You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This commit was created on GitHub.com and signed with GitHub’s verified signature.
Highlights / Основные изменения
English
Proxima (PROX)
Added support for all six Proxima pools available at release, including K1Pool, Innovlab, ProximaMine, AriaPool, ProximaPool, and Suprnova.
Pearl / Pearlhash
GPU series
Power consumption
Hashrate
RTX 50 series
−7%
+0.5%
RTX 40 series
−8%
+0.05%
RTX 30 series
−4%
+0.05%
RTX 20 series (default --kernel 2)
−3%
+1.5%
Startup and smart tuning
The miner starts faster.
Added and tested advanced --smart-tune on for GTX 1660 Super and CMP 170HX on Linux. Power consumption is down 4% and 2%, respectively, across supported coins with automatic tuning in place of manual core offsets.
Fixed an issue that could leave Turing GPUs tuning indefinitely and a bug that reset the tuning cache.
Moved the HiveOS tuning cache to /root/.cache/rgminer so subsequent updates can reuse saved GPU settings.
Sped up tuning on RTX 20, 40, and 50 series GPUs.
Telemetry and overclocking
Added GPU hotspot and memory temperature telemetry, along with core voltage at the active voltage/frequency (V/F) point where supported.
On Linux, use --vcore-target to set the core voltage target under load or --vcore-offset to apply a negative voltage offset. Both values are in mV.
Windows
Fixed Windows unlock bugs.
Русский
Proxima (PROX)
Добавлен майнинг Proxima на всех шести пулах, доступных на момент релиза, включая K1Pool, Innovlab, ProximaMine, AriaPool, ProximaPool и Suprnova.
Pearl / Pearlhash
Серия GPU
Потребление
Хешрейт
RTX 5000
−7%
+0,5%
RTX 4000
−8%
+0,05%
RTX 3000
−4%
+0,05%
RTX 2000 (ядро по умолчанию --kernel 2)
−3%
+1,5%
Запуск и автотюн
Ускорен запуск майнера.
Добавлен и протестирован расширенный режим --smart-tune on для GTX 1660 Super и CMP 170HX под Linux. При автоматической настройке вместо ручного подбора офсетов потребление снизилось на 4% и 2% соответственно на всех поддерживаемых монетах.
Исправлены бесконечный автотюн на Turing и ошибка сброса кеша настроек.
Кеш автотюна в HiveOS перенесён в /root/.cache/rgminer. Следующие обновления смогут использовать сохранённые настройки GPU.
Ускорен автотюн на видеокартах RTX 2000, 4000 и 5000.
Телеметрия и разгон
Добавлен вывод температуры горячей точки GPU и памяти, а также напряжения ядра в рабочей точке V/F на поддерживаемых GPU.
В Linux параметр --vcore-target задаёт целевое напряжение ядра под нагрузкой, а --vcore-offset задаёт отрицательное смещение напряжения. Оба параметра принимают значения в мВ.