CUDA miner

1121315171869

Comments

  • jakobitojakobito Member Posts: 9
    @cflush I followed the guide that @buechling shared some weeks ago and it worked perfectly.
    For 750Ti add -DCOMPUTE=50 in the cmake.
    Hope it helps.
    buechling said:

    I got it. Thank you for all the hints.. so for everybidy else: here is the ultimate guide for GTX 750 Ti on linux with 9,3 Mhash and 0% CPU:

    by the way: you only need a 16 GB Usb-Stick as hard drive.. and 4 GB RAM for the CPU ( but you really speed up compiling with 8 GB and can keep multiple DAG-Files with 32 GB Stick )

    Install Ubuntu Linux 14.03 x64

    sudo -s

    wget -O - http://llvm.org/apt/llvm-snapshot.gpg.key | sudo apt-key add -
    add-apt-repository "deb http://llvm.org/apt/trusty/ llvm-toolchain-trusty main"
    add-apt-repository -y ppa:ethereum/ethereum-qt
    add-apt-repository -y ppa:ethereum/ethereum
    add-apt-repository -y ppa:ethereum/ethereum-dev
    wget http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1404/x86_64/cuda-repo-ubuntu1404_7.0-28_amd64.deb
    dpkg -i cuda-repo-ubuntu1404_7.0-28_amd64.deb
    apt-get -y update
    apt-get -y upgrade

    apt-get -y install cpp-ethereum screen cuda build-essential git cmake libboost-all-dev libgmp-dev libleveldb-dev libminiupnpc-dev libreadline-dev libncurses5-dev libcurl4-openssl-dev libcryptopp-dev libjson-rpc-cpp-dev libmicrohttpd-dev libjsoncpp-dev libargtable2-dev llvm-3.8-dev libedit-dev mesa-common-dev ocl-icd-libopencl1 opencl-headers libgoogle-perftools-dev qtbase5-dev qt5-default qtdeclarative5-dev libqt5webkit5-dev libqt5webengine5-dev ocl-icd-dev libv8-dev

    git clone https://github.com/Genoil/cpp-ethereum
    cd cpp-ethereum
    nano cmake/EthCompilerSettings.cmake
    # remove "-Werror" around line 5
    nano libethash-cu/ethash_cu_miner.cpp
    # remove cudaDeviceSetSharedMemConfig(cudaSharedMemBankSizeEightByte); around line 140
    # then add cudaSetDeviceFlags(cudaDeviceScheduleBlockingSync);
    # then add cudaDeviceSetCacheConfig(cudaFuncCachePreferL1);
    cmake -DBUNDLE=miner -DETHASHCU=1
    make -j4
    reboot
    sudo -s
    cd cpp-ethereum/ethminer/
    screen
    ./ethminer -U -F http://(yourpool) --gpu-devices 0 1 2
    with 3 cards i now get 27,3 Mhash @ 270 Watt

  • PhantomPhantom Member Posts: 46
    @jakobito : Enable OC and fan controls in NVIDIA X Server Setting. Working with 750 ti, but not with GTX 960. With 960 rig, I go back my Window, but did not see much different between OC and Stock with Ethminer.

    sudo nvidia-xconfig -a --cool-bits=28 --allow-empty-initial-configuration (Source: https://bitcointalk.org/index.php?topic=826901.msg12279696#msg12279696)

    Note: My Gigabyte GA-N75TOC-2GI runs stable about 5 days with +135 GPU Clock and +650 Mem. Run 58Mh/s with about 475W . Don't limit power. Your cards will dead after 1-2 hours.

    @cfush: bueching wrote the setup guide on Ubuntu. You can go back to check it.
    https://forum.ethereum.org/discussion/2227/cuda-miner/p8

    And this is the link to build ccminer for quark, lyrav2... on Ubuntu
    https://gist.github.com/zcshiner/4b32980792d367222304

    I'm Windows guy, no experience with Linux. It took me many days to setup my rigs. However, need any help, just ask me.


