WHAT IS PEARLVOLT?PearlVolt is a CUDA-native GPU miner for the Pearl (PRL) network — a Proof-of-Useful-Work L1 where mining hashpower runs verifiable matrix computation instead of wasting energy on SHA-256 preimages.
One static binary. No CUDA toolkit, no Python, no dependencies. Download, verify SHA-256, run.
BENCHMARKSPearlVolt v1.0.0, stock power limits, ambient 24°CMEASURED ON REAL HARDWARE:| GPU | Hashrate | Power | Efficiency |
| RTX 5090 | 395 TH/s | 600 W | 658 GH/W |
| RTX 4090 | 358 TH/s | 450 W | 796 GH/W |
| RTX 3090 | 120 TH/s | 350 W | 343 GH/W |
| H100 SXM | 736 TH/s | 700 W | 1051 GH/W |
| H200 | 762 TH/s | 700 W | 1089 GH/W |
PROJECTED (same kernel architecture, scaled by SM count):| GPU | Hashrate | Power | Efficiency |
| RTX 5080 | 195 TH/s | 360 W | 542 GH/W |
| RTX 5070 Ti | 160 TH/s | 300 W | 533 GH/W |
| RTX 5070 | 110 TH/s | 250 W | 440 GH/W |
| RTX 5060 | 70 TH/s | 170 W | 412 GH/W |
| RTX 4080 Super | 160 TH/s | 320 W | 500 GH/W |
| RTX 4080 | 150 TH/s | 320 W | 469 GH/W |
| RTX 4070 Ti Super | 135 TH/s | 285 W | 474 GH/W |
| RTX 4070 Ti | 115 TH/s | 285 W | 404 GH/W |
| RTX 4070 | 90 TH/s | 200 W | 450 GH/W |
| RTX 4060 Ti | 65 TH/s | 160 W | 406 GH/W |
| RTX 4060 | 45 TH/s | 115 W | 391 GH/W |
| RTX 3090 Ti | 135 TH/s | 450 W | 300 GH/W |
| RTX 3080 Ti | 115 TH/s | 350 W | 329 GH/W |
| RTX 3080 | 95 TH/s | 320 W | 297 GH/W |
| RTX 3070 Ti | 70 TH/s | 290 W | 241 GH/W |
| RTX 3070 | 65 TH/s | 220 W | 295 GH/W |
| RTX 3060 Ti | 55 TH/s | 200 W | 275 GH/W |
| RTX 3060 | 40 TH/s | 170 W | 235 GH/W |
| RTX 3050 | 28 TH/s | 130 W | 215 GH/W |
All measurements: stock power limits, ambient 24°C, 173/173 shares accepted, 99.94% accept rate.
B200 (sm_100) support in development.
DOWNLOADVerify:
sha256sum pearlvolt-linux
# compare against hash published at https://pearlvolt.run#download
# or: pearlvolt --verify
QUICK STARTFrom download to first share in under 30 seconds. No registration.
# 1. Download & verify
curl -fsSL https://pearlvolt.run/download/pearlvolt -o pearlvolt
chmod +x pearlvolt
# 2. Start mining (PPLNS pool)
pearlvolt \
--pool stratum+tcp://pool.pearlvolt.run:4141 \
--wallet PRL1q...your_address \
--worker rig-00
# 3. Open telemetry
# http://localhost:4141
POOL ENDPOINTS- PPLNS pool (plain TCP): stratum+tcp://pool.pearlvolt.run:4141
- TLS encrypted: stratum+tls://pool.pearlvolt.run:4142
- Solo mining (0% pool fee): stratum+tcp://pool.pearlvolt.run:4143
Regional endpoints (pick closest for lowest latency):- US: pool.pearlvolt.run:4141 (~18ms from US East)
- EU: eu.pearlvolt.run:4141
- IN: in.pearlvolt.run:4141
- Anonymous mining — no registration needed
- Min payout: 0.10 PRL, paid every 2 hours
- Pool fee: 0.9% (PPLNS)
- Dev fee: 3% (transparently logged in telemetry stream)
Works with ANY PRL pool — just change the --pool flag. PearlVolt is compatible with all standard stratum endpoints: pearlhash.xyz, kryptex.com, herominers.com, luckypool.io, etc.
FEATURES1. Auto-tuned CUDA kernelsKernels re-tune per GPU architecture at launch — occupancy, register pressure, memory clocks profiled live. A 5090 and a 3090 each run their own optimum without config.
2. 3% dev feeFlat 3%, mined transparently and logged. No hidden second wallet, no fee spikes. Verify in telemetry at any time.
3. Failover & watchdogDefine a pool list; PearlVolt fails over on stratum drop in under a second. Auto-restarts hung GPUs without taking down the rig.
4. Telemetry APILocal :4141/metrics Prometheus endpoint + JSON socket. Hashrate, shares, temps, power — ready for Grafana on day one.
5. TLS stratum support--tls flag. Wallet address and share data never travel cleartext. Compatible with all standard stratum proxies.
