A RAM kit ships running at a conservative JEDEC default speed no matter what the box says, and it stays there until an XMP or EXPO profile is enabled in BIOS -- the two standards do the same job for Intel and AMD platforms respectively.
Every memory kit runs at a slower, conservative JEDEC default speed the moment it's installed, regardless of what speed is printed on the box or the retail listing. Reaching the rated speed requires enabling a profile in BIOS β XMP on Intel platforms, EXPO on AMD's current platforms β and skipping that step is the single most common reason a "6000 MT/s" kit measures as running well below that in practice.
The JEDEC standards body defines conservative, guaranteed-compatible speed and timing combinations that any compliant motherboard and any compliant memory module must support without special configuration. That baseline exists so that plugging in memory works reliably across an enormous range of board and module combinations without either side needing to negotiate anything. The tradeoff is that the guaranteed-safe speed is well below what most modules can actually run at β the manufacturer validates a much higher speed and timing combination, but running it isn't automatic because it isn't guaranteed to work on every board the module might end up in.
An XMP or EXPO profile is a small block of data stored on the memory module itself, containing the manufacturer-validated speed, timing and voltage settings above the JEDEC default. Enabling the profile in BIOS tells the board to read that block and configure the memory controller to those settings instead of the conservative default. It is a one-time BIOS setting, not something that needs to be re-applied, but it is also not something any board turns on automatically β the safe-by-default JEDEC behavior is deliberately the out-of-box state.
XMP (Extreme Memory Profile) is Intel's standard, and it's also the profile format nearly every memory module ships with regardless of brand, because the DDR5 memory market largely converged on it before AMD introduced its own standard.
EXPO (Extended Profiles for Overclocking) is AMD's standard, developed specifically for its DDR5-based platforms. A module needs to explicitly carry an EXPO profile to expose one β a module with only an XMP profile still works on an AMD board, but at the JEDEC default speed, unless it also carries EXPO data specifically.
The practical consequence: when shopping for a kit destined for an AMD platform, check that the listing states EXPO support rather than assuming an XMP-labeled kit will read out an equivalent profile on an AMD board. Many current kits carry both, but not all of them do, and a kit lacking the matching profile format does not fail outright β it just quietly runs at the slower default speed with no BIOS-level indication that anything is wrong. Once EXPO support is confirmed, the next useful step is filtering by the DDR5 5600-6000 MT/s speed tier rather than trusting the box's headline number alone β it narrows the catalog to kits actually rated at the data rate you're targeting.
After enabling XMP or EXPO in BIOS, the operating system's own reporting tools (or the BIOS's own memory status screen) will show the actual running speed. This is worth checking after any new build or memory upgrade specifically because a profile that fails to apply β due to an unsupported combination of modules, or a board that can't stabilize the requested settings β typically fails silently back to the JEDEC default rather than throwing an error. A kit that measures at its box-stated speed confirms the profile is active; one that doesn't means the JEDEC fallback is still in effect.
A profile is validated by the manufacturer against the specific module count it was tested with, most often two matched modules. Populating four slots β whether with two kits combined or a single four-module kit β asks the memory controller to hold the same profile's voltage and timing across more electrical load on the same channels, and it is common for a board to need a manually loosened timing, a lower speed, or simply more trial and error to stabilize four modules at a profile that two modules ran without issue. This is a real limitation of the platform and the profile mechanism, not a defect in any specific kit β it's the reason a two-module kit at a given capacity is generally treated as the safer bet for reliably hitting a rated profile than an equivalent four-module configuration.
A motherboard's chipset generation and BIOS version both determine which profile formats it can read at all. A sufficiently old board may recognize XMP but have no EXPO support whatsoever, or may cap out at an earlier XMP revision than a given module's profile was written against β in either case the practical result is the same as a missing profile: the module falls back to its JEDEC default rather than negotiating a partial match. Checking a board's supported profile format and revision before buying memory for it matters more on an older or budget-tier board than on a current flagship one, where broad profile support is close to a given.
A CL30 rating on a DDR5-6000 kit only delivers the 10.0 ns true latency worked out in CAS latency, true latency and why CL numbers mislead once the kit is actually running at 6000 MT/s with that CL30 timing β which requires the EXPO or XMP profile to be active. Running the same physical module at its JEDEC fallback speed changes both numbers in the true-latency formula, so a kit bought specifically for its rated latency figure needs its profile enabled to actually deliver it. This is a one-time BIOS setting, and checking it after any build is worth the minute it takes given how much of a kit's purchased performance depends on it.
CAMM2 and SODIMM are physically incompatible laptop memory formats. The difference is not speed at a given rating but how many channels each module carries, how the module attaches, and whether a machine can ever add memory rather than replace it.
DDR5 doubled the bandwidth and moved power regulation onto the module. Whether that is worth paying for depends entirely on which platform you are building.
CAS latency is a count of clock cycles, not a unit of time, so a lower CL number does not mean a faster module once two kits run at different clock speeds -- converting to true latency is the only way to compare them honestly.