Benchmarking XMRig on Your Own CPU

Measure your own hardware

We do not publish a hashrate table. Numbers collected from other people's machines mix different memory kits, BIOS settings, cooling and Windows power plans, and the same CPU model can land 30% apart on two desks. The benchmark that matters is the one you run yourself, on the machine you will actually mine with.

Running the built-in benchmark

cd C:\Mining\xmrig
xmrig.exe --bench=1M
xmrig.exe --bench=10M

The short run gives a quick figure; the longer one is closer to sustained behaviour because the CPU has time to reach its steady thermal state. Run both as administrator so large pages are available — a benchmark without them measures a configuration you would not use.

Reading the result

• Compare the fifteen-minute average, or the long benchmark, never the first ten seconds.
• Repeat each measurement. A single run that differs by a few percent from the next is noise, not an improvement.
• Change one thing between runs — thread count, affinity, page setting — or you will not know what caused the difference.
• Close the browser and background applications; they compete for the same cache.
• Note the room temperature. Summer results and winter results are not comparable.

What usually moves the number

Change	Typical effect
Enabling large pages	The largest single gain on most Windows systems.
Matching threads to L3 cache	Often faster than using every thread.
Faster or dual-channel memory	Helps RandomX consistently.
MSR tuning on supported CPUs	Noticeable, requires administrator rights.
Better cooling	Protects the number you already have over long runs.

Method and reasoning behind these are in the performance guide.

Benchmark questions

Does the benchmark mine?

No. It runs the algorithm locally and reports a hashrate; nothing is submitted anywhere.

Why is my result lower than tables I find online?

Usually large pages are off, memory is single-channel, or the CPU is throttling. Check those three first.