Топ майнинг ПО

Рейтинг

v3.13.6.0

12 нояб. 2026 г.

Version 3.13.6.0

SHA256

Platform qli-Client
Linux x64 d154eeb82429dcae91b9b0c3734e5d08bbd809988848707b3fb24185204aa38d
Linux arm64 5f027b3056c0b0892231a37475b33fa1f448aba40847f49768a6f6c42cd63f90
Windows x64 5112df7e78edf2140e5b3d1abe26d731ae3ff159b7e7ee117ed5a93a59a4c977
Android (bionic/arm64) 03b1516b581caf00fd7757dc52f893a85632bf5b57ab2f818c2d921c71d18edc
macOS arm64 cc82450a6bcae4dd2ef2fa8cf2b20b1d175509715efd56e728235e95348c400e
macOS x64 21068c0462ce22c209ec887cd631567c3297f86815cb6d9bf9c7b3573931c693

RGminer v1.1.5

11 нояб. 2026 г.

Highlights / Основные изменения

English

Pearl / Pearlhash

  • Improved energy efficiency by 3–8% across RTX 20–50 series, depending on the architecture.
  • Increased hashrate by 2% on RTX 50 series and approximately 0.4% on other architectures.
  • Added support for Peryn.ai and MaxPool (maxpool.pro).
  • Fixed telemetry reporting for Pearl Fortune.

Nockchain / Nock-ZK

  • Added support for three more pools: Gigahash, NockMining and TheFactory.
  • Increased CMP 170HX hashrate by 14.5%.
  • CMP 170HX reference result: 34.01 MH/s at 150 W, with 1450 MHz core / 600 MHz memory.

Quantus

  • GTX 16xx / RTX 20 series: hashrate +1.7%.
  • RTX 30 series: hashrate +3.5%.
  • RTX 40 series: hashrate +3%.
  • RTX 50 series: hashrate +5%.

NOID

  • Added NOID mining support.
  • Supported Stratum pools: Suprnova, InnovLab and RPlant.
  • Developer fee: 2%.
  • Reference hashrates, power readings and clock settings are listed in the performance section below.

Pool and runtime improvements

  • Improved pool handling during frequent reconnects. This reduces the impact of unstable connections but cannot fix problems on the provider's side.
  • Reduced miner RAM usage and CPU load.
  • Removed Exfer support.

Windows

  • Fixed GPU unlocking on multi-GPU systems that previously encountered errors.
  • Corrected unlocking for CMP 40HX and CMP 70HX.
  • Added compatibility fixes aimed at older systems.

Web panel

https://rgminer.net/#/

  • Improved the ways rigs connect to the web panel.
  • Improved the panel's UI/UX.
  • Added email linking and two-factor authentication (2FA).

Built-in referral system

RGMiner is the first miner with a built-in referral system.

Invite your friends and earn more together :)

How it works:

  1. A user follows your referral link and registers in the web panel.
  2. All rigs linked to that user's account send every tenth developer-fee share (dev share) to the wallet you specify in your referral settings.
  3. Developer-fee mining uses the pool selected by the invited user, following the same connection approach as the regular developer fee. This helps avoid additional pool connectivity problems for the invited user.

To connect a rig, add the token from the website to the miner's launch parameters:

--token "YOUR_WEBSITE_TOKEN"

--mem-control

Linux only

Automatic memory control is now enabled by default: --mem-control on.

  • Extended support for 405 / 810 MHz memory modes to GPUs where these settings were previously unavailable, including Windows.
  • The miner automatically selects a memory profile for the algorithm and GPU, using GPU P-state configuration.
  • Settings changed by the miner are restored when it stops.
  • If another miner changed the GPU's P-state settings without restoring them, a rig reboot may be required when upgrading to restore a clean starting configuration.

Explicit miner OC parameters have the highest priority. --lock-mclock overrides the automatic memory frequency, and --mclock overrides the automatic memory offset.

With --mem-control on, external memory changes that conflict with the selected automatic profile are returned to the profile values. To use your own memory settings, set --mem-control off or specify them through the miner's OC parameters.

--smart-tune

Added smart undervolting, controlled independently of memory control:

  • --smart-tune off — disabled by default.
  • --smart-tune on — automatic per-GPU tuning.

The tuner adjusts private GPU controls beyond the usual public core offset and fixed core clock. It works with a fixed core clock or with a power limit alone. With --mem-control on, the miner also selects the algorithm's memory profile and the supported internal clock settings before tuning.

Tuning proceeds iteratively, using CUDA runtime errors and GPU recovery status reported by the driver. Settings are adjusted to one hardware step, equivalent to approximately 5–15 MHz of core offset, depending on the architecture. Unsupported private controls are skipped; where GPC control is unavailable, the supported core-offset tuning path is used.

A typical tuning session takes approximately 30–90 minutes, depending on the starting settings. After tuning finishes, GPU errors still trigger stepwise corrections toward more stable settings.

