CUDA miner

1454648505169

Comments

  • GenoilGenoil 0xeb9310b185455f863f526dab3d245809f6854b4dMember Posts: 769 ✭✭✭
    @luthor that's getwork (-F) right? -S (stratum) does not actively report rate.
  • luthorluthor Member Posts: 34
    Genoil said:

    @luthor that's getwork (-F) right? -S (stratum) does not actively report rate.

    Ok, yes the one in blank its in stratum mode, this its a good test because its the the same card on different rigs, 2 on proxy and this one on stratum, lets see the next 24 hours
  • GenoilGenoil 0xeb9310b185455f863f526dab3d245809f6854b4dMember Posts: 769 ✭✭✭
    @luthor thanks for helping out!
  • Bensam123Bensam123 Member Posts: 62
    What level of interest do you have in doing a dual miner Genoil? Right now Claymore is like gold... Surprised it's not being mentioned here.
  • GenoilGenoil 0xeb9310b185455f863f526dab3d245809f6854b4dMember Posts: 769 ✭✭✭
    Bensam123 said:

    What level of interest do you have in doing a dual miner Genoil? Right now Claymore is like gold... Surprised it's not being mentioned here.

    The Claymore miner has its own thread here, no worries. Of course it would be great doing something like that, but I'm not sure my fork is the ideal basis for such an adventure. Right now there are more important things to fix, such as -as you suggested more than once- stability. Which leads me to a question: what exactly happens when a card in a rig crashes? What happens with ethminer, does it still give output? Or does it exit? I'm trying to simulate such crashing (i have exactly 1 headless 7950) to fix it. But if I just disable the device in Device Manager, the amd opencl dll just crashes. I'm not sure if that's what I need to workaround :)
  • dlehenkydlehenky Member Posts: 2,249 ✭✭✭✭
    @Genoil On linux, it appears the failed GPU locks up the driver (I think) and pegs the processor at 100%. When you try to kill the ethminer process, you can't, and it turns into a zombie process. That basically locks up the whole GPU complex, so a new ethminer process gets nowhere - I think it starts mining it's own chain. A reboot, and sometimes a full power cycle, is needed to clear the GPU fault. On Windows, I thought there was a watchdog that was supposed to break away from the hung GPU, but I have never mined with Windows, so I have no direct experience with it.
  • retherrether Member Posts: 258 ✭✭
    Bensam123 said:

    What level of interest do you have in doing a dual miner Genoil? Right now Claymore is like gold... Surprised it's not being mentioned here.

    Is the dual miner really worth it? My experience is that decred generates a lot more heat from increased power draw (see: http://cryptomining-blog.com/7501-nvidia-geforce-gtx-970-thermal-images-for-ethererum-and-decred-mining/ ). Not to mention, decred is pretty low market cap.

    Increased power, (possible) lost ETH hashing, paying the dev fee...doesn't seem worth it. It's a really cool idea but I don't think it's that amazing in practice. I'm open to data that says otherwise.

  • retherrether Member Posts: 258 ✭✭
    Does anyone have a copy of the latest compiled with CUDA 6.5?
  • Bensam123Bensam123 Member Posts: 62
    Genoil said:

    Bensam123 said:

    What level of interest do you have in doing a dual miner Genoil? Right now Claymore is like gold... Surprised it's not being mentioned here.

    The Claymore miner has its own thread here, no worries. Of course it would be great doing something like that, but I'm not sure my fork is the ideal basis for such an adventure. Right now there are more important things to fix, such as -as you suggested more than once- stability. Which leads me to a question: what exactly happens when a card in a rig crashes? What happens with ethminer, does it still give output? Or does it exit? I'm trying to simulate such crashing (i have exactly 1 headless 7950) to fix it. But if I just disable the device in Device Manager, the amd opencl dll just crashes. I'm not sure if that's what I need to workaround :)
    Usually it just locks up. It doesn't output any errors. It just sits there and doesn't do anything. The GPUs are still operating at max frequency and drawing the same amount of watts. The easiest way to tell which GPU exploded is by looking at a GPU monitoring program and see which one reset its clocks.

    Going back to talking about a dual miner, you technically wouldn't need to make a dedicated miner for Dcr for instance. You could talk to SP about integrating his and splitting the mining percentages between the two of you.

