acominer 0.4.0: Fast Ethereum miner for AMD/Linux (Open Source)

[ishitatsuyuki]

acominer is an open source, fast and reliable Ethereum miner.

Requirements:

• OS: Linux
  A recent distro is recommended. For Ubuntu, use 20.04 or later.
• Graphics card: AMD card with 6GB or more memory (zombie mode is not supported)
  Developed on and optimized for RDNA architectures. Should run fine on GCN but performance might not be competitive.

Features:

• Zero miner fee
• Powered by Vulkan
• Small work dispatch size, gives lower latency
• Desktop-friendly, mining affects GUI performance minimally
  Mine and play games simultaneously without any tuning or lag

Performance: 49.5 MH/s @ 107W on a 5700 XT, stock clock and memory timings

Downloads
Download from GitHub, extract and execute ./run.sh [stratum address]. Only plain text stratum is supported.
Builds are fully automated to make audit easy.

v0.4.0
acominer v0.4.0 features LatencyFleX (TM)! Taking inspiration from packet congestion control, LatencyFleX reduces the latency when submitting shares, allowing a lower rate of stale shares. It dynamically adapts to system load, making room for responsiveness while keeping the latency low all the time.

It's advised that you apply https://github.com/ishitatsuyuki/patches/tree/master/drm-hrtimer to get the intended performance.

Changes:

• Implemented LatencyFleX (TM) (activate with --latencyflex)
• Properly flush caches before reading results
  This prevents potential lost shares. In practice this had a very low chance to happen, though.

v0.3.2
Changes:

• Fixed compatibility with GCN cards
• Continued performance tweaks
• Resolve additional pool compatibility issues

v0.3.1
This is a hotfix release addressing log output being hidden by default.

v0.3.0
v0.3.0 is a performance focused release. It features a rewritten mining kernel, providing up to 9% power reduction (118W -> 107W) on 5700 XT.

With this improvement, acominer now performs best-in-class, even among other proprietary miners.

Changes:

• Kernel rewrite and various performance tweaks
• Resolve some pool compatibility issues
• Rewrite logging to use a more detailed engine
  Tip: You can now set RUST_LOG=debug to see all received jobs.

v0.2.2
First release on the forum.

[1715197694]

1715197694

[1715197694]

1715197694

[1715197694]

1715197694

[1715197694]

1715197694

[ishitatsuyuki]

v0.3.0 is a performance focused release. It features a rewritten mining kernel, providing up to 9% reduction (118W -> 107W) on 5700 XT. With this improvement acominer now performs best-in-class, even among other proprietary miners.

Changes:

• Kernel rewrite and various performance tweaks
• Resolve some pool compatibility issues
• Rewrite logging to use a more detailed engine
  Tip: You can now set RUST_LOG=debug to see all received jobs.

I'm currently accepting feature requests. Fan control? Monitoring API? Even NVIDIA support... maybe?

Ask in this thread or on GitHub. I'll prioritize the more demanded ones.

[cafemax]

Good job, but I am windows user.

Will you release a windows ver. build? 

[ishitatsuyuki]

Will you release a windows ver. build? 

Sorry, Windows support is out of reach for now due to various driver quirks.

[ishitatsuyuki]

v0.3.2 has been released.

Changes:

• Fixed compatibility with GCN cards
• Continued performance tweaks
• Resolve additional pool compatibility issues

Download from GitHub.

[Shlimann]

Windows, please!

[ishitatsuyuki]

acominer v0.4.0 features LatencyFleX (TM)! Taking inspiration from packet congestion control, LatencyFleX reduces the latency when submitting shares, allowing a lower rate of stale shares. It dynamically adapts to system load, making room for responsiveness while keeping the latency low all the time.

It's advised that you apply https://github.com/ishitatsuyuki/patches/tree/master/drm-hrtimer to get the intended performance.

Changes:

• Implemented LatencyFleX (TM) (Activate with --latencyflex)
• Properly flush caches before reading results
  This prevents potential lost shares. In practice this had a very low chance to happen, though.

Download from GitHub.