Memory calculator

How much memory a machine needs depends almost entirely on what you do with it. Choose a workload below for a capacity recommendation and an explanation of where the benefit stops.

Capacity is the decision that matters

Of the three memory specifications people compare, capacity, speed and latency, capacity is the one that changes whether a machine works well. The other two adjust performance at the margins.

The reason is that running out of memory is not a gradual slowdown. When physical memory is exhausted, the operating system starts moving data to disk to free space. Disk is far slower than memory even on a fast drive, and the effect is a machine that stutters, pauses and feels broken. Having enough memory removes that cliff edge entirely. Having more than enough does nothing at all.

Buy two modules, not one

Memory controllers on desktop platforms are dual channel, meaning they can talk to two modules at once. A single module leaves half that capability unused, and the difference is measurable, particularly where integrated graphics share system memory.

So 2 × 8 GB is a better purchase than 1 × 16 GB at the same capacity and usually the same price. It also leaves two slots free on a four-slot board for a later upgrade.

Why filling all four slots can slow you down

This is counterintuitive and catches people out. On most consumer platforms, populating all four memory slots makes it harder for the memory controller to maintain signal integrity, and the system compensates by running the memory slower than its rating.

The practical consequence: if you buy 2 × 16 GB now and add another 2 × 16 GB later, you may end up running all four modules below the speed the kit is rated for. Buying the capacity you want as a two-module kit from the start avoids this. It is also why mixing a new kit with an existing one is unpredictable, even when the specifications appear to match.

The profile you have to switch on

Memory installs and runs at a conservative default speed defined by the JEDEC standard, which is slower than the speed on the box. The advertised speed is a stored profile that has to be enabled in the motherboard firmware, labelled EXPO on AMD platforms and XMP on Intel.

A large number of machines run slower than the memory they contain because this step was missed. It is one checkbox in the BIOS, and it is the single easiest performance gain available after a build.

Speed and latency, briefly

Higher speed means more data transferred per second. Lower CAS latency means less delay before each transfer starts. They pull against each other, and a headline speed figure alone does not tell you which kit is quicker in practice.

For most builds this matters far less than the discussion around it suggests. Pick a kit at the speed your platform actually supports, at a sensible latency, and spend the difference elsewhere. The exception is AMD processors with stacked cache and integrated graphics of any kind, where memory performance does show up more clearly.

Related reading: How much RAM do you need? and DDR4 versus DDR5.