CUDA miner

1383941434469

Comments

  • blueboxbluebox Member Posts: 181 ✭✭
    @newkidONdablock Unless @Genoil has changed library dependency requirements that we all missed, yes, you need to have the CUDA toolkit installed for the version of ethminer you're running (older ones require 6.5, the latest ones 7.5).
  • LeChuckDELeChuckDE Member Posts: 46
    @thelosers106 , hashrates springing from 0 to something are most crashing drivers
  • GenoilGenoil 0xeb9310b185455f863f526dab3d245809f6854b4dMember Posts: 769 ✭✭✭
  • GenoilGenoil 0xeb9310b185455f863f526dab3d245809f6854b4dMember Posts: 769 ✭✭✭
    edited March 2016
    Uploaded 1.0.6b2 binaries. Changes:
    - fixed crash with -E flag when not using -R flag
    - set extra gpu mem to 0. rigs don't use displays anyway, useful for 2GB AMD cards these days.
    - fixed some building stuff

    https://github.com/Genoil/cpp-ethereum/blob/pick-devices/releases/ethminer-0.9.41-genoil-1.0.6b2.zip?raw=true

    I investigated the reported bug with --opencl-devices on a 4GPU AWS g2 instance. Worked flawlessly, so I'm not quite sure what the problem is. Can anybody verify that --opencl-devices 1 2 will indeed NOT use device 0? Ideally on an AMD rig, because the NV rig did fine.
  • workwork Member Posts: 2,084 ✭✭✭✭
    edited March 2016
    @genoil Newest build is for some reason taking a lot longer to "warm-up" then the previous build. CPU locked at 100% for over 5 minutes before any mining started. No, it didn't need to generate the DAG.

    5x R9 290 rig on win 10. It only took 1-2min to get going on 1.0.6b

    Edit: also, display values for new jobs have 0x in them twice, and solution found values don't have 0x at all. Minor detail, but looks pretty weird and makes it hard to see the number of 0s in the target as it whips by in the console. Unsure in which version this started.
  • GenoilGenoil 0xeb9310b185455f863f526dab3d245809f6854b4dMember Posts: 769 ✭✭✭
    @work thanks for your feedback. Fixed the 0x mixup :smile: . I;m also noticing slower startup, altough not as bad as you experience. I didn't change much in the code responsible for that. Do you get similar warmup time when using the benchmark? If not, then it's because of DAG size.
  • workwork Member Posts: 2,084 ✭✭✭✭
    edited March 2016
    Genoil said:

    @work thanks for your feedback. Fixed the 0x mixup :smile: . I;m also noticing slower startup, altough not as bad as you experience. I didn't change much in the code responsible for that. Do you get similar warmup time when using the benchmark? If not, then it's because of DAG size.

    It's running beautifully since it warmed up a bit over an hour ago. The processors on my rigs are the cheapest I could find, so that probably explains why my startup experience is worse then you observed.

    One thing I've noticed with your recent genoil builds, is that share submissions don't seem to tank hashrate anywhere near as badly as they used to. Before, if you monitored GPU usage, every time there was a share submission GPU usage dropped to 0% for a bit. Now, I'm only seeing tiny spikes to 0%. Pretty big improvement to total hash output, I'd estimate 2-5% increase in total share output.
    Post edited by work on
  • hozehoze Member Posts: 27
    Can someone provide me a link for installing ethminer in linux for nvidia 750ti ? Thx
  • Phoenix00Phoenix00 Member Posts: 61
    Can someone give me current hashrate for gtx760, gtx780 and gtx960, please?
  • miningpoolhubminingpoolhub Member Posts: 308 ✭✭
    This is the best miner currently.
    It works quite well on our mining pool hub using stratum+vardiff supported.

    @Genoil I couldn't find -S option from --help command.
    There's only -s simulation mode for similar option.

    Hope it adds detailed description about stratum.
    Thanks.
  • dexkzdexkz Member Posts: 3
    So. To sum up - If you want to mine on Windows 10 with CUDA, you need to:
    1. Uninstall the the current new driver
    2. Install 347 driver
    3. Install Cuda 6.5
    4. Use a 1.0.5 miner for cuda 6.5

    Am I mistaking somewhere?
  • blueboxbluebox Member Posts: 181 ✭✭
    work said:

    One thing I've noticed with your recent genoil builds, is that share submissions don't seem to tank hashrate anywhere near as badly as they used to. Before, if you monitored GPU usage, every time there was a share submission GPU usage dropped to 0% for a bit. Now, I'm only seeing tiny spikes to 0%. Pretty big improvement to total hash output, I'd estimate 2-5% increase in total share output.

    Noticed the same thing, and it happened when I switched from coinotron's getwork port to ethermine's proxy... I was using 1.0.5 in both instances, so it wasn't specific to 1.0.6b...

    ???
  • retherrether Member Posts: 258 ✭✭
    Phoenix00 said:

    Can someone give me current hashrate for gtx760, gtx780 and gtx960, please?