    Joint venture wouldn't be a bad idea.
  • dominusdominus Member Posts: 36
    edited April 2016
    Bensam123 said:


    Usually it just locks up. It doesn't output any errors. It just sits there and doesn't do anything. The GPUs are still operating at max frequency and drawing the same amount of watts. The easiest way to tell which GPU exploded is by looking at a GPU monitoring program and see which one reset its clocks.

    I can confirm that 100%.
  • Franco_ozFranco_oz Member Posts: 21
    As @dominus and @Bensam123 I can confirm 100% too.
  • cpixelscpixels Member Posts: 3
    edited April 2016
    Slightly modified version of @dr_pra instuctions for Ubuntu 15.04

    Ubuntu 15.04. OpenCL + CUDA

    wget http://developer.download.nvidia.com/compute/cuda/7.5/Prod/local_installers/cuda-repo-ubuntu1504-7-5-local_7.5-18_amd64.deb
    sudo dpkg -i cuda-repo-ubuntu1504-7-5-local_7.5-18_amd64.deb
    sudo apt-get -y install software-properties-common
    sudo add-apt-repository -y ppa:ethereum/ethereum
    sudo apt-get update
    sudo apt-get install git cmake libcryptopp-dev libleveldb-dev libjsoncpp-dev libjson-rpc-cpp-dev libboost-all-dev libgmp-dev libreadline-dev libcurl4-gnutls-dev ocl-icd-libopencl1 opencl-headers mesa-common-dev libmicrohttpd-dev build-essential cuda -y
    git clone https://github.com/Genoil/cpp-ethereum/
    cd cpp-ethereum/
    mkdir build
    cd build
    cmake -DBUNDLE=cudaminer ..
    make -j8
  • dominusdominus Member Posts: 36
    bitcanuck said:

    Franco_oz said:

    As @dominus and @Bensam123 I can confirm 100% too.

    Is this Linux or Windows? I've been using cpp-ethereum 1.2.2 ethminer for about a month now on both Linux (Ubuntu 14.0.4) and Windows 7e. A couple days ago I switched to genoil-1.0.6 on Windows, and so far no problems (and better hashrates seen at mph compared to regular ethminer + eth-proxy). The Win7e box is running a single MSI R9 380 OC to 1G/1.5G with 15.7.1 catalyst drivers.

    Windows 7x64, nvidia.
  • Mo35Mo35 Member Posts: 43
    rether said:

    Does anyone have a copy of the latest compiled with CUDA 6.5?

    ill try to find time this weekend and compile 1.0.7 , if im successful i`ll post it
  • ethfanethfan Member Posts: 458 ✭✭✭
    Since my request for Linux binaries was... not seen? / ignored? ... I was "forced" to do it personally. Compiled 107 for 14.04 and now testing on one of my 6 cards rig and on my test rig (with just a single GPU) at ethermine. Looking good. Hash-rate marginally lower. BUT, more importantly, the invalid share rate has dropped considerably. Where it was averaging 5% (with occasional spikes to >10% !) with qtminer, the 6 carder now shows a steady 2%. Even more remarkable the test rig shows ZERO invalids!
  • BeneKikiBeneKiki Member Posts: 17
    ethfan said:

    Since my request for Linux binaries was... not seen? / ignored? ... I was "forced" to do it personally. Compiled 107 for 14.04 and now testing on one of my 6 cards rig and on my test rig (with just a single GPU) at ethermine. Looking good. Hash-rate marginally lower. BUT, more importantly, the invalid share rate has dropped considerably. Where it was averaging 5% (with occasional spikes to >10% !) with qtminer, the 6 carder now shows a steady 2%. Even more remarkable the test rig shows ZERO invalids!

    Were you able to compile a version which supports stratum mode (-S)? I tried to compile several time, got a working executable, but the -S flag never worked.
  • ethfanethfan Member Posts: 458 ✭✭✭
    BeneKiki said:

    Were you able to compile a version which supports stratum mode (-S)? I tried to compile several time, got a working executable, but the -S flag never worked.