⚠️ Before enabling smart tuning

  • Avoid changing clocks during or after tuning, especially increasing them. The selected private settings belong to a specific operating point. The recommended procedure is to stop the miner, apply the new settings and start it again.
  • Changes to the configured fixed core clock, fixed memory clock or PL start a new tune for the affected operating point. Memory changes are evaluated after applying the --mem-control policy and explicit miner overrides.
  • On the first run, use a known-stable public core offset: it becomes the initial tuning value and is saved in the cache. When an applicable cache exists, changing only that public offset does not replace the saved starting value or trigger a retune.
  • During tuning, external public core-offset writes are blocked. Adding a public offset on top of the selected private settings can make the GPU unstable and may hang the rig. The original public offset is restored when the miner stops.
  • Tuning state and selected settings are saved automatically, allowing tuning to continue after a restart.

GPU recovery

Recovery follows the GPU's condition: the miner can restart an individual backend, perform a PCIe FLR reset, attempt bus recovery after a GPU falls off the bus, or reboot the host when required.

Some GPU failures can also block the NVIDIA driver while it waits for the device. In testing, a GPU reset alone did not always release these waits. --smart-tune on therefore selects the reboot watchdog policy, allowing a system reboot when recovery cannot complete. Reboots are forbidden when the effective watchdog policy is not reboot.

⚠️ Smart tuning may restart individual GPU backends or reboot the rig. Allow for this when choosing when to run it.

Smart tuning allows deeper undervolting, particularly on RTX 20, RTX 40 and RTX 50 series. RTX 30 series can also reach settings beyond those available through the public core offset alone. The feature will continue to evolve; feedback and test results are welcome. Further recovery improvements are planned to reduce the need for reboots.

Русский

Pearl / Pearlhash

  • Энергоэффективность увеличена на 3–8% на RTX 2000–5000 в зависимости от архитектуры.
  • Хешрейт увеличен на 2% на RTX 5000 и примерно на 0,4% на остальных архитектурах.
  • Добавлена поддержка Peryn.ai и MaxPool (maxpool.pro).
  • Исправлена передача телеметрии для Pearl Fortune.

Nockchain / Nock-ZK

  • Добавлена поддержка трёх пулов: Gigahash, NockMining и TheFactory.
  • Хешрейт CMP 170HX увеличен на 14,5%.
  • Результат CMP 170HX: 34,01 MH/s при 150 Вт, частоты 1450 МГц по ядру / 600 МГц по памяти.

Quantus

  • GTX 16xx / RTX 2000: хешрейт +1,7%.
  • RTX 3000: хешрейт +3,5%.
  • RTX 4000: хешрейт +3%.
  • RTX 5000: хешрейт +5%.

NOID

  • Добавлена поддержка майнинга NOID.
  • Поддерживаемые Stratum-пулы: Suprnova, InnovLab и RPlant.
  • Комиссия разработчика: 2%.
  • Хешрейт, потребление и частоты приведены в таблице производительности ниже.

Улучшения работы с пулами и оптимизация майнера

  • Оптимизирована работа при частых переподключениях к пулам. Это уменьшает влияние нестабильного соединения, но не устраняет проблемы на стороне провайдера.
  • Снижены потребление оперативной памяти и нагрузка на процессор.
  • Удалена поддержка Exfer.

Windows

  • Исправлен анлок на мульти-GPU системах, где ранее возникали ошибки.
  • Скорректирован анлок CMP 40HX и CMP 70HX.
  • Внесены исправления, направленные на поддержку старых систем.

Веб-панель

https://rgminer.net/#/

  • Улучшены способы подключения ферм к веб-панели.
  • Улучшен UI/UX интерфейса.
  • Добавлены привязка почты и двухфакторная аутентификация (2FA).

Встроенная реферальная система

RGMiner — первый майнер со встроенной реферальной системой.

Приглашайте друзей и зарабатывайте больше вместе :)

Принцип работы:

  1. Пользователь переходит по вашей реферальной ссылке и регистрируется в веб-панели.
  2. Все фермы, подключённые к его аккаунту, отправляют каждую десятую dev-шару на кошелёк, который вы указали в настройках реферальной системы.
  3. Для dev-майнинга используется пул, выбранный приглашённым пользователем, по аналогии с обычной dev-комиссией. Это помогает избежать дополнительных проблем с подключением к пулам у приглашённого пользователя.

Для подключения фермы добавьте токен с сайта в параметры запуска майнера:

--token "ВАШ_ТОКЕН_С_САЙТА"

--mem-control

Автоматическое управление памятью теперь включено по умолчанию: --mem-control on.

  • Расширена поддержка режимов памяти 405 / 810 МГц на GPU, где эти значения раньше были недоступны, в том числе в Windows.
  • Майнер автоматически выбирает профиль памяти под алгоритм и видеокарту, используя настройку P-state GPU.
  • При остановке майнер восстанавливает изменённые им настройки.
  • Если другой майнер изменил P-state и не вернул настройки, при обновлении может потребоваться перезагрузка фермы, чтобы восстановить исходное состояние GPU.

Явные параметры разгона майнера имеют наивысший приоритет. --lock-mclock переопределяет автоматическую частоту памяти, а --mclock — автоматический оффсет памяти.

