The Pearl pool that supports both dedicated mining (PRL rewards) and inference serving. Measured: inference at $0.107/1M output tokens on H100. Mining and inference are separate workloads — a GPU does one or the other.
Pearl mining is matrix multiplication — the same math that runs every LLM. We're the pool that turns it into sellable inference.
Measured on H100 NVL: $0.107/1M output tokens (Qwen3-30B-A3B, c=256). Competitive with Together/DeepInfra. Mining and inference are separate tiers — pick one per GPU.
Idempotent, crash-proof ledger — no double-pays, ever. Independent security audit before any real payout. Your earnings are accounted to the satoshi.
Datacenter H100/H200 via our inference gateway; consumer 3090/4090/5090 via standard stratum. The whole swarm, one pool.
Measured PRL throughput per card. Higher-tier cards (H100/H200) also unlock the inference revenue tier.
| GPU | PRL hashrate | Tier |
|---|---|---|
| NVIDIA H200 | ~615 TH/s | mine + inference |
| NVIDIA H100 | ~600 TH/s | mine + inference |
| RTX 5090 | ~350 TH/s | mining · can serve |
| RTX 4090 | ~262 TH/s | mining · can serve |
| RTX 3090 | ~117 TH/s | mining · can serve |
Pick your tier. Point your miner at the pool. You're earning.
Run the Pearl vLLM miner. Your card serves paid inference when there is demand, and falls back to PRL mining when there is not — the gateway switches for you. In private beta — request access.
PEARLD_RPC_URL=https://pool.linkrra.com PEARL_ADDRESS=<your-prl-address> WORKER=<name> # → joins the inference pool; mines PRL when idle
Algorithm pearlhash. Verified working: PeakMiner (NVIDIA, --coin pearl). Which miners actually support pearlhash →
Your card can also register as a GPU serving node for inference — one card serves whatever model fits its VRAM, and a multi-GPU rig can serve a bigger model. How serving nodes work →
peakminer --coin pearl \ --url stratum+tcp://stratum.linkrra.com:5567 \ --user <your-prl-address>.<worker> \ --password x