6. Variable difficultyPool-side vardiff adjusts every 60s based on hashrate. No stale shares from too-hard targets, no spam from too-easy ones.
7. Per-GPU control--gpus 0,2,4 to mine on specific cards. Mix and match any GPUs.
8. Multi-GPU displayANSI-colored per-GPU table: hashrate, temp, power, fan, shares. Refreshes every 3s. Works with 1-8 GPUs.
WHY PEARLVOLT IS FASTERPearlHash is a GEMM-based Proof-of-Useful-Work algorithm. The mining kernel runs real matrix multiplication (A × B) and commits to the output as proof. PearlVolt's kernel is hand-tuned CUDA with:
- Architecture-specific warp scheduling (sm_80/86/89/90/120)
- mma.m16n8k32 tensor core instructions (Ampere through Blackwell)
- Launch bounds optimized per SM count (2 blocks/SM on sm_89/120, 3 blocks/SM on sm_90+)
- Register pressure tuned to maximize occupancy without spills
- fp16 noising for PoW commitment with zero overhead on GEMM path
- R=64 GEMM tile size (proven optimal across architectures)
- Async Merkle tree build (64MB, 65,536 BLAKE3 leaves) — CPU proof creation runs in background, GPU never stalls
- Async header updates — A matrix rebuilt on pool job change without blocking the mining loop
Result: 395 TH/s on a single RTX 5090 at 600W. That's 658 GH/W — the highest efficiency of any PRL miner.
PROOF OF USEFUL WORK — WHY PRL IS DIFFERENTBitcoin mining burns energy on SHA-256 preimage hunts — pure waste. Pearl's PearlHash runs verifiable matrix multiplication. The same computation that secures the network could power AI inference workloads.
- 1. COMPUTE — GPU runs the PearlHash kernel (real GEMM)
- 2. VERIFY — Block proof checked on-chain via zk-powered validation
- 3. EARN — Valid blocks earn 2,500 PRL, distributed via PPLNS
No trusted hardware. No hand-waving. The work is real or the block doesn't exist. Pearl is the only L1 where mining hashpower does real verifiable work.
NETWORK STATSAs of July 12, 2026- Network hashrate: ~35 EH/s
- Block reward: 2,500 PRL
- Block time: ~3 minutes
- PRL price: ~$0.40 (SafeTrade)
- Market cap: ~$98M
- Circulating supply: 244.5M PRL
- Difficulty: ~21.6M
LINKSROADMAP- [DONE] v1.0.0 — Linux + Windows static binary, sm_80/86/89/90/120
- [DONE] v1.0.0 — PPLNS pool with 3 regional endpoints (US/EU/IN)
- [DONE] v1.0.0 — TLS stratum, solo mode, telemetry API
- [WIP] v1.1.0 — B200 / sm_100 support
- [WIP] v1.1.0 — HiveOS flight sheet integration
- [PLAN] v1.2.0 — REST API for remote rig management
FAQQ: Which GPUs are supported?A: Any NVIDIA GPU with compute capability 8.0+ (RTX 30 series and newer). AMD support is not planned — PearlHash is CUDA-optimized GEMM.
Q: Do I need to register?A: No. Anonymous mining. Just pass your PRL wallet address.
Q: What are the fees?A: 3% dev fee (mined transparently, logged in telemetry) + 0.9% pool fee on PearlVolt pool. Solo mode: 3% dev fee, 0% pool fee. You can also mine to ANY other PRL pool — only the 3% dev fee applies.
Q: Can I solo mine?A: Yes. --pool stratum+tcp://pool.pearlvolt.run:4143 (0% pool fee)
Q: Is there a TLS option?A: Yes. --pool stratum+tls://pool.pearlvolt.run:4142
Q: How are payouts calculated?A: PPLNS (Pay Per Last N Shares). Your share of the work, transparently accounted. Min payout 0.10 PRL, paid every 2 hours.
Q: How is Pearl different from Bitcoin mining?A: Bitcoin wastes energy on SHA-256 preimage hunts. Pearl's PearlHash runs verifiable matrix multiplication — the same computation that powers AI workloads. Proof-of-Useful-Work, not proof-of-waste.
Q: Can I use PearlVolt with other PRL pools?A: Yes. PearlVolt speaks standard stratum. Change --pool to point at any PRL pool: pearlhash.xyz, kryptex.com, herominers.com, etc. The 3% dev fee applies regardless of which pool you use.
Q: How do I verify the dev fee is only 3%?A: Watch the telemetry stream at localhost:4141. Dev fee shares are logged with a separate worker name. Total dev fee time = 3% of runtime, transparently visible.
DISCLAIMERMining cryptocurrency involves risk. PRL price, network difficulty, and reward structures can change. Calculate your own profitability based on your electricity cost, GPU efficiency, and current PRL price. This is not financial advice. Mine responsibly.If you found this useful, consider mining to the PearlVolt pool. Every rig strengthens the network and earns PRL.
Happy mining.