DDR4 vs DDR5 latency: higher bandwidth doesn't always mean lower latency

How this is calculated

CAS latency (CL) numbers are in clock cycles, not nanoseconds. DDR5 runs at higher frequencies, so its cycles are shorter. CL30 at 6000 MT/s is 10 ns. CL16 at 3600 MT/s is 8.9 ns. The DDR4 kit has lower absolute latency. But DDR5 runs at double the transfer rate, so for workloads that are bandwidth-bound (video editing, scientific computing, AI inference), DDR5 wins. For latency-sensitive workloads (games, databases with random access patterns), the difference is small and either generation at the same tier performs similarly. The real gains from DDR5 come from higher capacities per DIMM and better power efficiency, not lower latency.

Verdict

The headline bandwidth jump obscures the latency story. DDR5-6000 CL30 lands at about 10 ns true latency, while DDR4-3600 CL16 hits 8.9 ns, so the older kit is actually faster per access. DDR5 wins by running at double the transfer rate, which serves bandwidth-bound workloads like video editing and AI inference. It also ships higher per-DIMM capacity and better power efficiency, which matters more for servers than for the slight latency cost.