При --mem-control on внешние изменения памяти, противоречащие выбранному автоматическому профилю, возвращаются к значениям профиля. Чтобы использовать свои настройки памяти, отключите функцию через --mem-control off или задайте значения OC-параметрами майнера.

--smart-tune

Поддержка добавлена для Linux

Добавлен режим умного даунвольта, работающий независимо от управления памятью:

  • --smart-tune off — отключён по умолчанию.
  • --smart-tune on — автоматический подбор настроек каждой GPU.

Тюнер настраивает приватные параметры GPU, выходящие за пределы обычного публичного оффсета и фиксированной частоты ядра. Подбор работает как с фиксом ядра, так и только с ограничением PL. При --mem-control on перед тюном также выбираются профиль памяти для алгоритма и поддерживаемые настройки внутренних частот чипа.

Подбор идёт итерациями с учётом ошибок CUDA runtime и статуса восстановления GPU, сообщаемого драйвером. Точность — один аппаратный шаг, соответствующий примерно 5–15 МГц оффсета ядра в зависимости от архитектуры. Неподдерживаемые приватные механизмы пропускаются; при недоступном GPC используется поддерживаемый путь подбора оффсета ядра.

Обычно тюн занимает от 30 до 90 минут в зависимости от стартовых настроек. После завершения подбора ошибки GPU также приводят к пошаговой корректировке настроек в сторону стабильности.

⚠️ Перед включением тюна

  • Не рекомендуется менять частоты во время или после подбора, особенно повышать их. Подобранные приватные настройки относятся к конкретному режиму GPU. Рекомендуемый порядок: остановить майнер, применить новые настройки и запустить его снова.
  • Изменение заданного фикса ядра, фикса памяти или PL запускает новый тюн для изменившегося режима. Изменения памяти учитываются после применения политики --mem-control и явных параметров майнера.
  • При первом запуске используйте заранее проверенный стабильный публичный оффсет ядра: он становится стартовым значением тюна и сохраняется в кеше. При наличии подходящего кеша изменение только этого оффсета не заменяет сохранённое стартовое значение и не запускает ретюн.
  • Во время тюна внешнее применение публичного оффсета ядра блокируется. Его сложение с подобранными приватными настройками может вызвать нестабильность GPU вплоть до зависания фермы. После остановки майнера исходный публичный оффсет восстанавливается.
  • Состояние тюна и подобранные настройки сохраняются автоматически, позволяя продолжить работу после перезапуска.

Восстановление видеокарт

Способ восстановления зависит от состояния GPU: майнер может перезапустить отдельный backend, выполнить PCIe FLR reset, попытаться восстановить карту после выпадения из шины или перезагрузить хост при необходимости.

При некоторых сбоях GPU также блокируется драйвер NVIDIA, ожидающий ответа от устройства. Во время тестов сброс GPU не всегда освобождал такие ожидания. Поэтому --smart-tune on выбирает политику вотчдога reboot, разрешающую перезагрузку системы, если восстановление не завершается. Если действующая политика вотчдога отличается от reboot, перезагрузка запрещена.

⚠️ Во время тюна возможны перезапуски отдельных GPU-backend и перезагрузка фермы. Учитывайте это при выборе времени запуска подбора.

Алгоритм тюна позволяет глубже даунвольтить видеокарты, особенно RTX 2000, RTX 4000 и RTX 5000. Для RTX 3000 также доступны настройки за пределами обычного публичного оффсета. Функционал будет дорабатываться — ждём отзывов и результатов тестирования. Дальнейшие улучшения восстановления направлены на сокращение необходимости перезагрузок.

HiveOS

Pearl
{
  "isFavorite": false,
  "items": [
    {
      "coin": "pearl",
      "pool_ssl": false,
      "dpool_ssl": false,
      "miner": "custom",
      "miner_alt": "rgminer-1.1.5",
      "miner_config": {
        "url": "YOUR_POOL",
        "algo": "pearlhash",
        "miner": "rgminer-1.1.5",
        "template": "%WAL%.%WORKER_NAME%",
        "install_url": "https://github.com/Printscan/rgminer/releases/download/v1.1.5/rgminer-1.1.5-hiveos.tar.gz"
      },
      "pool_geo": []
    }
  ]
}
Quantus
{
  "isFavorite": false,
  "items": [
    {
      "coin": "quantus",
      "pool_ssl": false,
      "dpool_ssl": false,
      "miner": "custom",
      "miner_alt": "rgminer-1.1.5",
      "miner_config": {
        "url": "YOUR_POOL",
        "algo": "quantus",
        "miner": "rgminer-1.1.5",
        "template": "%WAL%.%WORKER_NAME%",
        "install_url": "https://github.com/Printscan/rgminer/releases/download/v1.1.5/rgminer-1.1.5-hiveos.tar.gz"
      },
      "pool_geo": []
    }
  ]
}
Nock-ZK
{
  "isFavorite": false,
  "items": [
    {
      "coin": "pearl",
      "pool_ssl": false,
      "dpool_ssl": false,
      "miner": "custom",
      "miner_alt": "rgminer-1.1.5",
      "miner_config": {
        "url": "YOUR_POOL",
        "algo": "zkpow",
        "miner": "rgminer-1.1.5",
        "template": "%WAL%.%WORKER_NAME%",
        "install_url": "https://github.com/Printscan/rgminer/releases/download/v1.1.5/rgminer-1.1.5-hiveos.tar.gz"
      },
      "pool_geo": []
    }
  ]
}
NOID
{
  "isFavorite": false,
  "items": [
    {
      "coin": "noid",
      "pool_ssl": false,
      "dpool_ssl": false,
      "miner": "custom",
      "miner_alt": "rgminer-1.1.5",
      "miner_config": {
        "url": "YOUR_POOL",
        "algo": "noid",
        "miner": "rgminer-1.1.5",
        "template": "%WAL%.%WORKER_NAME%",
        "install_url": "https://github.com/Printscan/rgminer/releases/download/v1.1.5/rgminer-1.1.5-hiveos.tar.gz"
      },
      "pool_geo": []
    }
  ]
}

