RAM Speed and Latency Explained: MHz and CL

What RAM speed means on the label

A listing such as "32GB DDR5 6000MHz CL30" gives you three facts: how much memory the kit holds, how fast it moves data and how long it waits before answering. Capacity comes first, and our guide on how much RAM you need covers it. This page is about the other two numbers. For the basics of what memory does, start with what RAM is.

The speed figure is a data rate. DDR stands for double data rate: the memory sends data twice in every clock cycle, so a kit rated at 6000 MT/s (megatransfers per second) has a real clock of 3000 MHz. Shops and box art still say MHz almost everywhere. When you compare listings, treat 6000MHz and 6000 MT/s as the same kit.

CL stands for CAS latency. It is the number of clock cycles the memory needs between receiving a request and returning the data. CL30 means 30 cycles. At the same speed, a lower CL is quicker. Across different speeds the CL number on its own tells you very little, because the length of a cycle changes with the speed.

How to work out true RAM latency

The fix is to turn cycles into time. Crucial explains it as cycle time multiplied by the number of cycles. Written with the figures you see in a listing, that becomes:

true latency (ns) = CL x 2000 / speed in MT/s

So DDR4 3200 CL16 gives 16 x 2000 / 3200 = 10 ns, and DDR5 6000 CL30 gives 30 x 2000 / 6000 = 10 ns as well. Crucial notes that CL figures have climbed along with speeds, which kept true latency roughly level from one generation to the next. When two kits have the same latency in nanoseconds, pick the faster one, because it moves more data every second.

The table works out the latency for common speed and CL pairs and adds the cheapest desktop kit we found at UK shops this week for each. The price covers any capacity, so the cheapest row can be a small kit.

Two rows deserve a second look. DDR4 3200 at CL22 is almost 4 ns slower than the same speed at CL16. And two DDR5 6000 kits can differ too: G.Skill rates its Flare X5 32GB kit at CL30 and its Ripjaws S5 32GB kit at CL36, both at 6000 MT/s. The headline speed matches; the latency does not.

Why a new kit runs slower than the box says

Each DDR generation has standard speeds set by JEDEC. According to Kingston, official DDR4 speeds stop at 3200 MT/s and DDR5 starts at 4800 MT/s. Anything faster relies on a profile stored on the module itself.

Intel calls its profile XMP (Extreme Memory Profile). AMD's version for DDR5 on socket AM5 is EXPO. You switch the profile on in the BIOS, and the memory then runs at the speed and timings printed on the box. Both Intel and AMD treat these profiles as overclocking, outside the official specification.

Leave the profile off and a DDR5 6000 kit usually boots at a lower default of 4800 or 5600 MT/s. Nothing is broken when that happens. It is the most common reason a new kit looks slower in Windows than what you paid for, and one BIOS setting fixes it.

DDR5 RAM speeds for AMD and Intel

Start with the platform, not the kit. AMD AM5 (Ryzen 7000 and 9000) and Intel Core Ultra 200S take DDR5 only. Intel 12th to 14th gen Core chips handle either type, depending on the motherboard. The Core i5 14600K, for example, is listed for up to DDR5 5600 or DDR4 3200.

On AM5, DDR5 6000 CL30 is the kit most builders choose. Newegg's 2026 DDR5 buying guide explains why: at 6000 MT/s the memory runs in step with the processor's memory controller and Infinity Fabric. That speed sits above what AMD guarantees, so it depends on EXPO. AMD lists DDR5 5600 for the Ryzen 9 9950X with two modules and only DDR5 3600 with four. You can compare current kits on our DDR5 6000 CL30 page.

For Intel, the Core Ultra 9 285K is specified for DDR5 up to 6400 MT/s. Engadget reports that Core Ultra 200S reaches 6400 with CUDIMMs, modules that carry their own clock driver, and 5600 with regular sticks. Newegg suggests a 6000 to 6400 MT/s kit for these chips.

DDR4 RAM speeds: 3200MHz or 3600MHz?

If your PC already runs DDR4, 3200MHz is the natural choice. It is the highest official DDR4 speed, so any processor that supports it will run the kit without a profile. 3600MHz kits are common and can work well, but they run through XMP and need a board and processor that cooperate. Slower 2400 and 2666 kits mainly make sense when you add to an existing set and want to match its speed.

The table shows what each common DDR4 capacity and speed costs this week, from the cheapest desktop kit we found in each scope. It also shows what a step up in gigabytes or in speed adds.

An 8GB DDR4 3200MHz kit still suits a spare machine, but Newegg's 2026 tiers put browsing, office work and most 1080p gaming at 16GB. If the gap between 3200 and 3600 on your size is small, the faster kit can be worth it, provided both have similar true latency. Every DDR4 kit we track, with price per GB, is on our DDR4 RAM prices page, and DDR5 kits are on our DDR5 RAM prices page.

How much does RAM speed matter in games?

Less than the price gaps suggest. Engadget writes that no game needs a particular DDR5 speed, and that any game that runs on DDR5 runs fine on the slowest DDR5 4800 kits. Moving from the slowest to a mid range kit may add a little. Kits at DDR5 7200 and 8400 add only a small gain in frame rate when the processor is the limit, at a much higher price.

There is also a reason faster is not always better on AMD. Engadget explains that the memory controller can match the memory clock up to roughly 3000 MHz. Beyond that it drops to half the memory clock, which adds delay. In the same article, Tom's Hardware results showed a DDR5 6000 CL30 kit ahead of a DDR5 6400 CL32 kit. On Intel, Tom's Hardware found the lowest scoring 32GB kit about 4 percent behind the fastest across heavy productivity work.

Standard JEDEC kits follow a pattern of their own. Engadget calculates that JEDEC timings keep latency near 14 ns all the way from DDR5 3200 to DDR5 8800, the official ceiling JEDEC set in 2024. A faster standard kit gives you more bandwidth, not a shorter wait.

Common mistakes

• Comparing CL across different speeds. CL30 at 6000 MT/s responds faster than CL22 at 3200 MT/s. Work out the nanoseconds.
• Leaving XMP or EXPO off. The kit then runs at its default speed and you lose what you paid for.
• Treating DDR5 6000 CL30 as AMD's official spec. It is a popular overclocked setting, not a speed AMD guarantees.
• Buying four sticks for a high speed. Two modules hold high speeds more easily, and AMD lists a much lower speed for four.
• Adding a second kit with a different speed or CL. Corsair points out that replacing all modules with one kit is the only way to guarantee they work together.
• Paying for speed before capacity. A fast 16GB kit does not help if your work needs 32GB.

We do not test memory ourselves. Each week we collect the prices and the specifications that UK shops publish, and we work out latency from those specifications. All prices on this page are including VAT.