workloads that gain little from flash latency.

Sustained write-heavy workloads on a budget. A mechanical drive has no wear ceiling. Where writes are heavy but latency is not critical, the endurance premium on write-intensive flash can exceed what the workload is worth.

Existing arrays. Dropping one SSD into a mechanical RAID group achieves nothing, because the group runs at its slowest member. Our RAID guide covers why.

For latency-critical tiers, flash wins clearly. See our interface guide for how NVMe fits, and the drive tier guide for where 15K still holds ground.

Monitoring wear

Enterprise SSDs report remaining endurance, usually as a percentage of life used. That figure is worth monitoring for the same reason SMART attributes are on mechanical drives — it converts a failure into a planned replacement.

Unlike mechanical failure, endurance exhaustion is predictable. A drive at 60 percent life used after two years will reach end of life at a calculable point, and you can budget for it rather than discover it.

Sourcing

We source enterprise SSDs to order against exact part numbers. Send us your workload write volume if you have it, or the server model and what the drive will do, and we will recommend an endurance class rather than quoting the most expensive option.

The bulk quote page explains what to include, or email sarah.jane@techsellerusa.com.

Common questions

Tell us the workload and write volume and we will recommend an endurance class rather than the most expensive drive.