Performance / Производительность

Pearl
GPU Hashrate, TH/s Power, W Core Offset Core, MHz Memory, MHz Mem offset Efficiency, TH/s/W
Tesla V100-SXM2-16GB ≈37.85 215† — 1350 — — ≈0.176
CMP 40HX ≈52.91 ≈122.82–128.78 255 1650 5000 — ≈0.411–0.431
RTX 2080 ≈76.50 ≈156.36–163.95 135 1650 5000 — ≈0.467–0.489
CMP 50HX ≈91.26 ≈191.39–200.68 255 1650 5000 — ≈0.455–0.477
CMP 70HX ≈56.20 ≈110.73–111.76 255 1650 — -2000 ≈0.370–0.388
RTX 3060 Ti ≈63.55 ≈82.74–86.75 255 1650 405 — ≈0.733–0.768
RTX 3070m ≈66.99 ≈79.02–82.85 255 1650 405 — ≈0.808–0.848
RTX 3070 ≈77.01 ≈97.89–102.64 255 1650 405 — ≈0.750–0.787
CMP 90HX ≈83.29 ≈110.04–120.71 300 1650 405 — ≈0.503–0.527
RTX 3080 Ti ≈133.73 ≈174.12–182.57 255 1650 810 — ≈0.732–0.768
RTX 3090 ≈136.64 ≈185.93–194.95 255 1650 810 — ≈0.701–0.735
CMP 170HX ≈210.74 ≈224.83–242.29 300 1455 600 — ≈0.909–0.953
RTX 4070 Ti ≈149.60 ≈117.67–125.14 345 2445 405 — ≈1.269–1.330
RTX 4090 ≈317.7 ≈267.83–280.83 315 2445 405 — ≈1.137–1.192
RTX 5070 Ti ≈173.91 ≈141.67–148.54 499 2445 405 — ≈1.171–1.228
Quantus
GPU Hashrate, MH/s Power, W Core Offset Core, MHz Memory, MHz Mem offset Efficiency, MH/s/W
P104-100 49.06† 87.94† 150 1650 — -2000 0.558†
CMP 100-100 68.44† 189.7† — 1455 837 — 0.361†
GTX 1660 Super ≈158.25 55.36† 150 1650 810 — ≈2.858
GTX 1660 Ti ≈172.69 56.19† 150 1650 810 — ≈3.073
RTX 2060 ≈215.43 80.20† 210 1650 810 — ≈2.686
RTX 2070 ≈256.03 93.74† 200 1650 810 — ≈2.731
RTX 2080 ≈328.26 120.80† 115 1650 810 — ≈2.717
CMP 50HX ≈401.02 119.11† 250 1650 810 — ≈2.371
RTX 3060 Ti ≈279.35 85.00† 300 1650 405 — ≈3.286
RTX 3070m ≈294.15 67.00† 250 1650 405 — ≈4.390
RTX 3070 ≈337.94 90.26† 300 1650 405 — ≈3.744
RTX 3080 ≈496.31 144.24† 300 1650 405 — ≈3.441
RTX 3080 Ti ≈581.55 166.57† 300 1650 405 — ≈3.491
CMP 90HX ≈366.74 165.99† 300 1650 405 — ≈2.209
CMP 170HX ≈461.52 141.78† 210 1450 600 — ≈3.072
RTX 4070 Ti ≈652.92 138.03† 300 2460 405 — ≈4.730
RTX 4090 ≈1393.95 294.11† 315 2460 405 — ≈4.740
RTX 5070 Ti ≈753.90 154† 499 2460 810 — ≈4.895
Nock-ZK
GPU Hashrate, MH/s Power, W Core Offset Core, MHz Memory, MHz Mem offset Efficiency, MH/s/W
RTX 2080 24.65† 155.13† 115 1650 5000 0 0.159†
CMP 50HX 29.18† 198.63† 250 1650 5000 0 0.147†
RTX 3060 Ti 20.33† 124.64† 300 1650 810 0 0.163†
RTX 3070m 21.42† 76.09† 250 1650 5000 0 0.282†
RTX 3070 24.62† 129.90† 300 1650 5000 0 0.190†
RTX 3080 Ti 42.83† 236.61† 300 1650 5000 0 0.181†
CMP 90HX 26.62† 176.31† 300 1650 — -2000 0.151†
RTX 4070 Ti 47.33† 148.45† 300 2460 5000 0 0.319†
RTX 4090 100.07† 304.23† 315 2460 5000 0 0.329†
RTX 5070 Ti 61.86† 191.16† 499 2460 5000 0 0.324†
CMP 170HX 34.01 150 — 1450 800 — 0.227
NOID
GPU Hashrate, MH/s Power, W Core Offset Core, MHz Memory, MHz Mem offset Efficiency, MH/s/W
RTX 3070m 40.234 72.23 — 1650 405 — 0.557
RTX 4070 Ti 92.4 147 — 2460 405 — 0.629
RTX 5080 139.45 225.67 — 2647 405 — 0.618
RTX 2080 18.31 110 — 1650 810 — 0.166
CMP 170HX 97.5 140 — 1450 600 — 0.696

