argfa/pearl-gpu repository overview

⁠ArgfaMining Pearl (PRL) GPU Miner — NAUTILUS

Official GPU miner for the Pearl (PRL) PoUW network, built for the ArgfaMining⁠ pool. Native multi-GPU — one process drives every supported NVIDIA card in the rig.

⁠What's new in v1.1.3

• The hashrate no longer drops to almost zero on long runs. After a long session (about 42 hours on an RTX 5070, longer on slower cards) the hashrate counter overflowed, so the miner reported a few TH/s while the pool kept seeing the real speed. Mining and shares were never affected.
• The hashrate in the [TOTAL] line is now the last 5 seconds instead of the average of the whole session, so a power-limit change, thermal throttling or a lost connection shows up right away.
• Still mines at rank 128, which Pearl requires since block 96,251 (new in v1.1.2). v1.1.1 and older mine at rank 512 and lose 3 of every 4 shares on pools that apply the rule (ArgfaMining included). If you run v1.1.1 or older, update.

⁠Quick start

docker run --rm --gpus all \
  -e WALLET=prl1youraddress \
  -e REGION=latam \
  -e MODE=solo \
  argfa/pearl-gpu:1.1.3

⁠Configuration (environment variables)

Regions: latam (Venezuela) · usa (Virginia) · eu (Frankfurt) · india (Mumbai).

⁠Multi-GPU

By default the miner uses all supported GPUs. To pin specific cards, append --devices:

docker run --rm --gpus all -e WALLET=prl1... argfa/pearl-gpu:1.1.3 --devices 0,1

Any extra arguments after the image name pass straight to the miner.

⁠Requirements (host)

• NVIDIA driver (CUDA 13.x) + nvidia-container-toolkit (pre-installed on vast.ai / RunPod / octa.space).
• GPU: NVIDIA Ampere (RTX 30xx, A100) / Ada (RTX 40xx) / Hopper (H100, H200) / Blackwell (RTX 50xx, B200, B300) — sm_80+. Turing (RTX 20xx) and AMD/Intel are not supported.
• The CUDA runtime is static in the binary; the base is a tiny ubuntu:22.04 and the driver is injected at runtime by --gpus.

⁠Notes

• Tags: 1.1.3 (pinned) · latest.

Support: https://discord.gg/nZTu2QC2YC⁠