Recovers the v1.7.8 hashrate regression and adds real gains across the board. Measured on the LAN fleet (70% power limit, same-rig vs v1.7.8):
RTX 5090: +30% (~355 TH/s)
RTX 4090: +5–15% (~210–271 TH/s)
RTX 5080: +14% (~204 TH/s)
RTX 4060 Ti: +14% (~72 TH/s) · RTX 4080 Super: +5%
RTX 4080 / 5070 Ti / 5060 Ti: flat (within noise)
Fleet aggregate +11% vs v1.7.8. No card regressed.
Kernel/throughput optimizations for NVIDIA Blackwell and Hopper. No backend changes — a pure performance release.
Blackwell (RTX 50-series): ~8% faster vs v1.7.7
Hopper (H100): ~680–700 TH/s expected
Dev fee removed — alpha-miner is now FEE FREE at 0%. Every TH/s you produce goes to your wallet.
RTX 50-series regression bugs fixed — addresses performance and stability issues on 5090 / 5080 / 5070 Ti / 5060 Ti observed in v1.7.6. RTX 50 owners: this is the release for you.
Continued performance refinements across the supported backend matrix (volta/ampere/ada/hopper/blackwell/blackwell-native/b200/b300).
B200 + B300 support carried forward from v1.7.6.
The v1.7 family delivered substantial measured hashrate gains over v1.6 on our internal LAN fleet
Pearl: Hashrate improvements on NVIDIA Ampere / Ada Lovelace / Blackwell:
RTX 3090: 100-110 TH/s
RTX 4090: 250-260 TH/s
RTX 5090: 350-360 TH/s