The RAM latency calculator turns CAS latency and speed into true latency in nanoseconds: CL × 2000 ÷ speed in MT/s. Got a RAM kit you want to check? Put its CAS latency and speed into the HypestKey RAM latency calculator above. Out comes true latency in nanoseconds. How fast the memory actually responds, in other words. Way more useful than the CL number on the box.
| You get | What it is |
|---|---|
| True latency | In nanoseconds |
| Rating | Excellent, Good, Average or Consider Upgrading |
| One clock cycle | Its length in ns |
| Real clock | In MHz, half the MT/s figure |
It works for DDR3, DDR4 and DDR5 at any speed. Runs in your browser. The number shows once you press Calculate Latency. Not sure how much memory to buy in the first place? The RAM calculator answers that part.
How to use the RAM latency calculator
Two numbers from the sticker: CAS latency and speed in MT/s. DDR4-3600 CL16, for example, comes out at 8.89 ns. Under 9 ns rates Excellent, under 11 ns Good.
- Find the CAS latency, CL for short, and the speed of your kit. Both are on the sticker or in the model name. DDR4-3600 CL16, say.
- Type the CL into the CAS Latency field. The tool accepts values from 1 to 100.
- Type the speed in MT/s into the RAM Speed field, for example 3600. Values from 100 to 20000 work.
- Press Calculate Latency, or Enter in a field. The result shows on the right, and a wrong entry gets explained.
- Compare your kit with the Common RAM Configurations list under the big number. Or tap a Quick Preset, then press Calculate Latency. DDR5-6000 CL30, for instance.
- Press Copy result to save the numbers, or Reset to clear both fields.
What latency is
The delay between the CPU asking memory for data and the data arriving. Measured in nanoseconds. Your CPU sends millions of requests per second. Tiny waits add up quickly.
A typical gaming kit such as DDR4-3600 CL16 has a CAS latency of about 8.9 ns. DDR5-6000 CL30 sits at 10 ns.
On RAM boxes you see specs like CL14, CL16 or CL30. Marketing loves these numbers. Problem is, CL by itself tells you almost nothing, because CL counts clock cycles, not time.
The same CL16 on a 3200 MHz stick and on a 2400 MHz stick gives totally different real world delays. 10 ns against 13.33 ns.
True latency solves this. CL and speed go in, nanoseconds come out. That works across DDR4, DDR5 and any speed rating.
The formula
The standard one: true latency in ns = CL × 2000 ÷ speed in MT/s. Why 2000? DDR moves data on both clock edges. The real clock is half the MT/s figure.
| Step | What the tool does |
|---|---|
| 1 | Checks that CL is between 1 and 100 and speed between 100 and 20000 MT/s |
| 2 | Calculates one clock cycle as 2000 ÷ speed: 0.556 ns at 3600 MT/s |
| 3 | Multiplies the cycle by CL: 16 × 2000 ÷ 3600 = 8.89 ns for DDR4-3600 CL16 |
| 4 | Rates the result against fixed bands |
| True latency | Rating in the tool |
|---|---|
| Under 9 ns | Excellent |
| 9 to under 11 ns | Good |
| 11 to under 14 ns | Average |
| 14 ns and above | Consider Upgrading |
DDR5-6000 CL30 seems worse on paper. Yet 30 × 2000 ÷ 6000 = 10 ns. Same as DDR4-3200 CL16.
One honest limit. This is CAS latency only, the first word delay inside the memory chip. Other timings, the memory controller and the CPU cache add more.
- tRCD.
- tRP.
- tRAS.
A full memory test such as AIDA64 usually reports a total of 60 ns or more. Comparing two kits? We’d still call the CAS number the fair and simple yardstick.
DDR5 against DDR4
DDR5 isn’t really worse, once you count in nanoseconds. Its CL numbers are higher, often CL30 to CL40. But it also runs at 4800 to 8000+ MT/s.