  • PhantomPhantom Member Posts: 46
    @TheFiendishOne : with only one card, you must joint pool to earn Eth. When you joint pool, you do not need to run Ethereum client with Ethminer. Just run Ethereum client when you need to see your balance.

    Your hashrate should be higher than that. Take a look at http://cryptomining-blog.com/ to download the newest ethminer for Windows.
  • jakobitojakobito Member Posts: 9
    @Phantom Thank you very much!
  • TheFiendishOneTheFiendishOne Member Posts: 18
    edited September 2015
    @Phantom Thanks for the info.

    I originally started with ethminer-cuda-02092015-2 and ran with about 3 MH/s and the one you pointed me to tells me to set "ethminer.exe -F http://site.com/miner=address -U". With "-U" it auto-closes , without it I run at about 400KH/s and when I put -G infront of -F i'm back to my expected 3 (per the 02092015-2 release). MH/s so I'm curious what I have to put to get more than 3 because they've set the hashrate parm url to 5 and I'm not getting anywhere near 5.

    *EDIT*
    After realizing the update pattern per this thread, I've got the proper exe's in place and I appear to have a 2% increase from 3 MH/s but still not getting anything over 4

    Thanks.
  • PhantomPhantom Member Posts: 46
    @TheFiendishOne : Get Ethminer here http://cryptomining-blog.com/5612-updated-ethminer-0-9-41-opencl-and-cuda-for-windows/

    Make sure you run Ethminer in Ethminer-Cuda-0.9.41 folder and set ethminer.exe -F http://xxx -U
  • TheFiendishOneTheFiendishOne Member Posts: 18
    edited September 2015
    @Phantom Yeah I did that, the batch file will not launch if I have -U at the end. If I take -U out I get 400 khash, if i add -G -F I get 3 Mhash

    I have latest drivers
  • PhantomPhantom Member Posts: 46
    My rig with 2 x gtx 960 + 3 x gtx 970 medium OC runs stable about 67 Mh/s - 710Watt in Windows 7. I will test on Ubuntu tomorrow.

    @Genoil : this rig starts very high like 83Mh/s, then quickly drop to 68 or 67 Mh/s. Do you know a reason?
  • PhantomPhantom Member Posts: 46
    With that rig, I am doing about 71.5 Mh/s when I run Geth client. Nothing connects between them, because I solo mine with Geth client on another rig.
  • GenoilGenoil 0xeb9310b185455f863f526dab3d245809f6854b4dMember Posts: 769 ✭✭✭
    @Phantom perhaps one or more cards start throttling down? What does each individual card do? Is there a different pattern when using -S instead of -F?
  • TheFiendishOneTheFiendishOne Member Posts: 18
    Should I Install Ubuntu Linux 14.03 x64 to overcome this Windows issue? I've only used Fedora, once, so it'll be a new learning experience for me, but at the end of the day more hash = linux?
  • GenoilGenoil 0xeb9310b185455f863f526dab3d245809f6854b4dMember Posts: 769 ✭✭✭
    @TheFiendishOne for now that is the most effective solution. I'm looking into the issue as we speak. For some reason, on Win7, I get full 8.7MH on the benchmark, but only 50% in real mining mode. The main difference is that in benchmarking, you actually mine on a mock genesis block with the highest possible difficulty. I also get 8+MH in simulation mode (-S flag on the latest release), which mines on the same mock genesis block with varying difficulty. So you would think it's in the block headers. But if I hardcode them into the kernel as 0's, I get the same results..
  • TheFiendishOneTheFiendishOne Member Posts: 18
    edited September 2015
    does the difficulty in the simulation have anything to do with it? I am not sure what I'm looking at I was just observing that once it gets above 28 it stabilizes. Or is that the simulated block difficulty?

    Does the Ubuntu Linux I get have to be 14.03 or can I get the latest 14.04.3?

    **EDIT**
    if -U doesn't work for me(batch file crashes) does that mean there's a bug in the code? I know I don't want -C, and -G is opencl and it's what I'm getting 3MH/s from.
    Post edited by TheFiendishOne on
  • mockthereummockthereum Member Posts: 0
    Hi @Genoil,

    I'm trying to build the cuda-opencl-ptx branch on an Ubuntu 14 host with 3GB of RAM. However whenever I try it ends in my host swapping itself to death and needing a power cycle.