BzMiner v100.80

11 нояб. 2026 г.

Release v100.80 of bzminer

All feedback is welcome ;)

Join us on Discord: https://discord.gg/NRty3PCVdB

  • improved pearl hashrate/efficiency on nvidia
  • 5060ti can now do over 2th/w
  • fatal errors such as config.txt/command line parsing now show up immediately
  • new screen with 'e' key to view errors/warnings
  • fixed --oc-reset false
  • dual mining with profit switching (--mo-dual [2..6])
  • added --oc-power-tuner option (WARNING, can crash gpu drivers)
GPU Profile Core lock Core offset Memory lock Hashrate Power Efficiency
RTX 5090 Efficiency 1192 MHz +200 7001 MHz 188.6 TH/s 176.5 W 1.069 TH/W
RTX 5090 Balanced 2475 MHz +250 7001 MHz 401.3 TH/s 444.1 W 0.904 TH/W
RTX 5090 Hashrate 2595 MHz +150 7001 MHz 419.3 TH/s 485.2 W 0.864 TH/W
RTX 4090 Efficiency 2220 MHz +150 405 MHz 286.8 TH/s 230.7 W 1.243 TH/W
RTX 4090 Balanced 2430 MHz +340 810 MHz 314.4 TH/s 255.8 W 1.229 TH/W
RTX 4090 Hashrate 2520 MHz +100 405 MHz 325.3 TH/s 328.5 W 0.990 TH/W
RTX 4070 Ti SUPER Efficiency 2200 MHz +200 405 MHz 146.8 TH/s 118.1 W 1.243 TH/W
RTX 4070 Ti SUPER Balanced 2580 MHz +260 405 MHz 167.4 TH/s 136.2 W 1.229 TH/W
RTX 4070 Ti SUPER Hashrate 2700 MHz +200 405 MHz 175.8 TH/s 162.7 W 1.081 TH/W
RTX 3090 Efficiency 1500 MHz +300 405 MHz 124.1 TH/s 165.7 W 0.749 TH/W
RTX 3090 Balanced 1575 MHz +260 405 MHz 130.3 TH/s 178.1 W 0.732 TH/W
RTX 3090 Hashrate 1635 MHz +200 405 MHz 135.3 TH/s 215.6 W 0.627 TH/W
RTX 3080 Efficiency 1425 MHz +260 405 MHz 97.9 TH/s 130.0 W 0.753 TH/W
RTX 3080 Balanced 1725 MHz +265 405 MHz 118.2 TH/s 175.7 W 0.673 TH/W
RTX 3080 Hashrate 1785 MHz +150 405 MHz 122.3 TH/s 227.5 W 0.538 TH/W
RTX 3060 Ti Efficiency 1275 MHz +300 405 MHz 48.8 TH/s 61.4 W 0.794 TH/W
RTX 3060 Ti Balanced 1785 MHz +270 405 MHz 68.4 TH/s 114.6 W 0.597 TH/W
RTX 3060 Ti Hashrate 1860 MHz +150 405 MHz 71.2 TH/s 145.3 W 0.490 TH/W
RTX 5060 Ti Efficiency 1380 MHz +400 405 MHz 50.0 TH/s 24.3 W 2.052 TH/W
RTX 5060 Ti Balanced 2835 MHz +575 810 MHz 102.3 TH/s 101.9 W 1.003 TH/W
RTX 5060 Ti Hashrate 2850 MHz +500 810 MHz 102.7 TH/s 113.2 W 0.907 TH/W
CMP 170HX Efficiency 900 MHz +0 HBM 1215 MHz 122.5 TH/s 114.5 W 1.070 TH/W
CMP 170HX Balanced 1140 MHz +0 HBM 1215 MHz 152.8 TH/s 159.3 W 0.959 TH/W
CMP 170HX Hashrate 1410 MHz * +0 HBM 1215 MHz 180.1 TH/s 248.5 W 0.725 TH/W

Mines Tari, Conceal, Ryo, Ergo, Ethereum Classic, EthereumPoW, Ravencoin, Neurai, Neoxa, Meowcoin, Clore, nexa, Pearl, Quantus, Monero, Zephyr, Salvium, VERUS, Warthog and XELIS, and ships sha256d as the open-source SDK example.