DDR5 specs confuse people. CL36? CL40? Looks awful next to DDR4. Faster clocks make up for the bigger CL values, so a bunch of DDR5 setups end up matching or beating DDR4 latency.
| Kit | True latency | Note |
|---|---|---|
| DDR4-3200 CL14 | 8.75 ns | Usually Samsung B-die |
| DDR4-3600 CL16 | 8.89 ns | Popular gaming kit |
| DDR4-3200 CL16 | 10 ns | Common budget kit |
| DDR4-3600 CL18 | 10 ns | |
| DDR5-6000 CL30 | 10 ns | Ryzen sweet spot |
| DDR5-6400 CL32 | 10 ns | |
| DDR5-7200 CL34 | 9.44 ns | High-end kit |
| DDR5-4800 CL40 | 16.67 ns | Base JEDEC speed |
DDR5-7200 CL34 beats most DDR4 kits you can buy. Only slow DDR5-4800 CL40 clearly loses to budget DDR4.
Latency and gaming FPS
It depends on where your PC is limited. In CPU-limited games lower latency can add a few percent of FPS and smoother 1% lows. At 4K with a strong graphics card? The GPU usually hits its limit first. Latency barely matters.
The CPU-limited kind:
- High refresh shooters at 1080p.
- RTS games with tons of units.
- Flight sims.
Faster RAM gets data to your CPU with less waiting. More frames, smoother gameplay in those scenarios.
But here’s the thing. With a beefy graphics card at high resolution your framerate is already capped by graphics horsepower. Most modern AAA titles at high settings lean on the GPU too.
Unsure which part holds you back? Run your parts through the bottleneck calculator or the FPS calculator.
Workstation users, this matters to you too. Compiling code, running databases and scientific computing constantly shuffle data between CPU and RAM, and faster memory speeds them up noticeably. Video editing is different: capacity counts more than speed. On Premiere or Resolve we’d get 32 GB before worrying about latency.
| Use case | Rough target |
|---|---|
| Esports and competitive gaming | Shoot for under 9 ns |
| General gaming at 1080p or 1440p | Under 11 ns works great |
| Tight budget builds | Anything under 12 ns is perfectly fine |
| Workstation stuff | Under 10 ns helps, but get enough capacity first |
Checking your speed and CL
The fastest way is free CPU-Z. The Memory tab shows the real clock and the CL. The SPD tab lists the rated timings.
| Kit | Real clock in CPU-Z | A sign that XMP or EXPO is probably off |
|---|---|---|
| DDR4-3600 | 1800 MHz, half the MT/s figure | Around 1066 to 1200 MHz on a faster DDR4 kit |
| Fast DDR5 | Half its MT/s figure | 2400 MHz on a fast DDR5 kit |
Look at the sticker on your RAM, too. The model number usually includes speed and CL rating.
Prefer not opening your PC? Get CPU-Z, a free program. Memory tab for speed, SPD tab for timings.
Task Manager shows speed under Performance, then Memory. No latency data there. For timings you need CPU-Z or the BIOS.
Before anything else, enable XMP on Intel or EXPO on AMD in the BIOS. Most RAM ships running slow default settings until you do.
Fail-safe first, speed second. Per Intel’s XMP page, an XMP kit boots at standard JEDEC settings. The profile gets picked in the BIOS after that. Not sure how to get into it? Our guide on how to enter the BIOS from Windows lists the keys.
| Step | Turning it on |
|---|---|
| 1 | Restart and press Delete or F2 during boot |
| 2 | Look for an XMP, DOCP or EXPO option |
| 3 | Enable it and save changes |
| 4 | Check in CPU-Z that the real clock is half of the rated MT/s |
RAM speed and part number in Windows
No CPU-Z at hand? Windows reads the module data too. PowerShell, then this:
Get-CimInstance Win32_PhysicalMemory | Select-Object Manufacturer, PartNumber, Capacity, ConfiguredClockSpeed
One line per stick. The PartNumber is the useful bit for latency. Search that code on the maker’s site and the spec sheet shows the rated speed and CL, ready for the calculator.
ConfiguredClockSpeed is the speed setting the memory runs at right now. Microsoft documents these fields in its Win32_PhysicalMemory reference. For the live CL, CPU-Z stays the clearer check.
