SSD Endurance and TBW Explained

Quick Recommendation

Don’t worry about SSD endurance for a home server. A typical 1 TB consumer NVMe SSD is rated for 600 TBW (terabytes written). A home server writing 20 GB/day (which is a lot) would take 82 years to exhaust that. Even Docker’s logging, database writes, and container updates don’t come close to stressing a modern SSD.

Buy any reputable 1 TB NVMe SSD ($50–60) for your home server. You’ll replace it because it’s too small or too slow long before endurance is a concern.

Exception: If you’re using SSDs as ZFS SLOG/L2ARC cache or running a write-heavy database (PostgreSQL with high throughput), buy a higher-endurance drive.

What Is TBW?

TBW (Terabytes Written) is the manufacturer’s rated endurance — how much total data you can write before the SSD is expected to start degrading. It’s a warranty metric, not a hard limit. Most SSDs survive well beyond their TBW rating.

SSD	Capacity	TBW Rating	Write Warranty
Samsung 870 EVO (SATA)	1 TB	600 TBW	5 years
Samsung 990 Pro (NVMe)	1 TB	600 TBW	5 years
WD Black SN770 (NVMe)	1 TB	600 TBW	5 years
Crucial P3 (NVMe)	1 TB	220 TBW	5 years
WD Red SN700 (NAS NVMe)	1 TB	2,000 TBW	5 years
Samsung 983 DCT (Enterprise)	960 GB	1,366 TBW	5 years

TBW per TB Comparison

A more useful metric — TBW normalized per TB of capacity:

Category	TBW per TB	Examples
Budget consumer	150–300	Crucial P3, Kingston NV2
Mainstream consumer	600	Samsung 870/990, WD SN770
NAS-optimized	1,300–2,000	WD Red SN700, Seagate IronWolf 525
Enterprise	1,400–3,600	Samsung PM893, Micron 5400

How Much Does a Home Server Write?

Typical Write Loads

Activity	Daily Writes	Monthly Writes
Docker container logs	100 MB – 1 GB	3–30 GB
Nextcloud sync (personal)	500 MB – 2 GB	15–60 GB
Plex/Jellyfin metadata updates	50–200 MB	1.5–6 GB
Pi-hole/AdGuard DNS logs	50–200 MB	1.5–6 GB
Database operations (SQLite/Postgres)	200 MB – 2 GB	6–60 GB
System updates + temp files	100–500 MB	3–15 GB
ZFS SLOG (if used)	5–50 GB	150 GB – 1.5 TB
Surveillance camera recording	20–100 GB	600 GB – 3 TB

Typical Total Daily Writes

Server Type	Daily Writes	Annual Writes	Years on 600 TBW Drive
Light (Pi-hole, RSS, bookmarks)	1–3 GB	365 GB – 1 TB	600–1,600 years
Medium (Nextcloud, Plex, 10 containers)	5–15 GB	1.8–5.5 TB	109–333 years
Heavy (database-intensive, CI/CD)	20–50 GB	7.3–18.3 TB	33–82 years
ZFS SLOG cache	20–100 GB	7.3–36.5 TB	16–82 years
Surveillance NVR	50–200 GB	18.3–73 TB	8–33 years

Bottom line: Unless you’re running a write-intensive NVR or ZFS SLOG, SSD endurance is a non-issue for home servers. Even the cheapest consumer SSDs will outlast the useful life of the hardware.

NAND Types Explained

NAND Type	Bits per Cell	Endurance	Speed	Cost	Best For
SLC	1	Highest (100,000 P/E)	Fastest	$$$$	Enterprise cache
MLC	2	High (10,000 P/E)	Fast	$$$	Enterprise, prosumer
TLC	3	Good (3,000 P/E)	Good	$$	Consumer, most servers
QLC	4	Lower (1,000 P/E)	Slower writes	$	Read-heavy, cold storage

P/E cycles = program/erase cycles per cell. More bits per cell = cheaper per GB but fewer P/E cycles.

For home servers: TLC is the default. High enough endurance for any home use case, good performance, reasonable price. QLC is fine for read-heavy workloads (media storage, static file serving). Avoid QLC for write-heavy workloads (databases, ZFS SLOG).

When Endurance Actually Matters

ZFS SLOG (Synchronous Write Log)

ZFS SLOG devices receive every synchronous write (NFS, iSCSI, databases with sync=always). This can mean 20–100 GB of writes per day on a busy NAS. For SLOG:

Recommended: WD Red SN700 (2,000 TBW/TB), Optane (near-unlimited endurance)
Acceptable: Samsung 970 EVO Plus (600 TBW/TB)
Avoid: QLC drives, budget consumer NVMe

SLOG size: 8–16 GB is sufficient. ZFS SLOG only buffers 5–10 seconds of writes. A larger drive wastes capacity — buy for endurance, not capacity.