Downloads

file size SHA-256
bzminer_v100.80_windows.zip 46.3 MB 325ab8f1363f1797651408e0889f60ed6ad433cbd577671e348061700360129d
bzminer_v100.80_linux.tar.gz 46.6 MB abaf5926ffb8a5be7f676c0825592e7b6c1441682f413c0718501050fab48370
bzminer_custom-v100.80.tar.gz 46.6 MB d5fab387273368f3cb2d6a3d0b90d2dad3d49d02a4695d9fe7888b4a213dbe5e
bzminer_mmpos-v100.80.tar.gz 46.6 MB 687bee24554563d2a0ec1b32cf1fb54da30eca8bee6906ca199b3c89067fcc2c
bzminer_v100.80_docker.tar.gz 13.4 KB d8612d481eb1bc545cca5f6cf9c79248a062615cd6b0cd55dca0f5c2f11eef01
bzminer_v100.80_macos.tar.gz 59.1 MB dbff8ade55ae24fec3384f8cacab0a3168dd74d81a2be4f5f3a539d92150bc7e

Verify before running:

# Linux / macOS
sha256sum -c <<< "<hash>  <file>"

# Windows (PowerShell)
Get-FileHash .\<file> -Algorithm SHA256

A checksum only proves the file is the one that was published — always download from the official release page.

Getting started

  1. Unpack the archive for your platform.
  2. Open a start_<algo> script and replace the placeholder wallet with your own.
  3. Run it, then open http://127.0.0.1:4014/ for the dashboard.

readme.txt inside each archive has the full quick start, the common options, and the hardware-safety settings. config.txt is a commented copy of every setting, generated from the shipped binary, so it always matches the build you have.

config.txt may be JSON or a command line. The shipped file is JSON — every setting inside one { } object. bzminer equally accepts the same file written as the options you would have typed, which is what a mining OS generates and the easier form to write by hand:

# one option per line, or a flag and its value on one line
-a pearl
-p stratum+tcp://pool:1234
-w YOUR_WALLET
--worker "My Rig"

Quote a value containing a space, and start a comment with #. An option you type on the command line beats the file's copy of it, so the file is a default you can override without editing it.

Docker

There is an image, and it carries the GPU userspace as well as the miner - no ROCm install, no driver packages, nothing on the host but a kernel driver:

docker pull bzminer/bzminer:v100.80
docker run --rm -it --gpus all bzminer/bzminer -a pearl -o stratum+tcp://pool:port -w YOUR_WALLET

The identical image is on GitHub as ghcr.io/bzminer/bzminer:v100.80 if you would rather not pull from Docker Hub.

Or keep your config, log and plugins on the host and run it as a service:

docker run -d --name bzminer --restart unless-stopped --gpus all \
    -v /opt/bzminer:/data -p 127.0.0.1:4014:4014 \
    bzminer/bzminer --set http_address=0.0.0.0

With no arguments it runs /data/config.txt; anything you pass goes straight to bzminer, so every command line on this page works unchanged. http_address has to be 0.0.0.0 for the dashboard, because the default 127.0.0.1 binds the container's loopback, which no published port can reach.

GPU flags differ per vendor. NVIDIA is --gpus all (with nvidia-container-toolkit on the host). AMD and Intel are --device /dev/kfd --device /dev/dri --group-add video --group-add render and --device /dev/dri. Overclocking, MSR tuning and huge pages additionally need --privileged; mining does not.

Update with a pull, not the miner's self-update - and the old tag stays published, so a rollback is just naming it:

docker pull bzminer/bzminer && docker restart bzminer

The binary inside the image is byte-identical to the one in bzminer_v100.80_linux.tar.gz above. Full documentation: docker run --rm bzminer/bzminer cat /opt/bzminer/docs/DOCKER.md.

Updating on a mining OS

Both fetch this exact version, so they can be pasted as-is - the version below is already the one on this page.

MMPOS

Put this in miner profile → advanced → "Initiate command prior to miner launch". It downloads once; on every later launch the if sees the archive already in /tmp and exits immediately, so it costs nothing per restart.

export version="v100.80"
if [ -f "/tmp/bzminer_${version}_linux.tar.gz" ]; then
exit 0
else
cd /tmp; wget https://github.com/bzminer/bzminer/releases/download/${version}/bzminer_${version}_linux.tar.gz; tar -xvf bzminer_${version}_linux.tar.gz; sudo cp -adpR bzminer_${version}_linux/bzminer /opt/mmp/miners/bzminer/
fi

Hive OS custom miner

The Linux release now includes bzminer_custom-v100.80.tar.gz, a complete Hive OS custom miner: the full bzminer binary, flight-sheet config adapter, log launcher, and native /hive_status reporting.

