The wrong NVMe drive in a Proxmox host does not announce itself on day one. It runs fine, it benchmarks fast, and then the wear counter climbs faster than anyone expected, because Proxmox writes to its disk far more than a desktop ever does. The cluster filesystem, the metrics database, the logs, and every VM fsync all land on the same SSD, and if the pool is ZFS each sync write is written again for the intent log. Sequential speed, the number every spec sheet leads with, is almost irrelevant to how long that drive survives.
Two numbers decide it instead: write endurance, measured in terabytes written, and whether the drive carries power-loss protection. The picks below cover the range a Proxmox box actually needs. A proven consumer drive for a single-node VM host, an enterprise M.2 with capacitor-backed power-loss protection for a cluster or a boot drive that has to survive a hard power cut, and a high-endurance data drive for a write-heavy pool. Most of these sit behind a mini PC or a homelab build, so form factor is M.2 first.
The 2026 flash shortage has pushed every one of these prices up and made the value ordering shift week to week, so treat each figure below as a band and confirm the live number before buying. The specs and prices here were checked against manufacturer datasheets and live listings in September 2026.
How the five drives compare
Every drive here is TLC, not QLC, and every one is M.2 2280, so it drops into a mini PC or a standard board without an adapter. The columns that decide the purchase are endurance and power-loss protection, not the headline read speed, which on all five is far more than a stack of VMs will ever use. Kingston does not publish a DRAM figure for the DC2000B, shown as n/s below.
| Drive | Capacity | NAND | DRAM | PLP | Endurance | Best for |
|---|---|---|---|---|---|---|
| Samsung 990 PRO | 2 TB | TLC | Yes | No | 1,200 TBW | Single-node VM host |
| Crucial T500 | 2 TB | TLC | Yes | No | 1,200 TBW | Value VM host |
| Kingston DC2000B | 960 GB | TLC | n/s | Yes | 700 TBW (0.4 DWPD) | Boot drive with PLP |
| Micron 7450 PRO | 960 GB | TLC | Yes | Yes | 1,700 TBW (1 DWPD) | Write-heavy pool |
| WD Black SN850X | 4 TB | TLC | Yes | No | 2,400 TBW | High capacity |
1. Samsung 990 PRO 2TB: the safe default for a VM host
For a single-node Proxmox host that runs a handful of VMs and containers, this is the drive to buy and stop thinking about.
The 990 PRO pairs 3-bit TLC V-NAND with a full LPDDR4 DRAM cache and Samsung’s own controller on a PCIe 4.0 x4 link, rated at 1,200 terabytes written and a five-year warranty at 2TB. The DRAM cache is the part that matters for Proxmox: it holds the flash translation table on the drive itself, so the small random writes that a VM host generates do not thrash the way they do on a DRAM-less drive. It is the most consistent consumer NVMe drive under sustained mixed load, which is exactly what a hypervisor produces. Expect roughly $360 to $400 at 2TB; check the live price.
Skip it if the host will run a busy database or a write-heavy log pipeline. It has no power-loss protection, so an ungraceful shutdown can lose the last in-flight writes, and 1,200 TBW is generous for a home host but modest next to an enterprise drive. For a boot-plus-VMs box that idles most of the day it is more than enough. For a drive that is hammered around the clock it is the wrong tool, and picks three and four exist for that.
2. Crucial T500 2TB: the same endurance for less
The T500 delivers the same 1,200 TBW rating and a full DRAM cache as the Samsung, usually for a little less money.
Crucial pairs Micron’s own 232-layer TLC with an LPDDR4 DRAM cache on a Phison E25 controller, rated at 1,200 TBW at 2TB with a five-year warranty. On paper and under a VM workload it lands within the margin of the 990 PRO: the same endurance, the same DRAM-backed random-write behavior, and a sequential rating no Proxmox host will saturate. Because Crucial is Micron’s retail brand, the NAND supply is in-house, which tends to keep it a step cheaper than Samsung when both are in stock. Figure about $360 to $400 at 2TB, though the shortage has narrowed that gap; check the live price.
The trade-off is the same as the Samsung’s: no power-loss protection, consumer endurance. It also stocks out more often than the 990 PRO, so which of the two is the better buy on any given day comes down to price and availability rather than a real performance gap. Pick whichever is cheaper and in stock when the VM host is being built.