    You need to compile the 107 branch. After git clone do a git checkout 107. Then do the compile.
  • GenoilGenoil 0xeb9310b185455f863f526dab3d245809f6854b4dMember Posts: 769 ✭✭✭
    @ethfan sorry I don't compile Linux binaries. But just got an offer from somebody who's going to build a PPA.
  • dlehenkydlehenky Member Posts: 2,249 ✭✭✭✭
    @BeneKiki I just got it built today with stratum. Didn't you find the build instruction in the ethpool.org Support section? Just run "cmake -DBUNDLE=miner .." from the build directory *you* create under the source code directory. I can tell you how to "install" the ethminer command and libraries under your home directory, if you need help, otherwise you'll have to run ethminer from the build directory, which is a bit cumbersome, at least to me.
  • ethfanethfan Member Posts: 458 ✭✭✭
    Genoil said:

    @ethfan sorry I don't compile Linux binaries. But just got an offer from somebody who's going to build a PPA.

    No need to say sorry. :) You are not obliged to do everything anyone asks!
    But great news about the PPA. Please inform when ready...
  • BeneKikiBeneKiki Member Posts: 17
    dlehenky said:

    @BeneKiki I just got it built today with stratum. Didn't you find the build instruction in the ethpool.org Support section? Just run "cmake -DBUNDLE=miner .." from the build directory *you* create under the source code directory. I can tell you how to "install" the ethminer command and libraries under your home directory, if you need help, otherwise you'll have to run ethminer from the build directory, which is a bit cumbersome, at least to me.

    I followed the guide and builtit with "cmake -DETHSTRATUM=ON -DBUNDLE=miner .." command and ran it from buid directory, but no luck... I missed something!
  • GenoilGenoil 0xeb9310b185455f863f526dab3d245809f6854b4dMember Posts: 769 ✭✭✭
    @BeneKiki if you specify -DBUNDLE=miner, ETHSTRATUM is set to ON by CMake, so it shouldn't even be required.
  • BeneKikiBeneKiki Member Posts: 17
    edited April 2016
    I succeeded! The build files were always good, but I failed to run etherminer from the right directory, becase I missed the "./" before it :-) So the linux build working fine!
  • dlehenkydlehenky Member Posts: 2,249 ✭✭✭✭
    @BeneKiki Great news! Happy mining! :)
  • dlehenkydlehenky Member Posts: 2,249 ✭✭✭✭
    @Genoil I'm going to have a go at eliminating one of the causes of the GPU errors (verifying results for last block when a new block just arrived). If I have any success, I'll let you know. I *think* I have a fairly straightforward solution.
  • retherrether Member Posts: 258 ✭✭
    Should I see any considerable difference using stratum mode rather than using this with eth-proxy? Currently mining on ethermine.
  • dlehenkydlehenky Member Posts: 2,249 ✭✭✭✭
    @rether Depends on what you call considerable. I see a bit of improvement over eth-proxy in the effective hash rate, which is the only thing that really matters. It's not a lot, though, maybe 2%-2.5% tops. It's hard to tell with the always present variance, both with your rig, the pool, and the network. I'm on ethpool.org. The network hash rate and difficulty has really come up in the last 12 hours or so, but the new DAG will hit tomorrow evening, so that should knock it back down for a bit.
  • user09342user09342 Member Posts: 8
    Here is my current 1.0.6 batch file:

    ethminer -U -F POOL --cl-extragpu-mem 3072 --cl-global-work 16384 --cl-local-work 128

    This is running on a GTX 980, but when it runs, the response time on my computer decreases (windows slow to respond, clicks slow to respond, mouse does not smoothly track, etc.). I read previously that "cuda-grid-size" could be used but I do not see that command in this version; any ideas on how to get this to be more responsive? I am OK to lose a bit of hashrate if I can use the computer while still browsing, etc. If needed I have a I7-2600K with 24 GB of RAM. Thanks.
  • GenoilGenoil 0xeb9310b185455f863f526dab3d245809f6854b4dMember Posts: 769 ✭✭✭
    @user09342 reduce global work until you're satisfied. I.e 2048 works fine for my 970.
  • user09342user09342 Member Posts: 8
    Genoil said:

    @user09342 reduce global work until you're satisfied. I.e 2048 works fine for my 970.

    Thanks a lot, that seems to have done the trick.
Sign In or Register to comment.