ZFS L2ARC (Level 2 Adaptive Replacement Cache)

L2ARC is a read cache — much lower write load than SLOG. Any consumer SSD works. The L2ARC populates gradually and mostly reads after warmup.

Database Servers

If your home server runs a write-heavy PostgreSQL or MySQL instance (unlikely for most self-hosters but possible with Nextcloud at scale, Gitea with CI/CD, etc.):

Calculate actual daily writes: iostat -x 1 and watch the wkB/s column
Multiply by 365 to get annual writes
Compare against TBW

In practice, even a heavy home database writes less than 10 GB/day — 3.65 TB/year, or 164 years on a 600 TBW drive.

Surveillance / NVR

Continuous video recording is the one home use case that can stress SSD endurance:

4 cameras at 1080p, 24/7: ~100 GB/day
8 cameras at 4K, 24/7: ~800 GB/day

For NVR: use HDDs for long-term recording, SSDs only for the live buffer or database. Purpose-built NVR SSDs (WD Purple SN580) are designed for this workload.

How to Monitor SSD Health

# Install smartmontools
sudo apt install smartmontools

# Check SSD health (NVMe)
sudo smartctl -a /dev/nvme0n1

# Key fields to look at:
# - Percentage Used: how much of TBW has been consumed (0-100%)
# - Data Units Written: total data written (multiply by 512 bytes × 1000)
# - Available Spare: remaining spare NAND blocks (100% = fresh)
# - Media and Data Integrity Errors: should be 0

# Check SSD health (SATA)
sudo smartctl -a /dev/sda
# Key fields:
# - Wear_Leveling_Count: percentage of life remaining
# - Total_LBAs_Written: total writes in 512-byte sectors

Automated Monitoring

Set up email alerts with smartd:

# Edit smartd config
sudo nano /etc/smartd.conf

# Add monitoring for NVMe
/dev/nvme0n1 -a -o on -S on -n standby,q -s (S/../.././02|L/../../6/03) -W 0,0,45 -m [email protected]

# Restart smartd
sudo systemctl restart smartd

Or use Netdata or Grafana with Prometheus for visual dashboards of SSD health metrics.

Best SSDs by Use Case

Boot Drive (Low Writes)

Any reputable 256 GB+ NVMe SSD. The cheapest option from Samsung, WD, Crucial, or Kingston is fine. Don’t overspend — a boot drive barely writes.

Pick: Kingston NV2 250 GB (~~$20) or Crucial P3 500 GB (~~$30).

General Server (VM Storage, Docker)

Mainstream consumer TLC NVMe. 1 TB for room to grow. 600 TBW is more than enough.

Pick: WD Black SN770 1 TB (~~$55) or Samsung 990 EVO 1 TB (~~$60).

NAS Cache (ZFS SLOG/L2ARC)

Higher-endurance NVMe rated for NAS/write-intensive workloads.

Pick: WD Red SN700 500 GB (~$45) — 1,000 TBW for the 500 GB model.

Enterprise/Heavy Write

If you genuinely need enterprise endurance (you probably don’t):

Pick: Samsung PM893 960 GB (~$100 used) — 1,366 TBW, power-loss protection capacitors.

FAQ

What happens when an SSD reaches its TBW rating?

The SSD doesn’t die at exactly TBW. The rating is conservative — most SSDs survive 2–5x their rated TBW. After TBW is reached, the manufacturer’s warranty no longer covers the drive. The SSD may start to slow down or develop bad blocks, but immediate data loss is rare. Monitor SMART data and replace proactively.

Is TBW or DWPD more useful?

DWPD (Drive Writes Per Day) tells you how many times you can write the full capacity per day over the warranty period. For a 1 TB drive with 1 DWPD and a 5-year warranty: 1 TB × 365 × 5 = 1,825 TBW. DWPD is more common in enterprise specs. For home use, TBW is easier to reason about.

Should I use TRIM on my server SSD?

Yes. TRIM tells the SSD which blocks are no longer in use, improving performance and endurance. Most Linux filesystems (ext4, XFS) support TRIM. Enable periodic TRIM:

sudo systemctl enable fstrim.timer

Can I use a used/refurbished SSD for a home server?

Check the SMART data first. If Percentage Used is under 30% and Available Spare is above 80%, it has plenty of life left. Used enterprise SSDs are excellent value — they’re built for write-heavy workloads and are often pulled from service with 90%+ life remaining.