3. Kingston DC2000B 960GB: power-loss protection in an M.2 boot drive
This is the drive most homelab builders do not know exists: an enterprise boot SSD with real power-loss protection in a plain M.2 2280 stick.
The DC2000B is Kingston’s data-center boot drive: 3D TLC behind a PCIe 4.0 x4 controller, with onboard capacitors that give it full hardware power-loss protection. It is rated at 0.4 drive writes per day, which works out to 700 terabytes written at 960GB over a five-year warranty, and its sequential write tops out around 1,300 MB/s against a 7,000 MB/s read. Those numbers look weak next to a consumer drive, and that is the point: the job is not a data pool, it is the operating system and the cluster filesystem, where the value is the capacitors, not the write ceiling. On a Proxmox cluster the DC2000B holds the pmxcfs cluster database through a power cut without a corrupt config, and it makes an honest ZFS boot mirror or a small SLOG in a way no consumer drive can. It carries an enterprise premium, roughly $700 to $850 at 960GB during the shortage, and it sells mainly through business and reseller channels rather than the usual retail shelf; check the live price and consider the 480GB model, which is plenty for a boot device.
Skip it as a VM store. At 0.4 DWPD and 960GB it is not built to absorb heavy guest writes, and it is slower under sustained load than any of the consumer drives here because it is tuned for steady-state consistency, not burst speed. Buy it for the boot pair or the SLOG on a cluster that must not lose its config, and put the VMs on one of the higher-endurance drives.
4. Micron 7450 PRO 960GB: enterprise endurance for a write-heavy pool
When a Proxmox node actually earns its endurance rating, a busy database or a log-heavy VM stack, this is the M.2 built for it.
The 7450 PRO is a full enterprise drive in M.2 2280 form: Micron 176-layer 3D TLC, an onboard DRAM cache, capacitor-backed power-loss protection, and a 1 drive-per-day rating that the datasheet puts at 1,700 terabytes written at 960GB. That is about two and a half times the endurance of the DC2000B and enough to run a write-heavy VM pool for years without touching the wear budget. It has the power-loss protection the consumer drives lack, so it is safe for synchronous database writes, and its steady-state performance holds where a consumer drive falls off after the cache fills. Plan on roughly $700 to $780 at 960GB; check the live price.
The price is the whole objection. At three to four times the cost per terabyte of a consumer drive, the 7450 PRO only makes sense once the workload genuinely writes enough to threaten a 1,200 TBW consumer drive inside its warranty, which most home nodes never do. Run the wear-rate check at the end of this guide before buying it. If a consumer drive would take fifty years to wear out on your workload, this one is money spent on endurance you will never use. On a node that chews through a consumer drive in two years, it pays for itself.
5. WD Black SN850X 4TB: capacity without dropping to QLC
Four terabytes is where consumer drives are tempted into QLC to hit a price. The SN850X stays on TLC, which is why it is the capacity pick here.
The SN850X uses BiCS TLC with a full DRAM cache on WD’s own controller, PCIe 4.0 x4, rated at 2,400 terabytes written at 4TB with a five-year warranty. At four terabytes it gives a single-node host room for a real VM library, ISOs, and backups on one drive without stepping down to the slower, lower-endurance QLC that most cheap 4TB drives use. Endurance scales with capacity, so the 4TB carries double the TBW of a 2TB drive, which makes it the most forgiving consumer option for a busy single-drive host. Expect around $620 to $680; check the live price.
It shares the consumer limitation of picks one and two: no power-loss protection, so it is a VM store and a capacity drive, not a place for synchronous database writes you cannot afford to lose. It also runs warm at 4TB and wants airflow or a heatsink in a sealed mini PC. Skip it if the host only needs a terabyte or two, where the 2TB drives are far better value per gigabyte, or if the workload needs PLP, where the enterprise drives are the answer regardless of capacity.
Why Proxmox writes so much to disk
A Proxmox host is a constant low-level write generator even when the VMs are idle. The cluster filesystem, pmxcfs, fsyncs its SQLite config database to disk on every change; the high-availability stack persists its LRM and CRM state; the built-in RRD metrics database updates every few seconds; and the system journal logs the lot. None of that is large, but it is relentless small random writes, which is the workload that ages an SSD fastest. A drive that looks bored on a throughput graph can still be burning through its wear budget.