    The problem seems to be that the -DBUNDLE=cudaminer flag to cmake isn't having the desired effect. I do:

    cmake -DBUNDLE=cudaminer -DETHASHCU=1 -DCOMPUTE=50 -DMAXREGCOUNT=255

    And the output ends in:
    ------------------------------------------------------------- components
    -- MINER            Build miner                              ON
    -- ETHKEY           Build wallet tools                       ON
    -- TOOLS            Build basic tools                        ON
    -- SERPENT          Build Serpent language components        ON
    -- GUI              Build GUI components                     ON
    -- TESTS            Build tests                              ON
    -- ETHASHCL         Build OpenCL components                  OFF
    -- ETHASHCU         Build CUDA components                    ON
    -- JSCONSOLE        Build with javascript console            OFF
    -- EVMJIT           Build LLVM-based JIT EVM                 OFF
    ...so with GUI and tools still on.

    This results in the compiler using >2.5GB of RAM. Judging by the output it's getting stuck on something to do with wordlists.
  • mockthereummockthereum Member Posts: 0
    Hi @Genoil,

    I'm trying to build the cuda-opencl-ptx branch on an Ubuntu 14 host with 3GB of RAM. However whenever I try it ends in my host swapping itself to death and needing a power cycle.

    The problem seems to be that the -DBUNDLE=cudaminer flag to cmake isn't having the desired effect. I do:

    cmake -DBUNDLE=cudaminer -DETHASHCU=1 -DCOMPUTE=50 -DMAXREGCOUNT=255

    And the output ends in:
    ------------------------------------------------------------- components
    -- MINER            Build miner                              ON
    -- ETHKEY           Build wallet tools                       ON
    -- TOOLS            Build basic tools                        ON
    -- SERPENT          Build Serpent language components        ON
    -- GUI              Build GUI components                     ON
    -- TESTS            Build tests                              ON
    -- ETHASHCL         Build OpenCL components                  OFF
    -- ETHASHCU         Build CUDA components                    ON
    -- JSCONSOLE        Build with javascript console            OFF
    -- EVMJIT           Build LLVM-based JIT EVM                 OFF
    ...so with GUI and tools still on.

    This results in the compiler using >2.5GB of RAM. Judging by the output it's getting stuck on something to do with wordlists.
  • cflushcflush Member Posts: 24
    @Phantom
    @jakobito

    Thanks! I will have time this weekend to take a shot at ubuntu. If all goes well, I plan to use it on 2 rigs.
  • GenoilGenoil 0xeb9310b185455f863f526dab3d245809f6854b4dMember Posts: 769 ✭✭✭
    @mockthereum

    1. For best performance, use cudaminer-frontier branch. It's way ahead. opencl-ptx branch is just an experiment.

    2. I also found that i needed 8GB to build, but some posts back somebody with 4GB just waited it out and ultimately succeeded. you can also build cudaminer branch with less RAM, but I'm not planning to merge all maxwell kernel optimzations into it.
  • skunkskunk Member Posts: 13
    Genoil said:

    @mockthereum
    1. For best performance, use cudaminer-frontier branch. It's way ahead. opencl-ptx branch is just an experiment.

    any chance to make it cuda 6.5 compatible? i won't upgrade because cuda 7.5 makes ccminer slower...
  • GenoilGenoil 0xeb9310b185455f863f526dab3d245809f6854b4dMember Posts: 769 ✭✭✭
    @skunk yes i'll commit a patch tomorrow, but it won't be as fast. You could also download both and use the version you want. You would only need to trick cmake to pick the right CUDA for the right app.
  • cflushcflush Member Posts: 24
    edited September 2015
    @jakobito

    Can I use the gnome terminal to install / compile all that? Sorry - major linux newb here.

    I was thinking about installing Teamviewer on it to monitor it - does that sound feasible?