    My 760 is at about 12mh/s
  • blueboxbluebox Member Posts: 181 ✭✭
    I've seen 10-12MH for the 960's, so same as 760 (128-bit bus). 780's are about the same as 980/970, ~17-18MH stock clocks (256-bit bus). 9xx take less power than 7xx, 7xx are cheaper used of course.
  • Phoenix00Phoenix00 Member Posts: 61
    Thanks to @rether and @bluebox for infos :smiley:
  • mrpmrp Member Posts: 7
    edited March 2016
    ethminer 1.0.6b2 fail to cycle "GPU gave incorrect result" after 3 hours minig on AMD 8хGPU rig

    on GTX970x4 rig working perfectly
    Post edited by mrp on
  • DavBEDavBE Member Posts: 26
    Why do you guys use a cuda miner with amd gpus anyway ?
  • workwork Member Posts: 2,084 ✭✭✭✭
    edited March 2016
    DavBE said:

    Why do you guys use a cuda miner with amd gpus anyway ?

    It's the most up-to-date miner at the moment, with built-in stratum, and CUDA is just a feature. It still works with opencl.
  • alphageekalphageek Member Posts: 9
    @Genoil thanks for creating this CUDA-miner! ;)
  • alphageekalphageek Member Posts: 9
    edited March 2016
    bluebox said:

    I've seen 10-12MH for the 960's, so same as 760 (128-bit bus). 780's are about the same as 980/970, ~17-18MH stock clocks (256-bit bus). 9xx take less power than 7xx, 7xx are cheaper used of course.

    @bluebox Thanks for sharing these observations :)
  • dominusdominus Member Posts: 36
    dexkz said:

    So. To sum up - If you want to mine on Windows 10 with CUDA, you need to:
    1. Uninstall the the current new driver
    2. Install 347 driver
    3. Install Cuda 6.5
    4. Use a 1.0.5 miner for cuda 6.5

    Am I mistaking somewhere?

    I have tried the procedure.
    I have just downloaded 347 driver for win7/8.1 and installed it. The newer drivers were uninstalled automatically during the process.
    Then I have installed cuda 7.5, unchecking gpu drivers and other not needed options.
    The latest miner is now working and doing around 17MH with stock 970.

    But the problem is that nvidia-smi tool does not allow me to change clocks with those drivers. Any workaround or other method for forcing the P0 state?
  • GenoilGenoil 0xeb9310b185455f863f526dab3d245809f6854b4dMember Posts: 769 ✭✭✭
    Released 1.0.6 today. Still had a bug this morning so I quickly ironed that out (thanks @sp_). Fixes since 1.0.6b2:
    - fixed crash when using -E flag on system that hasn't run ethminer before or when -R directory isn't created yet
    - fixed stratum bug on coinmine.pl
    - corrected output console log of new stratum work and solutions
  • Phoenix00Phoenix00 Member Posts: 61
    Do someone know the maximum of gpu that the linux/windows driver can drive?
  • Mo35Mo35 Member Posts: 43
    dominus said:

    dexkz said:

    So. To sum up - If you want to mine on Windows 10 with CUDA, you need to:
    1. Uninstall the the current new driver
    2. Install 347 driver
    3. Install Cuda 6.5
    4. Use a 1.0.5 miner for cuda 6.5

    Am I mistaking somewhere?

    I have tried the procedure.
    I have just downloaded 347 driver for win7/8.1 and installed it. The newer drivers were uninstalled automatically during the process.
    Then I have installed cuda 7.5, unchecking gpu drivers and other not needed options.
    The latest miner is now working and doing around 17MH with stock 970.

    But the problem is that nvidia-smi tool does not allow me to change clocks with those drivers. Any workaround or other method for forcing the P0 state?
    cant comment on latest drivers , since im using 347 and cuda 6.5 on win10 , had problems with overclocking them too , the only tool that actually allowed me to change gpu/mem clocks with this setup was NV inspector "Nvidia Inspector"
  • Mo35Mo35 Member Posts: 43
    dexkz said:

    So. To sum up - If you want to mine on Windows 10 with CUDA, you need to:
    1. Uninstall the the current new driver
    2. Install 347 driver
    3. Install Cuda 6.5
    4. Use a 1.0.5 miner for cuda 6.5

    Am I mistaking somewhere?

    you dont need to install cuda6.5 , cuda 6.5 dll in miners folder is just fine , but you should prevent win10 to update that driver , which it will do even if you disable driver updates in control panel , there is some tips on how to in this forum
  • DPS50DPS50 Member Posts: 79
    I want to run 2 AMD cards with the CUDA miner and a 3rd AMD card with a separate CUDA miner in a new command window.
    How can i do that ?

    Thanks in advance :smile:
  • GenoilGenoil 0xeb9310b185455f863f526dab3d245809f6854b4dMember Posts: 769 ✭✭✭
    edited March 2016
    Why? But this should work with 1.0.6:

    ethminer.exe -G --opencl-devices 0 1 -F ...
    ethminer.exe -G --opencl-devices 2 -R C:\DAG -F ...

    But CUDA on AMD? No..
  • hashmogulhashmogul Member Posts: 1
    Hi Genoil,
    I am new in using your ethminer 1.0.6. I tried to configure it with stratum support for nanopool, but I could´t make it. Can you please give an example how to configure the bat file with stratum support? THANKS!
  • DPS50DPS50 Member Posts: 79
    edited March 2016
    @Genoil can you please write me the full commands ? (for coinotron)
    0 1 does the define the GPU 1 and GPU 2 ?
    And as i understood i need a new dag file for the 3rd card that will run in a separate command window ....
    I want to run one of my cards in an other pool :)
  • mrpmrp Member Posts: 7
    is --farm-recheck needing in last versions ?
Sign In or Register to comment.