Get GPU credit, spin up an H100, start mining PEARL in under 5 minutes. Exclusive perks below ↓
Per-second billing on H100 / H200 SXM. 99.9% uptime, friendly UI. Spin up a PEARL mining pod in under 60 seconds — your free credit covers the first ~3 hours.
Community GPU marketplace. H100 PCIe from $1.20/hr. Pick your host, lock the rate, mine immediately. Best $/PEARL on the market.
Affiliate links — Lord Of Pearls earns a small commission at no cost to you. Funds keep the explorer free + ad-free.
PEARL is a Proof-of-Useful-Work blockchain. Instead of brute-force SHA-256 hashing, miners run large language model inference on a GPU. Same security guarantees as Bitcoin's PoW, but the compute does something useful. Block reward ~2,845 PEARL every ~60 seconds. Read the full PoUW explainer →
Pearl mining uses vLLM as its inference backend, which sets a hard hardware floor:
The gateway is what mines. The other two are dependencies.
For the full hands-on tutorial — including bare-metal commands, monitoring tips, profitability math and common errors — read our deep-dive: How to mine PEARL on a single GPU. It walks through everything from curl -LO pearl-linux-x64.tar.gz to your first BLOCK FOUND log line.
Watch the network state page for current hashrate and block reward. Multiply (your hashrate / network hashrate) × 3,456 blocks/day × current block reward to estimate your daily PEARL. Multiply by current PEARL price (USD) to get gross revenue. Subtract GPU rental cost ($48/day for an H100 at $2/hr). Profitable months and unprofitable months alternate as the network grows — single-GPU solo mining is feast-or-famine. Read the article above for the full framework.
Once you've mined your first block, paste your prl1… address into our homepage search bar and you'll get a dedicated wallet page with balance, mined blocks, orphans lost, miner rank vs every other Pearl miner, and full transaction history. It's the fastest way to see whether your setup is working.
An NVIDIA H100 (80 GB) is the minimum. H200 (141 GB) is recommended for headroom. Consumer cards (RTX 4090 and below) are not supported because the Pearl mining kernel relies on Hopper-only SM 9.0 features. You'll also need 64 GB system RAM, a 200 GB SSD, and 100 Mbps network.
No. PEARL uses a Proof-of-Useful-Work scheme that requires GPU-accelerated LLM inference. CPU mining is not supported.
An H100 on RunPod runs about $1.50–$2.50 per hour (≈$48/day). On Vast.ai you can find the same hardware for 30–50% less. Profitability depends on the network hashrate, your share of it, the current PEARL price, and the block reward (currently around 2,845 PEARL per block). Run the math before you start — see the profitability section above.
Approximately 2,845 PEARL per block, distributed to the miner's coinbase address. The chain produces roughly one block every 60 seconds. Block reward decreases over time on a Bitcoin-style halving schedule.
Watch your gateway logs for "share submitted" and "BLOCK FOUND" lines. Then paste your prl1… address into the search bar at the top of lordofpearls.xyz. You'll see your live block count, balance, miner rank, and any orphans you've lost. If blocks_24h is zero but you're submitting shares, your setup is finding shares but not winning blocks — usually a sign of insufficient hashrate.
An orphan block is a valid block that lost a chain race because another miner found a competing block at the same height almost simultaneously. Some orphan rate is unavoidable — it's a function of network propagation speed and block frequency. If your orphan rate is significantly higher than the network average (visible on /orphans), your gateway's network latency or peer count probably needs tuning.
The full Pearl node and gateway binaries are released on the official Pearl GitHub. The vLLM inference server is a separate Python package installed via pip install vllm. The complete setup walkthrough is in our how-to-mine guide.