ZFS makes it worse in the way that matters. Every synchronous write is committed twice, once to the ZFS intent log and again to the pool, and copy-on-write plus metadata updates add their own amplification on top. Without a dedicated log device that traffic lands on the same VM drive. This is why a consumer SSD that would last a decade in a desktop can wear out in a year or two under a busy ZFS Proxmox node, and why giving ZFS enough RAM for its cache is part of protecting the disk, not a separate concern.
Power-loss protection, and when you actually need it
Power-loss protection is a bank of capacitors on the drive that holds enough charge to flush its write cache to flash if power drops mid-write. Enterprise drives have it; consumer drives do not, no matter how fast they are. The practical effect on Proxmox is twofold. A drive with PLP can safely acknowledge a synchronous write the instant it hits the cache, which is why enterprise drives keep their sync-write speed under load while consumer drives collapse. And it protects the data already in flight when a node loses power, which for the cluster filesystem or a database is the difference between a clean reboot and a corrupt config.
Not every build needs it. A single-node host on a UPS, running VMs that are backed up, survives fine on a consumer drive. Where PLP earns its premium is a cluster that cannot afford a corrupt pmxcfs, a ZFS SLOG device that must honour every sync, and any node running a database where a lost write is a lost transaction. If the workload is synchronous and the data matters, the capacitors are the spec to buy for, not the read speed.
DRAM, HMB, and why QLC struggles in a VM store
Two spec-sheet details separate a drive that holds up as a VM store from one that does not. The first is the DRAM cache. A drive with onboard DRAM keeps its flash translation table on the drive; a DRAM-less drive borrows a slice of system RAM through the host memory buffer, which is fine for a desktop but stumbles under the sustained small random writes a hypervisor produces. Every drive here meant to hold VMs has a real DRAM cache for that reason. Treat DRAM-less as disqualifying for a VM host, however cheap it looks.
The second is the NAND type. TLC stores three bits per cell; QLC stores four, which raises capacity and lowers cost but cuts both write endurance and sustained write speed hard. A QLC drive writes fast into its SLC cache and then falls off a cliff once that cache fills, which under a VM workload happens routinely. Every drive here is TLC on purpose. QLC has a place for cold bulk storage and backup targets, not for the drive the VMs run on.
Match the drive to its job
The five picks map onto four Proxmox roles, and most builds want more than one drive. For the boot device, a small drive is enough, and on a cluster the DC2000B’s power-loss protection is worth the premium so the config survives a power cut. For the VM store on a single node, a 990 PRO or T500 at 2TB, or the SN850X at 4TB when capacity is the constraint, covers almost everything a home host runs.
For a ZFS SLOG or a write-cache device, PLP is not optional, so it is a DC2000B or a 7450 PRO, never a consumer drive. And for a genuinely write-heavy pool, a busy database, a metrics stack, a node hosting dozens of active VMs, the 7450 PRO’s endurance and steady-state performance are the reason it exists. Bulk cold storage does not belong on any of these; that is spinning disk behind an HBA or a separate NAS, not premium NVMe.
How long these drives actually last under Proxmox
The endurance rating and the real lifespan are not the same number, because write amplification sits between them. A quiet single-node host might write 20 gigabytes a day to its system drive, which against a 1,200 TBW rating is over a century of nominal endurance. A busy ZFS node with synchronous database writes can amplify host writes five to ten times before they reach the flash, which turns that same century into a handful of years. The rating tells you the ceiling; only measuring your own node tells you the rate.
Read the real figures straight off the drive with the NVMe SMART log:
sudo nvme smart-log /dev/nvme0 | grep -iE "percentage_used|data_units_written"Percentage_used is the drive’s own estimate of consumed endurance, where 100 means the rated write budget has been used up, and data_units_written counts the host writes sent to the drive in 512,000-byte units, so multiply by 512,000 to get total bytes written. That host figure is what the TBW rating counts too, so the two compare directly. Note both figures, wait a week of normal load, and read them again. The difference is the real write rate on your workload, and it turns the endurance rating into an actual number of years. If that number is comfortably longer than the warranty on a consumer drive, the enterprise drives are endurance you will pay for and never use. If the drive is on course to wear out inside a couple of years, that is the signal to move the write-heavy guests onto a 7450 PRO and let the wear counter tell the truth.
