What RAM timings are best for RandomX?

The best RAM timings for RandomX are the lowest stable timings your CPU and memory can run without giving up too much memory speed. There is no single winning set: DDR4-3200 CL14 and DDR4-3600 CL16 are useful reference points for DDR4 systems, while DDR5-6000 CL30 is a common starting point on compatible DDR5 desktops—not a guarantee of the best result.

Look at latency, not just the CL number

RAM timings describe how long your memory waits before carrying out certain steps. Lower timings can reduce those waits, helping RandomX, Monero’s mining algorithm, access data faster.

You will often see four numbers, such as 16-18-18-36:

• CL: The wait between requesting data from an open memory row and receiving it.
• tRCD: The delay between opening a row and accessing its data.
• tRP: The time needed to close a row before opening another.
• tRAS: The minimum time a row must stay open.

These numbers count memory clock cycles, not a fixed amount of time. You must consider memory speed too. For example, DDR4-3200 CL16 and DDR4-3600 CL18 both have a theoretical CL delay of about 10 nanoseconds. That does not mean their overall mining performance will be identical.

Choose a sensible starting point

For a desktop PC, start with the memory’s tested profile rather than copying someone else’s manual settings. These profiles are usually called XMP or EXPO and are selected in your motherboard’s firmware settings, often called the BIOS.

Useful starting points include:

• DDR4: A stable 3200 CL14 or 3600 CL16 setup, if your hardware supports it.
• DDR5: A supported low-latency kit, such as 6000 CL30 on many compatible desktops.
• Servers: The manufacturer-supported speed and timings for your installed memory arrangement.

Do not force these desktop targets onto a server. Server firmware may lock timing controls, and filling more memory slots can lower the supported speed.

On some CPUs, pushing RAM faster changes the relationship between the memory clock and its controller—the part of the CPU that manages RAM. That can increase delays and erase the benefit of higher speed.

Tighten secondary timings carefully

After the main timings, tRFC, the time reserved for refreshing memory cells, can be worth testing. A lower stable tRFC may improve performance, but setting it too low can cause errors.

Other secondary timings control spacing between memory operations. Their best values depend heavily on the memory chips, capacity, motherboard, and CPU.

Change one setting at a time. Avoid treating unusually tight settings from an online screenshot as safe defaults. A system that starts successfully can still produce memory errors under sustained load.

Measure your actual RandomX workload

Keep the mining thread count, cooling, and software unchanged while comparing settings. A thread is one stream of work handled by your CPU.

For each change:

• Run a thorough memory stability test.
• Run repeated RandomX benchmarks after temperatures settle.
• Compare hashes per second—the number of mining calculations completed each second.
• Check power use and errors, not just speed.

If you mine in a browser, test that browser workload too. Browser limits and background activity can hide small timing improvements.

The bottom line

Choose stable, low-latency RAM settings at a speed your CPU handles well. Start with a supported profile, tune gradually, and keep changes only when repeated RandomX tests show a real improvement.