M.2 Key Types Explained: A, B, E and M Keys, Sizes and Slots

There are four M.2 key types in common use: A, B, E and M. The key is the notch cut into a module’s gold edge connector, and a matching ridge in the socket, and its position decides which modules fit and which interfaces they can use: M key for NVMe SSDs on up to four PCIe lanes, B key for SATA SSDs, 4G and 5G modems and GNSS, and A and E keys for Wi-Fi, Bluetooth and other small add-on cards.

M.2 has become the standard for onboard expansion in embedded and industrial computing, replacing Mini PCIe and mSATA for storage and wireless. It began as Intel’s Next Generation Form Factor (NGFF) and was standardized by PCI-SIG and SATA-IO around 2013. With up to four PCIe lanes to the processor, an M.2 socket carries far more bandwidth than Mini PCIe or mSATA, which use a single lane.

This guide covers the technologies found on M.2 SSDs and expansion cards and the features that decide compatibility and performance before you install one in a single board computer, industrial motherboard, embedded box PC or industrial computer: the key types and pin positions, sizes and naming, signal interfaces and storage protocols, SATA vs NVMe performance, heat in fanless systems, common compatibility mistakes and how to install a module.

The table is the quick reference. Each key sits at a fixed position on the connector’s 75 contact positions, and the pins under the notch are removed, so a module only fits a socket with the same key. A socket keyed one way and a module notched for two keys, such as B+M, can still fit together.

Two rules come out of the table. First, a socket’s key tells you which modules fit, but not which interfaces are wired up: an M-key socket may carry PCIe x4 only, SATA only or both, so the datasheet matters as much as the notch. Second, most A and E key cards use one PCIe lane even though the key allows two. Our FAQ on M.2 keying has a shorter summary.

M.2 sockets have key characteristics that define which M.2 modules are supported. Before installing a module, check that it is compatible with the host on two counts: the form factor, meaning its width and length, and the keying, meaning the notch position and the interfaces behind it. The sections below take each in turn, then cover the signals and protocols that decide performance.

The M.2 specification defines twelve key positions, labeled A to M without I. Only A, B, E and M are used by mainstream products today; the others are reserved for future use, and key G is set aside for non-standard designs. Up to 67 contacts remain on a module with a single notch.

Manufacturers sometimes list host sockets as A+E or B+M, which means modules notched for either key will fit. A particular key may still be needed to get the full bandwidth. For instance, SATA SSDs are usually notched B+M so they fit both sockets, but in an M-key socket a SATA drive will not use the PCIe x4 bandwidth the socket offers, and it will only work at all if the socket also carries SATA.

The M.2 form factor is a measurement of the module’s width and length, given as a four- or five-digit code: the first two digits are the width and the rest the length, in millimeters. An M.2 2280 module is 22 mm wide and 80 mm long, the most common size for M-key SSDs. Widths in the specification are 12, 16, 22 and 30 mm, and lengths run from 16 to 110 mm. Some host boards provide mounting points at several positions, giving more choice of module or a larger SSD.

Our FAQ on M.2 form factor sizes lists the full set. The size and the key are independent: a 2242 module can be B-keyed, M-keyed or B+M, so both have to match the socket.

M.2 modules also differ in thickness. The circuit board itself is 0.8 mm thick, and components can sit on one side or both. Single-sided modules carry an S code and suit thin, space-constrained systems where the module lies flat on the board; double-sided modules carry a D code and fit more flash chips for higher capacity. Some modules, such as M.2 video capture cards, also carry a heatsink for their processing chips. The full name of a module combines all three properties, for example 2280-D2-M or 2242-S3-B-M.

In a small fanless PC, check the host’s supported height code before choosing a double-sided SSD: a board-mounted socket over other components, or a heat spreader pressing on the module, may only accept single-sided drives.

Much like Mini PCIe, M.2 carries several signal interfaces and storage protocols, and these decide whether a device can talk to the host. Manufacturers list the supported M.2 signal interfaces and protocols on their datasheets.

M.2 signal interfaces: PCI Express, SATA and USB

The M.2 interface supports PCI Express up to four lanes, SATA 3.0 at 6 Gb/s and USB 2.0 and USB 3. That range is what makes the format so flexible: high-speed storage, AI accelerators and video capture on PCIe, cellular and GNSS modules on USB, wireless and Bluetooth on PCIe and USB, and I/O expansion modules on PCIe. Current industrial boards mostly offer PCIe 3.0 or 4.0 on their M.2 sockets, with PCIe 5.0 arriving on the newest platforms.

M.2 storage protocols: AHCI and NVMe

M.2 SSDs use one of two storage protocols. SATA drives use AHCI (Advanced Host Controller Interface), a protocol designed in the era of spinning hard drives that handles one command queue. NVMe (Non-Volatile Memory Express) was written for flash memory on PCIe and handles many queues in parallel with lower overhead, so it gets far more from NAND flash. A SATA M.2 SSD works only in a socket wired for SATA, and an NVMe SSD only in a socket wired for PCIe, regardless of whether the notch lets it fit.

Our guide to PCIe generations and bandwidth explains the lane figures in more depth.

The main difference between M.2 SATA SSDs and M.2 PCIe NVMe SSDs is speed. SATA tops out at 6 Gb/s, which caps a SATA SSD at roughly 550 MB/s. Each PCIe 4.0 lane runs at 16 GT/s, about 2 GB/s, and an M-key SSD uses up to four lanes. If you want the fastest NVMe SSDs, choose an M.2 M-key SSD and an M-key socket wired for PCIe x4.

