CUDA miner

1363739414269

Comments

  • LeChuckDELeChuckDE Member Posts: 46
    edited March 2016
    the GTX950M is a GM107 Card was also wondering about it but find the result with cuda support of ethminer fantastic. round about 50 % Powerboost at Cuda against CL.

    Got 5,5 MH/s at 65 W now will look forward to get some GTX9*0M Devices together for mining.

  • 0x000x00 Member Posts: 55
    Does the genoil-1.0.5 version support stratum connections to ethpool or do I still have to go through ethproxy?
  • GenoilGenoil 0xeb9310b185455f863f526dab3d245809f6854b4dMember Posts: 769 ✭✭✭
    @0x00 yes it does. but i can't guarantee bugfree operation. i don't actively mine so i don't have much experience in how stable it is. some people have reported instabilities which i have tried to resolve but i'm not sure how succesful i've been at it.

    when i'm testing new code (like at this moment for 1.0.6) I always use stratum against coinotron. works fine for relatively short periods.
  • 0x000x00 Member Posts: 55
    @Genoil Is there a specific release I must use for stratum support? because I can't seem to find the options for it, normally I would use something like -s and -u for qtminer. If there is, could you point me to a link? What I downloaded was from https://github.com/Genoil/cpp-ethereum/tree/master/releases
  • GenoilGenoil 0xeb9310b185455f863f526dab3d245809f6854b4dMember Posts: 769 ✭✭✭
    ethminer.exe -S coinotron.com:3344 -O User.rig:pass
  • LeChuckDELeChuckDE Member Posts: 46
    @0x00 and don't forget to enable it in CMakefile before compiling it
  • LeChuckDELeChuckDE Member Posts: 46
    Has someone tested NVIDIA Quadro FX 5800 Devices for mining ?
  • blueboxbluebox Member Posts: 181 ✭✭
    Genoil said:

    /offtopic I'm still puzzled by the naming strategy of mobile NVidia GPU's...You'd expect that 950M was a GM206 (Maxwell 2) part, while actually it's only a GM107 part (Maxwell 1). Hence the 5.5MH/s rate, which puts it on par with a desktop 750Ti.

    @Genoil /offtopic-response Both nVidia and AMD play the same game. An R9 M390 is a cut-down Tonga-based 380 chip, not a Hawaii class 390.

    They can use the same numbering scheme for mobile and desktop, but the "M" changes it enough that they wouldn't be legally liable for false advertising, etc. And you thought there was a technical reason... :D
  • blueboxbluebox Member Posts: 181 ✭✭
    Genoil said:

    when i'm testing new code (like at this moment for 1.0.6) I always use stratum against coinotron. works fine for relatively short periods.

    Unfortunately, coinotron's stratum and getwork ports have been puking rpc errors far too often in the past few weeks or so, losing more mining time. Not sure what's going on. Haven't tried your stratum function against another pool yet, but I'm going to switch over to ethermine.org tonight, stratum or proxy...
  • eddy4eddy4 Member Posts: 30
    How can I check the hash rate with ethminer for each GPU if you have a rig with 2 GPUs?
  • excavateexcavate Member Posts: 7
    First post!

    Thank you @Genoil for your hard work with the software.

    I am curious if my setup is producing the right hashrate. 5 GTX 960s, 1000W gold level, PSU, 16gb ddr3, i5, ssd on Ubuntu 14.04. This is my output from nvidia-smi

    +------------------------------------------------------+
    | NVIDIA-SMI 352.63 Driver Version: 352.63 |
    |-------------------------------+----------------------+----------------------+
    | GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
    | Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
    |===============================+======================+======================|
    | 0 GeForce GTX 960 Off | 0000:01:00.0 On | N/A |
    | 23% 68C P2 76W / 160W | 1645MiB / 2044MiB | 100% Default |
    +-------------------------------+----------------------+----------------------+
    | 1 GeForce GTX 960 Off | 0000:03:00.0 Off | N/A |
    | 26% 69C P2 83W / 160W | 1406MiB / 2047MiB | 100% Default |
    +-------------------------------+----------------------+----------------------+
    | 2 GeForce GTX 960 Off | 0000:04:00.0 Off | N/A |
    | 29% 71C P2 81W / 160W | 1406MiB / 2047MiB | 100% Default |
    +-------------------------------+----------------------+----------------------+
    | 3 GeForce GTX 960 Off | 0000:05:00.0 Off | N/A |
    | 30% 71C P2 76W / 160W | 1406MiB / 2047MiB | 100% Default |
    +-------------------------------+----------------------+----------------------+
    | 4 GeForce GTX 960 Off | 0000:06:00.0 Off | N/A |
    | 19% 67C P2 77W / 160W | 1406MiB / 2047MiB | 100% Default |
    +-------------------------------+----------------------+----------------------+