    Thx
    Cflush
  • antonio8antonio8 Member Posts: 44
    So my main rig just had a driver crash. Miner won't run now. Rebooted computer and it it still crashes.

    No idea what is going on. None of the versions run. Even if I try the cards one at a time.

    Other algos work fine.
  • PhantomPhantom Member Posts: 46
    My rigs are still running fine, but do not earn anything about 30 hours.
  • cflushcflush Member Posts: 24
    @Phantom
    Sorry to double post but I am working on this build tonight..
    Can I use the gnome terminal to compile with buechling's instructions? Or do I have to use xterm?

    Ya see, outta the gates linux confuses me. On windows you have cmd and for stuff like this u use ...cmd!! Why so many terminals / text editors ? arrghh!

    Thanks for your help on this guys.
  • PhantomPhantom Member Posts: 46
    @cflush: Ctrl + Alt + T to open terminal (like cmd on Windows)
  • jakobitojakobito Member Posts: 9
    @cflush I compiled it all in a terminal and its pretty easy (im linux newbie also).
    To control it remotely, I use ssh + screen.
    Install ssh server in unix using: apt-get install openssh-server
    Access to your unix machine from windows using ssh client like Putty (freeware)
    "screen" command allows you to close the ssh session and keep the miner running.

    But other options as Teamviewer, VNC, Xming, Exceed, etc. could work as well, if u want a remote desktop.
    cflush said:

    @jakobito

    Can I use the gnome terminal to install / compile all that? Sorry - major linux newb here.

    I was thinking about installing Teamviewer on it to monitor it - does that sound feasible?

    Thx
    Cflush

  • jakobitojakobito Member Posts: 9
    @cflush

    Unix commands (maybe a Linux expert will correct me :# ) in terminal (Xterm):

    To edit a file (without quotes): "sudo nano texfilename"
    In nano editor, save with ctrl+ X , etc

    If you are using the ssh server:
    To save a "screen" (without quotes): "sudo screen -S miner"
    To recover the "screen" when you log in again with a ssh client: "sudo screen -r miner"

  • cflushcflush Member Posts: 24
    edited September 2015
    @Phantom @jakobito When I get to nano libethash-cu/ethash_cu_miner.cpp, there is nothing there..just an empty file. Ugghh. Not sure what to do

    (EDIT): Ahhh, fook it. Past midnight and i am stuck. Linux never fails to give me a rash....gonna install a fresh copy of Win 7 Pro 64bit and see if i can give my 750 Tis some wings.

    again, thank you guys for your help. you are gentleman and scholars. :smiley:
    Post edited by cflush on
  • PhantomPhantom Member Posts: 46
    @cflush: Hey man, those steps don't need to do any more. No hope with Windows, don't waste your time. Gtx 750ti and Linux are a perfect couple :)

    Anyway, I re-write those steps now.

    git clone https://github.com/Genoil/cpp-ethereum

    cd cpp-ethereum

    cmake -DBUNDLE=miner -DETHASHCU=1

    make -j4

    Reboot

    Anytime you wanna mine:
    sudo -s

    cd cpp-ethereum/ethminer/

    ./ethminer -U -F http://(yourpool) -t n (n = how many gpu you have)
  • GenoilGenoil 0xeb9310b185455f863f526dab3d245809f6854b4dMember Posts: 769 ✭✭✭
    @Phantom i made the cudaminer-frontier branch default today, so your steps arent't correct any longer. It should now be this:

    git clone https://github.com/Genoil/cpp-ethereum

    cd cpp-ethereum

    cmake -DBUNDLE=cudaminer
    (optionally add -DCOMPUTE=50 for GTX750Ti-only rig)

    make -j4

    Reboot

    Anytime you wanna mine:
    sudo -s

    cd cpp-ethereum/ethminer/

    ./ethminer -U -F http://(yourpool) -cuda-devices 0 1 2 3 4 5
    (6 GPU's there)

  • SuchFakeAccountSuchFakeAccount Member Posts: 14
    @Genoli

    Building cudaminer-frontier now on linux fails and gives me bunch of errors.
Sign In or Register to comment.