The figures are typical datasheet peaks, so a PCIe 3.0 NVMe SSD is roughly six times faster than SATA on sequential reads. Sustained performance, especially on writes, depends on the drive’s controller, cache and temperature.

Capacity follows physical size. Larger capacities need more NAND flash chips, which leads to longer or double-sided modules. Mini PCs and small SBCs often trade speed or capacity for size, which is where 2230 and 2242 drives come in. For industrial use the flash type, endurance and temperature rating matter as much as speed: our FAQs on NVMe vs SATA on M.2, SSD TBW and power-loss protection cover the details.

M.2 modules run from a single 3.3 V supply, and their power has risen with every PCIe generation. Many PCIe 3.0 SSDs draw a few watts; PCIe 4.0 drives can reach around 8 W under sustained writes, and PCIe 5.0 drives can exceed 10 W. When the controller or flash gets too hot, the drive slows itself down to protect the data, which our guide to thermal throttling explains.

In a desktop PC a fan over the drive deals with this. In fanless PCs there is no airflow, so the heat has to travel through a thermal pad into a heat spreader or the chassis wall, the same principle as conduction cooling. We size storage to the workload: if the application records a few hundred megabytes per second, a PCIe 3.0 or 4.0 SSD has the bandwidth with less heat, and stays clear of throttling at the top of the temperature range.

Temperature rating is the other half. Industrial SSDs are offered in grades up to −40 to 85 °C, against 0 to 70 °C for most commercial drives, and our explainer on the industrial temperature range sets out the grades. In vehicles and on machines, the screw at the end of the module also holds it against vibration, so use the stand-off and screw the host supplies.

Most M.2 problems come from assuming that a module which fits will also work. These are the cases we see most often when specifying industrial systems.

A spare M.2 socket is also a common route to adding capability later. An AI accelerator such as the Hailo-8 in an M-key or B+M socket can add neural-network inference to a computer without a PCIe card slot, and our Hailo AI accelerator range lists systems built that way.

To work out which M.2 slot to use, read four things for each socket on the datasheet: its form factor, its key, the signals it supports and its bandwidth. Two systems from our range show how that looks in practice.

The same check works for any host. On the Sintrones SBOX-2625, for example, the M.2 key B slot takes either a 4G or 5G module with a SIM or an M.2 2242 SATA SSD, so a design that needs both cellular and extra storage has to put one of them elsewhere. For cellular builds our guides to industrial cellular computers and 5G at the edge go further, and our FAQ on GNSS vs GPS covers location modules.

Before M.2, onboard expansion and storage used Mini PCIe and mSATA. The two are hard to tell apart at first sight because they share the same card outline and connector: full-size cards measure 30 × 50.95 mm and half-size cards 30 × 26.8 mm. Sharing a connector means an mSATA card will physically fit an mPCIe slot and the other way round, but the protocol decides whether it works. Mini PCIe carries a single PCIe lane plus USB 2.0 for peripheral cards, while mSATA carries SATA for storage.

The PCI Express Mini Card was introduced in 2003 for Wi-Fi, Bluetooth and cellular cards in laptops and mobile devices, and quickly became an embedded standard across box PCs and single board computing designs with similar space constraints. Although it is now a legacy port, Mini PCIe is still widely used by industrial manufacturers alongside M.2, and our Mini PCIe capture cards remain popular for that reason.

SATA-IO announced mSATA in 2009, using SATA and AHCI to put solid-state storage onto a card. It made onboard storage practical on small boards such as 3.5 inch SBCs and Pico-ITX boards, removing the need for a 2.5 inch drive bay. Because not every system supports mSATA, an mSATA SSD may fit an mPCIe socket that only carries PCIe and USB, and simply not be seen.

Bandwidth was the limitation that led to M.2. With a single lane of PCIe or SATA, both formats are capped, and adding lanes was the way to raise throughput for SSDs. Our FAQ on M.2 vs mSATA vs mPCIe compares the three side by side.

M.2 is the current standard for small internal expansion cards fitted directly to a motherboard. Compared with Mini PCIe and mSATA it is faster and more flexible, supporting a wide range of storage and expansion cards on one connector family, with lower software overhead on NVMe and better power efficiency.

Benefits of M.2 expansion

High-speed technology: M.2 modules on PCIe 4.0 transfer about 2 GB/s per lane over up to four lanes, and M.2 NVMe SSDs are several times faster than SATA SSDs. That suits SSDs, Wi-Fi and Bluetooth, capture cards and accelerators.

Small form factors: M.2 cards take less board area than Mini PCIe cards. Small form factor computers, such as our industrial mini PCs, have little room for components, and M.2 lets them add Wi-Fi or a modem without giving up space.

Flexible lengths and widths: manufacturers choose the M.2 lengths their single board computers, motherboards and industrial PCs accept, either to keep the design small or to give users a choice of modules.

Power efficiency: M.2 runs from a 3.3 V supply and low-power modules draw very little, though fast NVMe SSDs need the thermal planning described above.

Installing M.2 expansion

When the module and motherboard have been confirmed compatible, installation is simple: insert the card into the socket at an angle of about 30°, press it down parallel with the motherboard, and fix the semicircular notch at the end of the card with the screw or latch provided.

These products show the main M.2 roles. We fit and configure M.2 SSDs, modems, Wi-Fi, capture cards and accelerators as part of each build, install the drivers, image the operating system and test the system with every module fitted before it ships.