What is the RandomX dataset and how often is it regenerated?

The RandomX dataset is a large block of working data that Monero mining software uses while testing possible solutions for the next block. It is regenerated whenever Monero’s RandomX seed changes—every 2,048 blocks, or roughly 2 days and 20 hours at the average two-minute block time.

What the dataset contains

RandomX is Monero’s mining algorithm: the set of rules your computer follows to do mining work. Monero has used it since November 30, 2019.

Its dataset contains 2,080 MB of data in the memory units used by the RandomX specification—just over 2 GiB. It is not a copy of the blockchain, a list of transactions, or a collection of coins.

Instead, it is a large, repeatable collection of values derived from a seed, a starting value taken from the Monero blockchain. Your mining software uses those values while running the calculations that produce a hash—a digital result tested against the network’s mining target.

Given the same seed, correctly working miners generate the same dataset.

When the seed changes

Monero changes the RandomX seed on a schedule measured in blocks, not hours:

• A new seed is selected every 2,048 blocks.
• The seed comes from the hash of a designated block.
• That seed becomes active 64 blocks later, roughly two hours and eight minutes on average.

That 64-block delay gives the seed block time to settle into the blockchain before miners depend on it. It also gives mining software an opportunity to prepare the next dataset.

The delay does not make the regular refresh interval 2,112 blocks. Once the schedule is running, changes remain 2,048 blocks apart.

Because blocks do not arrive at perfectly even intervals, the actual time between dataset changes varies. Monero’s mining difficulty—the measure of how hard it is to find an acceptable result—adjusts every block, but the dataset does not regenerate every block.

What your computer does at a change

In RandomX’s fast mode, your mining software builds the full dataset in RAM, your computer’s short-term working memory. It then reuses that dataset across many mining attempts until the seed changes.

Building a replacement takes processor time and memory activity. Depending on your software and available RAM, you may see a brief pause or lower mining speed. Software that prepares the replacement ahead of time may make the transition smoother, but temporarily needs more memory.

Restarting your miner can also cause it to rebuild the dataset. That is a local rebuild, not a change to the network’s seed schedule.

Why this matters for browser mining

RandomX also supports light mode, which uses a 256 MB cache, a smaller starting collection of data. It calculates needed dataset values on demand rather than keeping the whole dataset ready in memory. This saves RAM but is much slower for mining.

For browser mining, the distinction matters. A browser must work within memory and execution limits, and the full dataset is only part of the miner’s total memory use. You should not assume a mining page can deliver native mining software’s performance.

The bottom line

The RandomX dataset is reusable mining work data. Its contents change every 2,048 blocks, roughly 68 hours, with each new seed activated after a 64-block delay. Your miner normally handles generation automatically; it does not need to rebuild the dataset for every block or mining attempt.