CAS latency calculator in reverse
Turn the formula around and it answers a buying question. What CL does a faster kit need to match the latency you have now? CL = target ns × speed in MT/s ÷ 2000.
Round the result down to a whole CL. The table shows the highest CL that still hits 10 ns or 9 ns.
| Speed | Highest CL for 10 ns | Highest CL for 9 ns |
|---|---|---|
| DDR4-3200 | CL16 | CL14 |
| DDR4-3600 | CL18 | CL16 |
| DDR4-4000 | CL20 | CL18 |
| DDR5-5600 | CL28 | CL25 |
| DDR5-6000 | CL30 | CL27 |
| DDR5-6400 | CL32 | CL28 |
| DDR5-7200 | CL36 | CL32 |
| DDR5-8000 | CL40 | CL36 |
Moving from DDR4-3600 CL16 to DDR5? To keep 8.9 ns, look at DDR5-6000 at CL27 or below. Or DDR5-7200 at CL32. A CL30 kit at 6000? 10 ns, a hair slower to answer, with far more bandwidth.
Building the whole PC from scratch? The PSU calculator sizes the power supply. Check the board and CPU against the Windows 11 requirements too. A fresh build needs a license as well: our Windows 11 Pro key or Windows 11 Home key, instant to email. More free checks sit on the tools page.
Reading your result
| You see | What we’d do |
|---|---|
| Under 9 ns, Excellent | Nothing. That kit is as quick as it needs to be |
| 9 to under 11 ns, Good | Fine for gaming and work; leave it alone |
| 11 to under 14 ns, Average | Check that XMP or EXPO is on before blaming the kit |
| 14 ns and above | The tool says Consider Upgrading, and we’d agree if the board takes a faster kit |
| A message about a wrong entry | CL has to be 1 to 100, speed 100 to 20000 MT/s |
One caution. Type the numbers your RAM actually runs at, not just the ones on the sticker. With XMP off they differ.
What to buy in 2026
| Platform | Pick | True latency | Note |
|---|---|---|---|
| Older, DDR4 | DDR4-3600 CL16 | 8.89 ns | The safe pick; CL14 kits hit 8.75 ns but cost more |
| New, DDR5 | DDR5-6000 CL30 | 10 ns | The safe pick, with good bandwidth |
| Either | DDR5-4800 CL40 | 16.67 ns | Avoid if you can: slower than budget DDR4 |
DDR5-7200 kits are worth a look too. But memory prices move a lot, so we’d compare price per kit before paying extra for 0.5 ns.
Platform tips:
- Ryzen: DDR4-3600 or DDR5-6000 for Infinity Fabric sync.
- Intel is more flexible with frequencies.
- Two sticks always. Dual channel doubles bandwidth.
- Matched kits only. Random sticks cause issues.
Mistakes people make all the time
| Mistake | What goes wrong |
|---|---|
| Buying high MHz without checking CL | DDR4-4000 CL20 sounds fast but works out to 10 ns, while DDR4-3200 CL14 at a lower clock hits 8.75 ns |
| Forgetting XMP or EXPO | Fast RAM runs at slow defaults until you enable it in the BIOS |
| A single stick setup | One 32 GB module has half the bandwidth of two 16 GB modules |
| Mixing mismatched kits | The motherboard forces both kits to run at the slower one’s speed and timings; sometimes it refuses to boot at all |
Each of these can cost you more speed than the kit’s price difference. Always run the numbers here first.
Two sticks go in the A2 and B2 slots, usually.
Bought 16 GB two years ago, now adding another 16 GB from a different brand? Buy a matched set if you need more capacity. Weighing an upgrade against a new build? Check the PC price calculator.
Words on this page
| Word | Short explanation |
|---|---|
| CL | CAS latency, counted in clock cycles |
| MT/s | The speed on the sticker; the real clock in MHz is half of it |
| True latency | CL × 2000 ÷ speed, in nanoseconds |
| XMP, DOCP, EXPO | BIOS options that switch a kit to its rated speed |
| JEDEC | The base speed a kit runs at before XMP or EXPO |
| 1% lows | The slowest frames in a game; smoother when they rise |