  1. Click Copy on the Flight Sheet JSON below.
  2. Open the Hive OS farm page that lists all workers.
  3. Open the Flight Sheets tab.
  4. Click Import from Clipboard.
  5. Create/save the flight sheet and add your PEARL wallet.
  6. Apply the flight sheet to the workers.

The sample mines Pearl on HeroMiners. Change url, pool_urls, algo, and coin for another pool or algorithm. The package maps those fields, %WAL%, the Hive worker name, and pass directly to bzminer. Put optional bzminer CLI arguments in user_config.

Flight Sheet JSON

{
  "name": "PEARL_BZMINER_CUSTOM",
  "isFavorite": false,
  "items": [
    {
      "coin": "PEARL",
      "pool_ssl": false,
      "pool_urls": [
        "us.pearl.herominers.com:1200"
      ],
      "dpool_ssl": false,
      "miner": "custom",
      "miner_alt": "bzminer_custom",
      "miner_config": {
        "url": "us.pearl.herominers.com:1200",
        "algo": "pearl",
        "pass": "x",
        "miner": "bzminer_custom",
        "template": "%WAL%",
        "install_url": "https://github.com/bzminer/bzminer/releases/download/v100.80/bzminer_custom-v100.80.tar.gz",
        "user_config": ""
      },
      "pool_geo": []
    }
  ]
}

Hive OS built-in miner (manual binary replacement)

Run it over SSH or in the Hive web shell.

You do not need to know which bzminer version is installed. Hive keeps one folder per version under /hive/miners/bzminer/, and the flight sheet decides which one runs - so this replaces the binary in every bzminer folder it finds and whichever one your flight sheet points at gets the new build. That is the whole reason it does not name a version directory.

version=v100.80
cd /tmp && wget -q https://github.com/bzminer/bzminer/releases/download/${version}/bzminer_${version}_linux.tar.gz && tar -xf bzminer_${version}_linux.tar.gz || { echo "download failed"; exit 1; }
miner stop
n=0; for d in /hive/miners/bzminer/*/; do [ -d "$d" ] && cp -f "bzminer_${version}_linux/bzminer" "$d" && n=$((n+1)); done
[ "$n" -gt 0 ] && echo "updated $n bzminer folder(s)" || echo "no /hive/miners/bzminer/<version>/ found - install bzminer from the flight sheet first"
miner start

It prints how many folders it updated. 0 means bzminer has never been installed from a flight sheet on that rig, so there is nothing to replace yet.

Reporting problems

Please include: your OS, GPU model(s) and driver version, the algorithm and pool, and the first ~40 lines of the log (it prints the build, the devices it found, and what each backend reported). --list-metrics output helps for sensor issues.

SRBMiner-MULTI v3.8.7

11 нояб. 2026 г.
  • Minor efficiency improvements on algorithm 'pearlhash' for NVIDIA 3000, 4000 and 5000 GPUs

  • Fixes to efficiency fine-tuner

  • Added parameter '--gpu-power-tuner-value'*

  • Default allowed hashrate loss for efficiency fine-tuner is now 3%

  • Bug fixes


Documentation for efficiency fine-tuner : https://github.com/doktor83/SRBMiner-Multi/blob/master/Parameters


After each completed tuning step, SRB reports average hashrate, power and efficiency versus your base OC. An improved step and the final improved winner include an opaque SRB value in square brackets. Copy that value without the brackets to use the measured combination directly:

--gpu-power-tuner-value VALUE --gpu-id 0,2 --gpu-power-tuner-value VALUE_FOR_GPU0,VALUE_FOR_GPU2

One value applies to all selected NVIDIA GPUs. A list follows --gpu-id order; without --gpu-id it follows selected NVIDIA device order. Manual values override --gpu-power-tuner and --gpu-power-tuner-loss. They do not start an automatic scan or load/overwrite its saved winners.

Manual values are independent of OC, algorithm, loss limit...; the target GPU/driver must support the requested controls. Manual values also work with OC configured outside the miner, for example through HiveOS. You do not need to repeat those OC settings in SRB's parameters.

Unlike automatic tuning, manual application does not require a locked core clock. If a GPU fault is attributed to an active manual value, SRB recovers the GPU and suspends that value for this recovery session rather than repeating it indefinitely.

Use manual values at your own risk! A value that was optimal for one OC can make the GPU unstable or crash with a different OC, even on the same GPU.


Custom miner for Hiveos

To use it you need to select “Custom miner” when you create Flight Sheet and paste this into 'Installation url' : https://github.com/doktor83/SRBMiner-Multi/releases/download/3.8.7/srbminer_custom-3.8.7.tar.gz

After that you must set everything manually (--algorithm, --pool, --wallet) through the 'Extra config arguments' field by using parameters. Parameters list here : ( https://github.com/doktor83/SRBMiner-Multi/blob/master/Parameters )


Archive MD5 hashes ff5f70ca61872ab350af3b49ceb028cb *SRBMiner-Multi-3-8-7-win64.zip 7471402c42ae0b347b1eebc731eef6ce *SRBMiner-Multi-3-8-7-Linux.tar.gz

v3.13.5.0

11 нояб. 2026 г.

