Do more CPU cores always mean a higher RandomX hashrate?

No. More CPU cores can give you a higher RandomX hashrate, but only if those cores have enough cache, memory bandwidth, and power to work efficiently. A CPU with fewer strong, well-supported cores can beat one with more cores that spend time waiting for data.

Cores are only part of the picture

RandomX is the mining algorithm, or set of calculations, that Monero uses. It favors CPUs, but it does more than ask each core to repeat a simple calculation.

Your hashrate is the number of mining calculations your device completes each second. More cores can handle more calculations at once—until another part of your system becomes the limit.

Also, check what a processor’s advertised numbers mean:

• Physical cores are the actual processing units inside your CPU.
• Threads are work streams those cores handle. Technologies such as simultaneous multithreading let one physical core handle two threads.

A 16-core, 32-thread CPU does not have 32 full cores. Running 32 mining threads will not automatically double its output compared with running 16.

Cache can limit useful thread counts

Cache is small, fast memory built into your processor. RandomX needs about 2 MB of L3 cache per mining thread for efficient operation. L3 is a cache level that is often shared by several cores.

For example, a CPU with 16 MB of L3 cache has a rough cache budget for eight mining threads. It might support more threads, but enabling all of them can create competition for that cache rather than increase useful work.

That is a starting estimate, not a guaranteed best setting. Cache layout matters too. On some processors, L3 cache is split into groups, so every core cannot access every portion equally efficiently.

This is why you should compare cache capacity and layout, not just the core count printed on the box.

Memory, power, and cooling set the next limits

RandomX’s fast mode uses a dataset of about 2,080 MB. Its light mode uses a 256 MB cache, but is much slower. Fast mode also needs memory beyond the dataset for other working data.

Having enough RAM does not mean every extra core can get data quickly enough. Two important limits are:

• Memory bandwidth: how much data RAM can deliver each second.
• Memory latency: how long a core waits for requested data.

Large servers can run into these limits despite having many cores. On servers with multiple CPU sockets, accessing memory attached to another socket can also reduce performance.

Power and cooling matter as well. When you activate more cores, the processor may lower its running speed to stay within power or temperature limits. Extra threads can then add little hashrate while increasing electricity use.

Test the thread count you actually need

Start with a moderate mining thread count, then increase it in small steps. Keep each test long enough for temperatures and performance to settle.

Watch your:

• Total hashrate.
• Electricity use.
• CPU temperature.
• Computer’s responsiveness.

For browser mining, test in the actual browser miner you plan to use. Browser limits and the miner’s implementation can prevent it from matching dedicated mining software. A dedicated-software benchmark is not a promise of browser performance.

The bottom line

More cores help only while the rest of your system can keep them busy. For RandomX, choose and tune your CPU around cache, memory, cooling, and measured performance—not the largest core count alone.