Five finalists.One honest number.
We run the same GPU on the same overclock across every Pearl pool at once and measure what actually lands in the wallet — not what a block explorer claims. Daily results posted in the open, then filed as permanent history.
The divergence
Identical rigs should pay nearly the same. They don't. These daily figures — credited PRL (balance + payouts) and 24h pool hashrate per pool — are the only measured inputs; everything in The board below is computed from them. Switch the metric to see day-by-day income or pool hashrate.
Cumulative PRL earned · per pool
The board
So who actually pays best? Ranked by total earnings over the window, straight from the daily figures above. Hashrate shows what the miner produces versus what the pool credits back (the gap is hashrate the pool "lost"); PRL / TH·day is what the wallet was credited per TH the card actually produces — a pool can’t inflate it by mis-reporting hashrate.
| # | Pool | Miner · ver | Hashrate (m→p) | Fee (pool·dev) | PRL / TH·day | Earned Last Day | Total |
|---|
Stage 2 — Key takeaways
Fourteen days measured. Here's what the numbers support — and what they don't.
- pearlhash finished first, but it's a statistical tie at the top. pearlhash paid 32.45 PRL and Kryptex 31.44 PRL — a 3.2% gap that's smaller than a single card's day-to-day variance on a PROP pool (see below). We don't call that a win.
- PPS-style pools pay a stable, reproducible number. Fixed-per-share schemes barely fluctuate day to day, while PROP payouts behave like a lottery until you average many cards or a long window — a real difference in predictability on a small fleet.
- Identical GPUs on pearlhash.xyz produce different results. Run several identical cards on pearlhash.xyz under different wallets and each ends the same period with a different balance. Our control: three identical GPUs over the same 6-hour window paid 0.64 / 0.80 / 0.91 PRL — a 42% spread, roughly ±9% per day per card. A single-wallet PROP figure reflects luck more than the pool, which is why the top result is a tie. Details →
- HeroMiners is out for Stage 3 — it paid about 20% below the field over the 14 days.
The protocol
Those numbers only mean something if the test is clean. A benchmark is only worth reading if anyone can reproduce it — and only honest if the pools can't game it.
The bench
Identical by design. The only things that change between rigs are the stratum endpoint and the pool's designated miner. Here's the exact setup, so anyone can rebuild it and check our work.