# CL or ns: how fast is memory really?

Updated 2026-10-30

Every memory kit is sold with two numbers: the speed in MT/s (for example DDR5-6000) and the CAS latency (for example CL30). Many buyers compare only one of them. That is misleading, because CAS latency is counted in clock cycles, and how long a cycle lasts depends on the speed.

## The formula

**True latency in ns = CL × 2000 ÷ speed in MT/s**

The factor 2000 comes from DDR memory making two transfers per clock cycle: DDR5-6000 runs at 3000 MHz, so one cycle lasts about 0.33 ns.

## Examples

- DDR4-3200 CL16: 16 × 2000 ÷ 3200 = **10.0 ns**
- DDR5-4800 CL40: 40 × 2000 ÷ 4800 = **16.7 ns**
- DDR5-5600 CL46: 46 × 2000 ÷ 5600 = **16.4 ns**
- DDR5-6000 CL36: 36 × 2000 ÷ 6000 = **12.0 ns**
- DDR5-6000 CL30: 30 × 2000 ÷ 6000 = **10.0 ns**
- DDR5-6400 CL32: 32 × 2000 ÷ 6400 = **10.0 ns**

So DDR5-6000 CL30 answers just as fast as DDR4-3200 CL16, even though its CL number is almost twice as high, and it moves almost twice as much data per second.

## Speed or latency: which matters more?

Both, for different things:

- **Speed (MT/s)** sets the bandwidth: how much data flows per second. Video editing, compression, integrated graphics and large data sets benefit from it.
- **Latency (ns)** sets how quickly a single request is answered. Games benefit most, especially at high frame rates.

At the same speed a lower CL is always better. Across speeds, compare the ns values. The difference between 10 and 12 ns is usually small in games; between 10 and 16.7 ns it is clearly noticeable.

## Other timings

CL is only the first of several timings (tRCD, tRP, tRAS and more), usually printed on the module as a series such as 30-36-36-76. For a quick comparison, latency in ns is the best single number; if you overclock, look at the others too.

## What suits your processor

- **AMD Ryzen 7000 and 9000 (AM5):** AMD calls DDR5-6000 the "sweet spot". At that speed the memory controller and the memory run at a 1:1 ratio. Faster kits often add little there; low latency at 6000 MT/s matters more.
- **Intel Core Ultra 200S:** officially up to DDR5-6400; CUDIMM modules allow higher speeds.
- **Older DDR4 platforms:** DDR4-3200 CL16 and DDR4-3600 CL18 (both 10 ns) are the usual good compromises.

Whether a kit runs at its rated speed also depends on the memory profile in the BIOS: see [EXPO and XMP](https://ddrchecker.com/guide/en/expo-xmp/).

## Using the price list

The [price list](https://ddrchecker.com/uk/en/) shows an **ns** column for every listing. Under "More filters → True latency" you can set an upper limit, for example [DDR5 desktop kits of 10.7 ns or less](https://ddrchecker.com/uk/en/#g=5&f=UDIMM,CUDIMM&e=0&l1=10.7). Sorted by price per GB, the cheapest fast kit is then at the top.

## More guides

- [RAM prices now: what does memory cost per GB?](https://ddrchecker.com/guide/en/ram-prices/)
- [EXPO and XMP: why new RAM often runs slower than advertised](https://ddrchecker.com/guide/en/expo-xmp/)
- [Single or dual rank: what 1R and 2R mean for RAM](https://ddrchecker.com/guide/en/single-dual-rank/)
- [DRAM chips: Micron, Samsung and SK hynix](https://ddrchecker.com/guide/en/dram-chips/)
- [Which RAM fits: DDR4 or DDR5, form factor and capacity](https://ddrchecker.com/guide/en/which-ram-fits/)
- [To the UK price list](https://ddrchecker.com/uk/en/)
