[{"data":1,"prerenderedAt":245},["ShallowReactive",2],{"blog-2":3},{"id":4,"title":5,"author":6,"body":7,"createdAt":235,"description":236,"extension":237,"image":199,"meta":238,"navigation":239,"path":240,"seo":241,"stem":242,"tags":243,"__hash__":244},"blog\u002Fblog\u002F2-mxbm-an-open-source-beamhash-iii-gpu-miner.md","MXBM — an open-source BeamHash III GPU miner","Maxnflaxl",{"type":8,"value":9,"toc":226},"minimark",[10,14,17,25,30,33,36,92,95,99,112,116,119,127,131,172,176,183,187,190,200,203,206,209,216,223],[11,12,13],"p",{},"Beam has had GPU miners for years, but none of the fast ones are open source.",[11,15,16],{},"MXBM is an open one. A BeamHash III miner written from scratch against Beam's published spec, Apache-2.0, all source on GitHub.",[11,18,19],{},[20,21,22],"a",{"href":22,"rel":23},"https:\u002F\u002Fgithub.com\u002FMaxnflaxl\u002FMXBM",[24],"nofollow",[26,27,29],"h2",{"id":28},"speed","Speed",[11,31,32],{},"On a 4070 Ti SUPER at stock, CUDA does 74.4 sol\u002Fs and OpenCL does 63 sol\u002Fs. lolMiner does about 54 sol\u002Fs on the same card and won't go higher at any setting.",[11,34,35],{},"It depends on your power cap though:",[37,38,39,55],"table",{},[40,41,42],"thead",{},[43,44,45,49,52],"tr",{},[46,47,48],"th",{},"board cap",[46,50,51],{},"MXBM",[46,53,54],{},"lolMiner 1.98a",[56,57,58,70,81],"tbody",{},[43,59,60,64,67],{},[61,62,63],"td",{},"180 W",[61,65,66],{},"54.7 sol\u002Fs",[61,68,69],{},"53.3 sol\u002Fs",[43,71,72,75,78],{},[61,73,74],{},"220 W",[61,76,77],{},"69.9 sol\u002Fs",[61,79,80],{},"53.4 sol\u002Fs",[43,82,83,86,89],{},[61,84,85],{},"285 W",[61,87,88],{},"74.8 sol\u002Fs",[61,90,91],{},"53.8 sol\u002Fs",[11,93,94],{},"From about 120 W up MXBM is faster and more efficient. Below that lolMiner is still ahead by a few percent, so if you run your cards capped very low it may still be the better pick for now. That's what I'm working on next.",[26,96,98],{"id":97},"what-you-need","What you need",[100,101,102,106,109],"ul",{},[103,104,105],"li",{},"An NVIDIA card from the GTX 16 \u002F RTX 20 series or newer for the CUDA backend. Older NVIDIA and AMD cards run on OpenCL, which is slower.",[103,107,108],{},"3 GB of VRAM. With 8 GB or more you get the fastest memory layout.",[103,110,111],{},"Linux or Windows. Apple Silicon works too via Metal, see below.",[26,113,115],{"id":114},"dev-fee","Dev fee",[11,117,118],{},"1%. One 36-second round per hour. That's what the closed miners charge, so going open costs you nothing extra.",[11,120,121,122,126],{},"It shows up on its own row in the stats and the API. Fee rounds are tracked separately from your own shares. ",[123,124,125],"code",{},"--dev-fee"," raises it, nothing lowers it.",[26,128,130],{"id":129},"other-things-it-does","Other things it does",[100,132,133,136,150,156,166,169],{},[103,134,135],{},"Multi-GPU, including mixed rigs where some cards run CUDA and others OpenCL.",[103,137,138,139,142,143,142,146,149],{},"Overclocking from the command line: ",[123,140,141],{},"--pl",", ",[123,144,145],{},"--cclk",[123,147,148],{},"--mclk",", offsets, fan.",[103,151,152,155],{},[123,153,154],{},"--tune"," measures your card's power curve and tells you where to cap it.",[103,157,158,161,162,165],{},[123,159,160],{},"--apiport"," serves a small browser dashboard and a ",[123,163,164],{},"\u002Fsummary"," JSON endpoint.",[103,167,168],{},"A watchdog that exits with a restartable code if a card stalls.",[103,170,171],{},"lolMiner-style console and command line, so your scripts mostly just work.",[26,173,175],{"id":174},"for-fun","For fun",[11,177,178,179,182],{},"There's a native Metal backend for Apple Silicon. It does 18.8 sol\u002Fs on an M3 Max, which nobody should mine with, but it was interesting to write. There's also a CPU solver (",[123,180,181],{},"--solver ref",", needs the Beam core repo at build time) that's far too slow to mine with; it's there to check the GPU results.",[26,184,186],{"id":185},"what-would-help","What would help",[11,188,189],{},"Numbers from cards I don't own. Run",[191,192,197],"pre",{"className":193,"code":195,"language":196},[194],"language-text","mxbm --report\n","text",[123,198,195],{"__ignoreMap":199},"",[11,201,202],{},"and it benchmarks your card, measures its power curve and prints a markdown block you can paste into an issue. Nothing is uploaded. So far I have a 4070 Ti SUPER, 4070 SUPER, 3060 Ti, GTX 1660 Ti and an M3 Max. Turing looks slower than it should and I'm looking into it.",[11,204,205],{},"There's no AMD (HIP) backend yet, so AMD cards run on OpenCL. Contributions welcome.",[11,207,208],{},"Try it out and let me know what you think!",[11,210,211,212],{},"Discussion: ",[20,213,214],{"href":214,"rel":215},"https:\u002F\u002Fforum.0xmx.net\u002Fd\u002F204-mxbm-an-open-source-beamhash-iii-gpu-miner\u002F",[24],[11,217,218,219],{},"Telegram: ",[20,220,221],{"href":221,"rel":222},"https:\u002F\u002Ft.me\u002Fmxbm_miner",[24],[224,225],"mxbm-download",{},{"title":199,"searchDepth":227,"depth":227,"links":228},2,[229,230,231,232,233,234],{"id":28,"depth":227,"text":29},{"id":97,"depth":227,"text":98},{"id":114,"depth":227,"text":115},{"id":129,"depth":227,"text":130},{"id":174,"depth":227,"text":175},{"id":185,"depth":227,"text":186},"2026-08-28T19:45:23Z","A BeamHash III miner written from scratch, ahead of lolMiner.","md",{},true,"\u002Fblog\u002F2-mxbm-an-open-source-beamhash-iii-gpu-miner",{"title":5,"description":236},"blog\u002F2-mxbm-an-open-source-beamhash-iii-gpu-miner",[51],"-RYPIxGBxT6EYFyTDcOTyDh_c-5fAb-sNJC69ipXn2w",1791057490272]