RAM Latency Calculator
Enter DDR speed and CAS latency to get latency in nanoseconds from CL × 2000 / MT/s, plus peak bandwidth. This does not measure the installed RAM. A higher box number is not automatically faster once latency is included.
Latency calculator
The number after DDR5-, e.g. 6000 for DDR5-6000.
The first timing, e.g. 30 in 30-38-38-96.
Low: calculated first-word latency in a range common on gaming kits. Latency in ns = CL × 2000 / MT/s = 30 × 2000 / 6000 = 10.00 ns. One cycle is 0.333 ns. One 64-bit channel is 6000 MT/s × 8 bytes = 48.0 GB/s peak. Dual channel is twice that peak (96.0 GB/s), not a promise that games run twice as fast, and not a benchmark result.
Extra timings (optional)
Row to column delay
Row precharge
Row active time
Compare kits
| Kit | MT/s | CL | Price | Latency | 2-ch peak | Price per GB/s | Remove |
|---|---|---|---|---|---|---|---|
| #1 | 8.89 ns | 57.6 GB/s | -- | ||||
| #2 | 10.00 ns | 96.0 GB/s | -- | ||||
| #3 | 12.00 ns | 96.0 GB/s | -- |
Up to 4 kits, saved in this browser. Price per GB/s appears only after you enter a price. It divides that price by an assumed dual-channel peak (one 64-bit channel is MT/s × 8 bytes, times two). That peak is not measured bandwidth, and it does not mean a game runs twice as fast.
What CL would match?
Same formula rearranged: CL = latency in ns × MT/s / 2000, rounded up to a whole CL. Many kits only use even CL values, so you may need to round up again.
About RAM speed and latency
MT/s vs MHz: the box is doubling
DDR stands for double data rate: it moves data on both edges of the clock, so DDR5-6000 runs a 3000 MHz clock and makes 6000 million transfers per second. Many boxes and spec sheets print "6000 MHz" when they mean 6000 MT/s. Neither number alone tells you speed. Latency in cycles, the CL, often grows as the clock gets faster, so you need both for the delay in nanoseconds: latency in ns = CL × 2000 / MT/s. Some listings call that result true latency. It is the formula above, not a stopwatch on your RAM.
Why DDR5-6000 CL30 is not twice as fast
DDR4-3600 CL16 works out to 8.89 ns. DDR5-6000 CL30 is 10 ns, and DDR5-6000 CL36 is 12 ns. So the newer, bigger-number kit is not quicker to deliver its first word of data on this formula. Where it wins is peak bandwidth: 96 GB/s for two 64-bit channels against 57.6 GB/s. One channel is MT/s × 8 bytes. Dual channel is that figure times two, only as a peak-transfer number, not a promise that games run twice as fast. A higher number on the box is not automatically faster once latency is included. A DDR4-4000 CL22 kit (11 ns) is slower on this formula than DDR4-3200 CL14 (8.75 ns).
When latency or bandwidth matters
Games make many small, scattered memory reads, so they lean on latency, especially CPUs with smaller caches. Chips with a large stacked cache (AMD X3D) care much less. Video editing, rendering, compression and integrated graphics move large blocks and scale with bandwidth. Many DDR5 gaming builds use 6000 to 6400 MT/s at CL30 to CL32. A faster rating can raise peak bandwidth without lowering this calculated latency, and this page cannot tell you the change in frame rate.
XMP and EXPO are off by default
The speed on the box is an overclock profile. Out of the box, most boards run the kit at a slower JEDEC default, such as DDR5-4800 CL40 (16.67 ns on the formula above). Enable XMP (Intel) or EXPO (AMD) in the BIOS to get the advertised numbers, then check the running speed in Task Manager or CPU-Z. This calculator cannot see or change that.
CAS latency is only one step
CAS latency only covers one step inside the DRAM. A full memory access, often about 60 to 90 ns in tools such as AIDA64, also waits on the memory controller, the CPU ring or mesh, AMD Infinity Fabric, tRCD and other subtimings, and cache misses. Use this calculator to compare kit labels with the same formula, not to predict a benchmark.
Related tools
Check browser memory bandwidth with the RAM test (a JavaScript loop, not an error test and not this formula). Load the CPU with the CPU stress test, or the GPU with the GPU stress test (WebGL frames on that page, not a game score).
Frequently asked questions
Does this measure the RAM in my computer?
No. You type the speed in MT/s and the CAS latency from the kit label, the box, or a spec sheet. The page applies a formula. It does not read the installed sticks, the BIOS, or whether XMP or EXPO is on.
What formula does the latency number use?
Latency in ns = CL × 2000 / MT/s. DDR moves data twice per clock, so one cycle is 2000 / MT/s nanoseconds. Multiply that cycle by the CL number. DDR5-6000 CL30 is 10 ns. DDR4-3600 CL16 is 8.89 ns.
Is a higher number on the box always faster?
No. A higher box number is not automatically faster once latency is included. DDR5-6000 CL30 has more peak bandwidth than DDR4-3600 CL16, but its calculated first-word latency is higher (10 ns versus 8.89 ns).
How is the bandwidth figure worked out?
One 64-bit channel is MT/s × 8 bytes, shown in GB/s. Dual channel is that figure times two, and only as a peak-transfer number. It is not a promise that games run twice as fast, and it is not what a benchmark will show. The page cannot see how many channels your PC is using.
When does price per GB/s show up?
Only after you type a price for that kit, and only when the data rate is valid. An empty price stays as a dash. The price can be any currency. It is divided by the assumed dual-channel peak, not by a measured transfer rate.
Why is this lower than the latency in AIDA64?
CAS latency is one step inside the memory chip. A full access also waits on the memory controller, the CPU interconnect, other timings such as tRCD, and cache misses. Those tools often land around 60 to 90 ns. Use this page to compare kit labels with the same formula, not to predict a benchmark.
What if I enter 0 or leave a field blank?
A blank, zero, or non-numeric MT/s or CL shows a dash, and so does a value outside the allowed range. You will not see Infinity. The same dash is used when a kit row is incomplete.