Version 3.13.5.0

SHA256

Platform qli-Client
Linux x64 2a10d8b1b7f84840a3a70094044906461020632da0bab4e1db9f00029099a565
Linux arm64 efc97d1c15c8d90850a65f84df4a6c27dccc7ca8ea753a79538c87244d431947
Windows x64 0ab1362539fd78daf41446c6c65330b0efe0024a4667c31ca2b35683cbf17635
Android (bionic/arm64) 36207a77c793080b0fb68239ca38d88eef0364c56794ad16596cf90720a388f2
macOS arm64 60a0abc41f0e54517538bc4873c5bed668f8632b410c2ac732d9d53ef20d4528
macOS x64 73451d4639a54c71a1d76bc8e4772169d32ac1984b20bef626baf5ddfda387d8

v3.13.4.0

11 нояб. 2026 г.

Version 3.13.4.0

SHA256

Platform qli-Client
Linux x64 4ac4248a5137a0aede33086cade6bddd0b62f5b8738f2cf0cc4da8241c71550d
Linux arm64 d0a06de31b9478ef7b3bc3b0e13763f5009801085366eedbc433aab1ef6d9c69
Windows x64 1bd91db70b76e29de4d16476aa435ae51fcb8e424dff00a45be7be8019bf7ff1
Android (bionic/arm64) 54d22340439625127fd5b31fea7680ad6fa85f1edfca147f6d4a39a43e2b1d48
macOS arm64 4ad9a510e2d50b5ad1354952f7a7b109a349d4b95fe35e4949f8179d57ab4c03
macOS x64 b348f78d656cd8a885d1c40f82ebc05da539d9fba7e0fa7c1ed5a3fa4ca0cf21

Keryx-miner-v0.6.2-PoM

11 нояб. 2026 г.

Stratum / pools

  • TLS pool connections: use stratum+ssl://host:port (or stratum+tls://). The pool certificate is checked against the public CA roots, so pools with a Let's Encrypt or similar certificate work out of the box. --pool-ca-file adds a private CA for self-hosted pools.
  • stratum+tcp:// still works; the miner now warns at startup that the connection is not encrypted.

Multi-GPU rigs

  • A GPU that runs out of memory on its tier now falls back to the largest smaller tier already downloaded, --force-model included, and keeps mining it across era changes. The rig announces and serves exactly the models its GPUs mine.

Fixes

  • Rigs sharing a payout address now count prompt processing, not only generation, before a later rig takes over a request.

Credits

  • Unencrypted stratum connection spotted by @tonythewizard

WildRig Multi 0.52.1

9 нояб. 2026 г.

MANDATORY UPDATE in case you were using 0.52.0

  • improved pearlhash power efficiency up to 8% for NVIDIA gpu's
  • improved pearlhash for AMD gpu's
  • now memory can be locked to 405/810MHz on NVIDIA gpu's including CMP series

WildRig Multi 0.52.0

9 нояб. 2026 г.

UPDATE TO 0.52.1 WARNING: DO NOT USE THIS RELEASE, it can produce rejects for NVIDIA RTX 3000/4000 gpu's, probably same for 5000 series, but not yet confirmed

Keryx-miner-v0.6.1-PoM

8 нояб. 2026 г.

Performance

  • About +40% overall versus 0.6.0 on an RTX 3090.
  • Faster PoM v5 walk kernel, checked against the reference kernel at startup: 1.94 → 2.15 MH/s on an RTX 3090 (tier 3). Native SASS for sm_103.
  • New lock-step walk on top of it: several nonces share a warp and weights are read in contiguous blocks. The width is picked per card at startup, after the same check. 2.13 → 2.72 MH/s on the same card.
  • Model weights are loaded in a canonical layout, so the walk reads them in place with no extra copy.

Multi-GPU rigs

  • During an inference, only the GPU serving it pauses; the other GPUs keep mining.
  • A GPU that hits a CUDA fault is dropped on its own when another GPU still mines its model. The rest of the rig keeps going.

Stratum / pools

  • Traced inference and audit openings served through the pool, with the pool deciding when a request is traced.
  • PoM proofs sent in base64 when the pool supports it (hex otherwise).
  • Trace records uploaded to the pool when it asks for them.
  • The miner announces its models to the pool at connect and keeps it updated.
  • Sturdier pool connection: unknown messages and oversized lines are skipped instead of dropping the link, longer job timeout, lower latency on the socket, pool password option.
  • In pool mode, the chain height is read from the public API, so models of past eras are no longer downloaded.

Fixes

  • Some prompts that could not be served before are now answered normally.
  • Audit state is kept across restarts. Transient inference errors are retried, and broken UTF-8 in model output is handled.
  • Rigs sharing a payout address take over sooner when the rig ahead of them stays silent. The log line of a rig that was waiting its turn no longer blames it for the strike.

Credits

  • Lock-step walk and canonical weight layout ported from keryx-miner-supr by @ocminer.
  • Prompt serving issue spotted by @rogersmitha51.
  • Silent-rig takeover issue spotted by GiovannieR.