What is NVMe storage, and does it matter for hosting?

NVMe in plain words

Every server stores your website on some kind of drive. When a visitor opens a page, the server reads files and database records from that drive. NVMe (Non-Volatile Memory Express) is a standard for how solid state drives connect to and talk to the computer. It was designed for flash memory from the start, rather than borrowing an interface designed for spinning hard disks.

So "NVMe hosting" means your account sits on solid state drives that use the NVMe interface. It describes how fast and how many requests the storage can handle at once, not how much space you get.

Three kinds of storage you will see in hosting

	HDD (hard disk)	SATA SSD	NVMe SSD
How data is stored	Magnetic platters that spin, with a moving read head	Flash memory chips, no moving parts	Flash memory chips, no moving parts
How it connects	SATA interface	SATA interface	PCI Express (PCIe) lanes
Command protocol	AHCI (designed for hard disks)	AHCI	NVMe
Parallel commands	One queue of up to 32 commands	One queue of up to 32 commands	Many queues, each able to hold many commands
Main weakness	Slow random access because the head has to move	Limited by the SATA link and single queue	Cost, though prices have fallen
Where you still see it	Bulk storage, archives, cheap backup space	Older or budget servers	Most current hosting servers

Why the interface matters

PCIe vs SATA

A SATA drive connects through a host controller that was designed when every drive was a hard disk. The current SATA revision tops out at 6 Gbit/s on the link, and SSDs soon became fast enough to fill it.

An NVMe drive connects directly to the computer's PCI Express bus, the same high speed connection used by graphics cards. The NVM Express organisation describes an NVMe host using PCIe to access NVMe SSDs, without the separate host adapter used by SAS or SATA storage. PCIe lanes carry far more data than a SATA link, and the path between processor and storage is shorter.

Queues and queue depth

The bigger difference on a server is parallelism.

• AHCI, the protocol used by SATA drives, gives each port a single command list of up to 32 commands.
• NVMe supports many I/O queues. According to the NVM Express organisation, the protocol theoretically allows about 64K I/O queues, each with about 64K outstanding commands. Real drives support a fraction of that; vendors report between 16 and 256 queues with depths from 32 to 2,048.

Queue depth is how many requests can be waiting for the drive at once. A hosting server is not one person reading one file. It is dozens or hundreds of websites, each running PHP, reading small files and querying databases at the same moment. That workload is many small, random reads and writes in parallel, which is exactly where deep, multiple queues help. Each CPU core can have its own queue, so cores do not have to wait for each other to hand requests to the drive.

What NVMe means for your website in practice

Be wary of claims that NVMe makes a site "10 times faster". The real effect depends on where your site spends its time.

Where NVMe helps

• Busy shared servers, where many accounts read from disk at once. Faster, parallel storage means one account's activity is less likely to slow others.
• Database heavy sites, such as WooCommerce shops with large product and order tables, when data is not already cached in memory.
• Admin work: WordPress dashboard pages, searches and reports often bypass page caches.
• Backups and restores, which read or write the whole account.

Where NVMe does not help much

• Cached pages. If a page is served from a cache in memory, storage is barely touched.
• Heavy front ends. A 5 MB home page with uncompressed images and many scripts will be slow on the visitor's phone regardless of the server's drive.
• Slow code or plugins. A plugin making dozens of external API calls on every page is limited by those calls.
• Network distance. Time to First Byte also depends on how far the visitor is from the server, DNS lookup and connection set up. Google's web.dev guidance on TTFB lists these alongside server processing time.

In short, NVMe removes storage as a common bottleneck. It does not replace good site building. Our guide on improving website speed covers caching, images and the other factors that usually matter more.

How to read NVMe claims on a plan

When comparing plans, check:

1. Is the storage type stated for the plan you are buying? Some providers use NVMe on some products and not others.
2. Is it NVMe for all data, or only part of it? For example, NVMe for files but databases on other storage.
3. How much NVMe storage is included? Size and type are separate questions; our hosting terms guide helps you estimate how much you need.
4. What are the other resource limits? CPU, memory and process limits on shared hosting often matter more than disk speed.

ODHOST note: All five ODHOST shared hosting plans list NVMe storage: 10 GB on Spark, 25 GB on Boost, 50 GB on Thrive, 100 GB on Turbo and 200 GB on Elite. VPS and dedicated server plans list storage size, but the storage type is not stated, so ask by ticket if NVMe matters for your workload. See shared hosting and the full pricing page.

Technical notes for developers

• Form factors. NVMe drives come as M.2 cards, U.2 and U.3 drives (common in servers) and newer EDSFF formats. The form factor does not change the protocol.
• Random I/O matters more than sequential. Web and database workloads are dominated by small random reads. Headline sequential throughput figures are less relevant for hosting than random read latency and IOPS at realistic queue depths.
• Memory still wins. A database that fits in RAM rarely touches disk for reads. On a VPS, adding RAM for the database buffer pool often does more than faster storage.
• Virtualisation adds a layer. On a VPS, your virtual disk sits on the host's storage through the hypervisor. The underlying drive type still matters, but so does how busy the host is.