    Wattage and fan speed is lower than I expected. Is this just due to power efficiency or is there more room to push these cards? Here is what I'm hashing at:

    miner 21:59:52|ethminer Mining on PoWhash #14fe9b4e… : 52428800 H/s = 10485760 hashes / 0.2 s
    miner 21:59:52|ethminer Mining on PoWhash #14fe9b4e… : 52428800 H/s = 10485760 hashes / 0.2 s
    miner 21:59:52|ethminer Mining on PoWhash #14fe9b4e… : 52428800 H/s = 10485760 hashes / 0.2 s
    miner 21:59:53|ethminer Mining on PoWhash #14fe9b4e… : 52428800 H/s = 10485760 hashes / 0.2 s
    miner 21:59:53|ethminer Mining on PoWhash #14fe9b4e… : 62914560 H/s = 12582912 hashes / 0.2 s
    miner 21:59:53|ethminer Mining on PoWhash #14fe9b4e… : 52428800 H/s = 10485760 hashes / 0.2 s
    miner 21:59:53|ethminer Mining on PoWhash #14fe9b4e… : 52428800 H/s = 10485760 hashes / 0.2 s
    miner 21:59:53|ethminer Mining on PoWhash #14fe9b4e… : 52428800 H/s = 10485760 hashes / 0.2 s
    miner 21:59:54|ethminer Mining on PoWhash #14fe9b4e… : 52428800 H/s = 10485760 hashes / 0.2 s
    miner 21:59:54|ethminer Mining on PoWhash #14fe9b4e… : 62914560 H/s = 12582912 hashes / 0.2 s
    miner 21:59:54|ethminer Mining on PoWhash #14fe9b4e… : 52428800 H/s = 10485760 hashes / 0.2 s

    Stratum pool shows average 50MH/s.

    My other rig has two R9 390's and does 58 MH/s. I know AMD chipsets for ethereum outperform NVIDIA at the moment but does this hashrate look correct?

    Would it be safe to slighly overclock these EVGA cards? They state they are factory overclocked already?
  • GenoilGenoil 0xeb9310b185455f863f526dab3d245809f6854b4dMember Posts: 769 ✭✭✭
    edited March 2016
    @excavate 10MH per 960 sounds about right. Are you using stock OpenCL or my CUDA miner? The latter is faster.

    Overclocking the RAM should give you better speeds.

    The main reason the 950 and 960 are relatively slow compared to their AMD gaming counterparts is the narrow bus width.

    @eddy4 not possible at this time. It was already on my future feature list. First things first though.

    1.0.6 will have pool failover, automatic dag file cleanup and OpenCL device picking. The latter is in alpha btw but nobody seems to have noticed.
  • Bensam123Bensam123 Member Posts: 62
    How about automatic cuda device picking? Right now you have to specify cuda devices... I guess it's not that big of a issue as you can add like seven cuda devices even if you have one and it just wont use the other six so it works on pretty much any machine.
  • GenoilGenoil 0xeb9310b185455f863f526dab3d245809f6854b4dMember Posts: 769 ✭✭✭
    @Bensam123 yes that should be fixed as well. I haven't tested it with a multi GPU setup yet. Feel free to download alpha from the pick-devices branch.
  • excavateexcavate Member Posts: 7
    @Genoil I'm using your cudaminer. I'm not experienced with overclocking GPU's but I've got everything ready to go with nvidia-settings (gpu's on p0). The top end for overclock seems a bit outrageous... what would be a safe amount to bump the RAM?

    Do the attributes for your miner look correct? I just played around with numbers on cuda-streams and mining-threads. Not sure how those settings relate to number of video cards.

    --farm-recheck 100 --cuda-devices 0 1 2 3 4 -v 2 --cuda-schedule sync --cuda-streams 2 --cuda-grid-size 16576 --mining-threads 5

  • GenoilGenoil 0xeb9310b185455f863f526dab3d245809f6854b4dMember Posts: 769 ✭✭✭
    --mining-threads 5 and --cuda-devices 0 1 2 3 4 essentially is the same. But until 1.0.6 you have to specify ---cuda-devices. You can omit --mining-threads 5. --cuda-streams is per GPU. It allows for a tiny bit of overlap between kernel runtime and copying results back to the host. Anything over 2 is pointless.
  • philakonephilakone Member Posts: 163 ✭✭
    @Genoil Do you ever sleep man? haha
  • GenoilGenoil 0xeb9310b185455f863f526dab3d245809f6854b4dMember Posts: 769 ✭✭✭
    1.0.6b is out. It appears to work fine, but I'm just leaving it in beta for a bit because I couldn't test the new opencl-devices flag and fix of the cuda-devices flag yet. So please test and let me know if it is working or not:

    Changes:
    --opencl-devices: flag allows you to specify which opencl devices you want to use. 0 based, space delimited.

