Github repo, issue tracker, and wiki Newest binary, always updated to match newest git revision Works on AMD Radeon 4xxx and up (SDK 2.1 or 2.4/2.5 recommended, avoid 2.2 and 2.3), and Nvidia 8xxx and up (make sure you have their newest driver). Run ./DiabloMiner-YourOS.sh -u youruser -p yourpass matching the user and pass set in your bitcoin.conf. On Windows, run DiabloMiner-Windows.exe instead. I use what used to be ArtForz's excellent kernel, although it seems to have mutated since then, with phatk bits thrown in. This code may not be entirely endian safe. DiabloMiner now implements never fail async networking, and is the only miner guaranteed Eligius proof. DiabloMiner also supports connecting to multiple pools concurrently, just use commas. Radeon users: Catalyst 10.9 through 10.11 + SDK 2.1 seems to be the golden combination for 4xxx and 5xxx miners. Catalyst 11.x + SDK 2.5 seems to be golden combination for 6xxx. 2.4/2.5 does not suffer from the CPU use bug 2.2/2.3 have. Do not use 2.2/2.3. OSX users: You need at least OSX 10.6 (Snow Leopard) to run OpenCL apps of any kind. Building from source: Run mvn package to compile, you need Maven installed. Performance tweaking This is only a series of recommendations. This is not a replacement for actual testing. Each computer is different. Workgroup size: By default, DiabloMiner uses whatever the hardware maximum is. This is usually 256. See optimum settings below. Vectors: By default, DiabloMiner does not use vectors. -v 1: The same as off, it uses a single group of single wide variables (uint) -v 2: The same as poclbm and phoenix, it uses a single group of double wide variables (uint2) -v 3: Two groups of variables with instruction interleaving, one uint2 one uint. -v 4: Two sets of uint2 -v 5: Two sets of uint2 and a uint -v 6: Three sets of uint2 -v 18: Two sets of uint -v 19: Three sets of uint -v 20: Four sets of uint -v 21: Five sets of uint -v 36: uint4 -v 40: Two uint4 Optimum settings Radeon 4xxx on SDK 2.1 and 2.4: -w 64/128 Radeon 5xxx on SDK 2.1: -v 2 -w 128/192/256 Radeon 5xxx/68xx/69xx on SDK 2.4: -v 2 -w 128/192/256 Radeons with underclocked memory or 5970s or 6990s: Some have to use -w 192 or -w 256 instead of -w 128 to maintain performance Nvidia: -w 64, and you may require -f 240 or higher to maintain desktop interactivity if desired. Radeon Stream SDK: Stream SDK 2.1 is the fastest SDK for mining, but not all devices work with it. 2.4 on a 2.1-capable device loses about 10% of your mining speed. Not supported for OpenCL: Any R600 or earlier chip: 2xxx, 3xxx, 42xxM, IGP 3xxx, IGP 42xxM, FirePro 2250, FireStream 9170, any FireGL, any FireMV SDK 2.1 supports: Any R700 desktop: 43xx and up Any R700 laptop: 43xxM and up, 5xxVM, 51xxM Any R700 workstation: FirePro 2450, RG220, V3750, V5750, V7750, V87xx, M5725, M7740, Firestream 92xx Any 5xxx-like VLIW5 desktop (aka Cedar, Redwood, Junpier, and Cypress): 5xxx, 6750, 6770 Any 5xxx-like VLIW5 laptop: 54xxM through 58xxM, 63xxM, 65xxM, 68xxM Any 5xxx-like VLIW5 workstation: FirePro 2270, 2460, V3800, V4800, V5800, V7800, V8800, V8900, M5800, M7820, FireStream 93xx SDK 2.4 supports everything 2.1 does, plus: Any 6xxx-like VLIW5 (aka Barts): 6790, 68xx, 69xxM Any 6xxx-like VLIW4 (aka Caicos, Turks, Cayman): 64xx, 65xx, 66xx, 69xx, 64xxM, 66xxM, 67xxM Any Fusion APU VLIW5 (aka Wrestler, WinterPark): 62xx, 63xx, 64xx |