Skip to content

Uminer GPU v0.1.24 (Testing Release)

Choose a tag to compare

View all tags
@github-actions github-actions released this 14 Jul 09:10
· 1 commit to main since this release
v0.1.24
fa1726d

Uminer GPU v0.1.24 Testing Release

English

Highlights

  • Hardens unattended mining against slow-drip and half-alive pool connections, incomplete
    handshakes, stuck GPU lifecycle operations and misleading stale performance gauges.
  • Fixes CSD extranonce allocation across clean jobs so pools with connection-level duplicate
    detection do not reject later valid shares as duplicates.
  • Improves the aligned two-minute console panel, bounded INFO event aggregation, local API and
    Prometheus data for multi-GPU operations.

Minimal Usage

# BTX
./uminer --coin btx -o stratum+tcp://global.btxpool.org:23333 -u <BTX_ADDRESS>.rig01 -p x

# CSD; tcp/tls is selected by the URL and is never forced globally.
./uminer --coin csd -o stratum+tls://csd.luckypool.io:443 -u <CSD_ADDRESS>.rig01 -p x

Changes

  • Pool reads now have a wall-clock deadline in addition to socket timeouts, preventing a peer
    from holding a session indefinitely by dripping bytes.
  • Handshake, job freshness, submit accounting and reconnect paths use bounded health rules.
    A backup pool is used only when the user configured multiple pools; Uminer does not invent or
    replace the user's pool choice.
  • GPU setup, rebuild, work and Fee operations have explicit liveness boundaries. A failed GPU is
    isolated while healthy GPUs continue; only loss of the last usable GPU stops mining.
  • BTX low-difficulty work is bounded before a batch can overflow the result queue. CSD fractional
    vardiff and per-session extranonce allocation remain exact.
  • The status panel reports current-round and process-total A/S/R, real per-device work rates,
    power, clocks, VRAM, temperature, limit state, host usage and Fee totals with fixed-width
    alignment. Semantic terminal colors never use bold and are omitted from non-TTY/file output.
  • New-job and share-result events remain INFO-visible; after ten detailed events in one minute,
    further events are emitted as a classified aggregate and detailed output resumes in the next
    minute.
  • CSD pool-reported hashrate negotiation remains optional and endpoint-scoped. Unsupported
    extensions degrade to share-only reporting without stopping GPU work.

Compatibility

  • Linux x86_64 only; Windows is not included.
  • glibc 2.30 or newer.
  • Embedded native SASS: sm_75, sm_80, sm_86, sm_89, sm_120.
  • NVIDIA driver 535 or newer for Turing/CMP, Ampere and Ada.
  • NVIDIA driver 580.95 or newer for Blackwell.
  • No CUDA Toolkit installation is required at runtime.

Developer Fee

  • BTX: 2%.
  • CSD: 3%.
  • The per-coin policy is compiled into this release and cannot be reduced through the CLI.
  • A GPU changes to developer work only after a trusted Fee route has connected, authorized and
    supplied a valid job. If no route reaches READY, user mining continues and the missed window is
    not recovered. The user-pool connection remains online during a Fee window.

Verification and Scope

The workspace tests and strict lint checks passed, two clean CUDA release builds were
byte-identical, the package manifest and active Fee policy were verified, and the archive and
provenance carry detached signatures. This is a Linux testing-stage release under an explicit
fast-release exception; ongoing long-duration production qualification is not claimed here.

繁體中文

重點

  • 強化無人值守挖礦面對慢速滴流、半存活礦池連線、未完成握手、GPU 生命週期卡住及過期算力
    顯示時的自癒與可觀測性。
  • 修正 CSD 在 clean job 切換後重用 extranonce 的問題，避免採用連線級重複檢測的礦池把後續
    有效 share 判成 duplicate。
  • 改善固定對齊的兩分鐘控制台面板、有限 INFO 事件聚合、本機 API 與 Prometheus 多卡資料。

最簡使用方式

# BTX
./uminer --coin btx -o stratum+tcp://global.btxpool.org:23333 -u <BTX_ADDRESS>.rig01 -p x

# CSD；tcp/tls 完全由 URL 選擇，不做全域強制。
./uminer --coin csd -o stratum+tls://csd.luckypool.io:443 -u <CSD_ADDRESS>.rig01 -p x

變更

  • 礦池讀取除 socket timeout 外新增牆鐘截止時間，避免對端持續滴入少量位元組而永久佔住會話。
  • 握手、任務新鮮度、submit 記帳與重連均採有界健康規則。只有使用者明確設定多個礦池時才會
    切換備援；Uminer 不會自行新增或替換使用者選擇的礦池。
  • GPU 開啟、重建、工作及 Fee 操作都有明確存活邊界。單卡失敗只隔離該卡，其餘健康 GPU
    繼續；最後一張可用 GPU 失效才停止挖礦。
  • BTX 低難度批次在結果佇列溢出前主動限制工作量；CSD 小數 vardiff 與每會話 extranonce
    分配保持精確。
  • 狀態面板以固定欄寬顯示本輪及累計 A/S/R、逐卡真實工作速率、功率、頻率、顯存、溫度、
    限制狀態、主機資源與 Fee 總計。互動終端語義顏色不使用粗體，非 TTY 與檔案輸出不含 ANSI。
  • 新任務與 share 結果保持 INFO 可見；一分鐘內超過十條明細後改為分類聚合，下一分鐘自動
    恢復逐條輸出。
  • CSD 算力自報協商維持可選且按端點隔離；礦池不支援擴充時只停用自報，不會停止 GPU 工作。

相容性

  • 僅提供 Linux x86_64，不含 Windows。
  • glibc 2.30 或更新版本。
  • 內嵌 native SASS：sm_75、sm_80、sm_86、sm_89、sm_120。
  • Turing/CMP、Ampere、Ada 需 NVIDIA 535 或更新驅動。
  • Blackwell 需 NVIDIA 580.95 或更新驅動。
  • 執行時不需安裝 CUDA Toolkit。

開發者費用

  • BTX：2%。
  • CSD：3%。
  • 各幣種費率已固化於本版本，CLI 無法下調。
  • 只有受信任 Fee 路由完成連線、授權並取得有效任務後，GPU 才會切換。若沒有路由進入
    READY，使用者挖礦持續且該時間窗不追補；Fee 時間窗內使用者礦池連線保持在線。

驗證與發佈邊界

workspace 測試與嚴格 lint 已通過，兩次 clean CUDA release build 位元組完全一致，套件
manifest 與 active Fee 策略已驗證，歸檔及 provenance 均附獨立簽名。本版依明確快速發佈特例
提供 Linux 測試版；仍在進行的長時間生產資格驗收不在本版宣稱範圍內。

Assets 10
Loading