Timings
RAM timings explained: CL, tRCD, tRP, tRAS and the rest
RAM timings are small delays, counted in clock cycles, that the memory needs between steps when it reads or writes data. The four numbers on a kit, like 30-36-36-76, are the main ones: CL, tRCD, tRP and tRAS. Lower is quicker at the same speed, but you can only compare timings fairly once you turn them into nanoseconds, because a clock cycle gets shorter as the speed goes up.
For buying RAM, the speed and the first two numbers tell you almost everything. The rest are set for you when you turn on XMP or EXPO.
What do the four numbers mean?
Memory chips store data in a grid of rows and columns. To read something, the memory first opens the right row, then picks the column, then closes the row before it can open a different one. Each timing is the wait at one of those steps.
| Timing | Full name | What it measures, in plain words |
|---|---|---|
| CL | CAS latency | From asking for a column to getting the first data, when the row is already open |
| tRCD | RAS to CAS delay | From opening a row to being able to read from it |
| tRP | Row precharge | Closing the current row so a different one can open |
| tRAS | Row active time | The minimum time a row has to stay open |
So a kit labeled DDR5-6000 30-36-36-76 runs at 6000 MT/s with CL 30, tRCD 36, tRP 36 and tRAS 76. Some kits list a fifth value, the command rate (1T or 2T), which is how many cycles the controller waits before sending a command. Your board picks it automatically.
Why are DDR5 timings so much higher than DDR4?
Because DDR5's clock cycles are much shorter. Timings count cycles, not time. A faster kit has more cycles per second, so the same real delay takes more of them.
To compare kits fairly, convert CL into nanoseconds:
Latency in ns = CL × 2000 ÷ speed in MT/s
| Kit | Math | First-word latency |
|---|---|---|
| DDR4-3200 CL16 | 16 × 2000 ÷ 3200 | 10.0 ns |
| DDR5-6000 CL30 | 30 × 2000 ÷ 6000 | 10.0 ns |
| DDR5-6000 CL36 | 36 × 2000 ÷ 6000 | 12.0 ns |
| DDR5-6400 CL32 | 32 × 2000 ÷ 6400 | 10.0 ns |
| DDR5-7200 CL34 | 34 × 2000 ÷ 7200 | 9.4 ns |
A good DDR5 kit responds about as quickly as a good DDR4 kit did, and moves far more data per second. That's why CL30 on DDR5 isn't "worse" than CL16 on DDR4.
The same formula works for tRCD and tRP. A 6000 kit with tRCD 36 waits 12 ns to read from a newly opened row; one with tRCD 40 waits about 13.3 ns.
Which timing matters most?
CL gets all the attention, but it only covers the best case, when the data sits in a row that's already open. Plenty of reads miss, and then the memory has to close the old row (tRP), open the new one (tRCD), and only then wait CL. That's why two kits with the same CL can feel different:
- DDR5-6000 30-36-36-76: the usual "good" 6000 kit.
- DDR5-6000 30-40-40-96: same CL, but slower to switch rows.
When you compare kits at the same speed, look at CL first and tRCD second. tRAS barely matters for buying; it's set from the others.
How much does any of it change in practice? Not a lot. Going from CL36 to CL30 at 6000 is a few percent in CPU-heavy games and close to nothing at 4K. Our guide on CL30 vs CL32 vs CL36 has the details.
What are secondary and tertiary timings?
Behind the four main numbers sit dozens more. The kit's XMP or EXPO profile sets some of them, and the motherboard picks the rest automatically. You never have to touch them, but a few names come up a lot:
- tRFC (refresh cycle time): memory cells leak and have to be refreshed regularly, and nothing else can happen on that part of the chip while it refreshes. tRFC is how long a refresh takes. Out of the box it's usually set very loose, which is why people tuning their RAM start here. Overclockers UK calls lowering tRFC the biggest single reduction in latency you can get from secondary timings.
- tREFI (refresh interval): how often refreshes happen. A higher number means fewer interruptions and more speed, but memory that runs hot needs refreshing more often, so pushing it too far causes errors.
- tFAW and tRRD: limits on how quickly the memory can open rows one after another. Tightening them helps when lots of rows are opened in a burst.
- tWR and tRTP: how long the memory waits after writing or reading before it can close a row.
- tRC: a full row cycle, open plus close. It's roughly tRAS plus tRP.
These matter if you tune your RAM by hand. They're also the reason two kits with identical labels can perform slightly differently: one maker's profile may set tighter values underneath. Reviews that test real latency, rather than just listing the label, catch that.
Does the memory controller change the picture?
Yes, a little. The CPU's memory controller has its own clock, and on DDR5 it often runs at half the memory's speed. On AMD Ryzen, DDR5-6000 is popular because the controller can still run at full speed with the memory; above about 6400 it usually drops to half speed, which adds latency that tighter timings can't win back. We explain that in Gear 2 and UCLK 1:2.
How do you see your RAM's timings?
- CPU-Z, Memory tab: shows the current speed (as half the MT/s, in MHz) and the main timings. If it shows 2400 MHz on a DDR5-6000 kit, XMP or EXPO isn't on.
- CPU-Z, SPD tab: lists the profiles stored on the sticks, including XMP and EXPO.
- ZenTimings (AMD) or ASUS MemTweakIt: show the secondary timings too, including tRFC and tREFI.
- The BIOS: the memory or DRAM timing menu lists everything, usually with "Auto" next to the values the board chose.
The short version
- Timings are delays in clock cycles. Turn CL into nanoseconds (CL × 2000 ÷ speed) before comparing kits at different speeds.
- When comparing kits at the same speed, check CL first, then tRCD.
- DDR5's higher numbers aren't worse: DDR5-6000 CL30 responds as quickly as DDR4-3200 CL16.
- Secondary timings like tRFC are set for you. Only touch them if you're tuning by hand.
- For gaming, speed and timings make a few percent of difference. Capacity and turning on XMP or EXPO matter more.