    -FF/--farm-failover flag allows you to specify failover farm URL. this was requested by one of my donators with one specific goals in mind: the ability to run a failover stratum proxy for added stability. clever. this is also the main reason i haven't added a native stratum failover.

    --failover-retries. the name speaks for itself

    -when mining with multiple CUDA cards, you shouldn't have to specify all cards any longer with --cuda-devices. Instead it will try to use them all. Please test on multi-GPU rig.

    -E/--erase-dags : Erase unneeded DAG files. Default is 'none'. Possible values are:
    none - don't erase DAG files (default)" << endl
    old - erase all DAG files older than current epoch" << endl
    bench - like old, but keep epoch 0 for benchmarking" << endl
    all - erase all DAG files After deleting all files, setting changes to none." << endl

    You'll still need about 3 times the DAG size of disk space when running the 'old' setting. 1 for current DAG, 1 for next DAG and one extra during transition to next DAG. I think I could make the requirement just 2 times the DAG size, but this was easier to implement :)

    https://github.com/Genoil/cpp-ethereum/raw/pick-devices/releases/ethminer-0.9.41-genoil-1.0.6b.zip

    If you have ideas for new features, you're welcome! Currently I have this on my list:
    - change --cuda-grid-size and --cl-global-work to more common "--intensity" parameter
    - config file
    - per device configuration
    - mixed platform mining (AMD + NVidia on a single instance)
    - serial DAG loading for multiple GPU's
    - NVidia specific OpenCL optimizations (just for kicks)
    - more stable stratum support


  • newkidONdablocknewkidONdablock interwebzMember Posts: 121
    Holy smoking batman, man you're just amazing. I will be making a mixed (2/2 nvidia/amd) rig in the next couple of days so I can perhaps help you/community benchmark/debug/test certain issues.
  • LeChuckDELeChuckDE Member Posts: 46
    edited March 2016
    @Genoil , just for info.
    Tried to compile it without stratum support enabled, faild referenting to missing stratum libs.
    Enabled stratum support in CMakefilelist.txt, builds without any problems.

    Ubuntu 15.10, Cuda 7.5 Driver 352 NVIDIA GTX950M

    Speedtests will be come later.

    Edit:
    -U = 5.6 MH/s with GTX950M
    -G = 4.0 MH/s with GTX950M

    possible that the fix has speed up CL Mining on NVIDIA a little bit, but i think it is normal varrity at the hashrate.
  • sikuntsikunt Member Posts: 13
    my 270x cards are 2gb and since the new dag they wont run due to not being able to take the whole chunk, any workaround for this?
  • LeChuckDELeChuckDE Member Posts: 46
    try GPU_SINGLE_ALLOC_PERCENT=100... ur 270x reservers memory and the left space isn't enough
  • GenoilGenoil 0xeb9310b185455f863f526dab3d245809f6854b4dMember Posts: 769 ✭✭✭
    also add --cl-extragpu-mem 0 when you're not connecting a display on these cards.
  • bitmaxtbitmaxt Member Posts: 81
    Hi can you give me the hashrate of gtx660 and gtx680 on ETH?
  • LeChuckDELeChuckDE Member Posts: 46
    if u send me both cards, sure i could give u the benchmark results
  • newkidONdablocknewkidONdablock interwebzMember Posts: 121
    bitmaxt said:

    Hi can you give me the hashrate of gtx660 and gtx680 on ETH?

    getting 15-16 on my gtx680
  • sikuntsikunt Member Posts: 13
    Genoil said:

    also add --cl-extragpu-mem 0 when you're not connecting a display on these cards.

    cheers this fixed it for me
  • maksimlyamaksimlya Member Posts: 5
    what should i write for my gtx860m?
  • maksimlyamaksimlya Member Posts: 5


    are these numbers fine for 2 x gtx970?
    running them with
    "ethminer.exe --farm-recheck 200 -U -F http://127.0.0.1:8080/ --cl-extragpu-mem 0
    pause"
    monitor connected to intel's vga card.
    but on my y50 laptop with 860m getting just 0.6mh/s on windows on stock erh miner, on linux it raises to
    3-4 mh/s
    what could i do to improve it?
    ewr.jpg 550.8K
  • cgladuecgladue Member Posts: 2
    How long do you guys think it will be before the DAG gets to 2GB ?
Sign In or Register to comment.