Im bit stuck with latest 1.0.3 miner , unfortunately i had to slice up partition to fit win7 for mining purposes , and i have limited amount of disk space , so 1.0.3 versions DAG loc switch is must for me. i have 970 and 2x 950 cards , but with --cuda switch im unable to force miner to use all cards it just sees 1st 970 , using --opencl works with all cards and properly announces them as cuda1.2. Any suggestions what im missing ? why with --cuda only 1st card shows up ?
@restless only with compute 5.2 cards. 750ti is Compute 5.0
@Genoil GTX750Ti under Win7 does not work - this is clear. But is the Linux another story? Here and there I've got 4 GTX750Ti cards in different PC all under Win7. I wonder will it do any good if I place them into one Linux farm?
@Mo35 hmm this sounds like a bug. I've read the same somewhere else today. Anyway, try adding --cuda-devices 0 1 2 to the command arguments. This should force the use of three cards.
thanks that did the job , 3 cards 970+ 2x950 nets me close 40Mhz , impressive ! i just wish i can get rid of win 7 , hate it ... i was happily mining with my old 347.52 driver till i got 950 cards , support for 950 cards apeared just in already screwed 355.69 ... maybe sugggestions how to fool win10 to use 347.52 win7 driver for 950 (it does support 960) , probably some inf tinkering or something deeper is necesary ?
Well , actually i did managed to force all cards 970 and 950 to work with 347.52 under Windows 10 even 950 wasn't around at the time of driver release.
if anyone wants to tinker , here is what i did , i found under devices my hardware ID for gtx950 , looks something like this PCI\VEN_10DE&DEV_1402&SUBSYS_85551043 , 1402 is model hw identificator , since gtx960 was close pick , i checked in driver inf files for 960 ID , it was 1401 , so i crawled through infs and searched for 1401 id. And then made changes to fool installer, that my 1402 card actualy is one that that is supported , bunch of entries like this :%NVIDIA_DEV.1401%= Section034, PCI\VEN_10DE&DEV_1402 so on.
You will have to reboot in recovery with driver signature verification disabled to install this changed driver , also retarded Microsoft driver update prevention doesn't really works as it should, it still updates nvidias drivers even its prohibited via Contr.Panel\system\advanced settings\hardware\device installation settings, so i ended up making local computer group policies to prevent windows from touching those devices. Hit run type gpedit.msc and go to Computer Config\Administrative templ.\System\Device Installation\Device Installation Restriction and open policy named "Prevent installation of devices that match any of these Device IDs" , go to your device manager and look under you GPU`S details\Matching Device Id for your card ID values (something like pci\ven_10de&dev_13c2 ), enable policy and add your card ID strings to policy , windows wont be able to replace driver for those devices. Reboot , enable net , enable updates , recheck that your drivers sits on 347.52 and you are set to mine under Win10 for now at least with --opencl at full hashpower.
There is one thing tho with --cuda , you wont be able to mine with 1.0.3 unfortunately , because miner crashes with "cuda driver version is insufficient for cuda runtime version" , even if i copy cudart64_65.dll into miner dir , it wont pick it up , probably we need someone on linux to recompile 1.0.3 with older imho 1.1 cuda support.
Anyone ?
p.s. there is older release with cuda 1.1 support around like 1.0.1 ?, ask author , tho it performs worse (~-2-3Mhz from 40Mhz) and lack newly implemented functionality unfortunately , but at least you can mine with cuda not cl
@Mo35 I think the message means that the binary should be rebuilt with a version of CUDA that was around at the time the 347.52 drivers were released. So 6.5 or 7.0.
If you can do what you did there with Windows, you should try building ethminer ;-)
does 350.12 works with full hashpower too ? unfortunately even 350.12 dosnt include support for newest cards , like gtx950 sooo still , inf tinkering is must.
@Genoil , im not much into linux`es , but for thos 3 additional Mhz , and -R functionality i will probably have to look into , i was just hoping someone with setup for compilation can do that less painfully
@Mo35 you don't need Linux to roll your own ethminer. Building on Windows is a bit botched at the moment, but I'll try to make that a bit more straightforward.
@Genoil , could you please point me to "how to" for win , if its available , i will try to mop up something tonight , thats if i figure out what im suppose to do
@restless this driver workaround is for Compute 5.2 (GTX900 series) on Win8/10.
@Mo35 1. Download and install CMake and Visual Studio 12 2013 (community edition is free and fine) 2. Download and install CUDA Toolkit 6.5 to make sure you don't get the error message (not 100% sure if this is the cause, but there's no performance hit anyway for using an older CUDA toolkit). Make sure to only install the toolkit and not the bundled graphics driver 3. Download or check out the source code from github 4. run getstuff.bat in cpp-ethereum/extdep. This will take a while. 5. Create a directory "build" inside "cpp-ethereum" and go there with a command prompt 6. run cmake -DBUNDLE=cudaminer -G "Visual Studio 12 2013 Win64" .. 7. open ethereum.sln in build directory 8. go to right pane, right click ethminer and choose "Set as startup project" 9. Make sure build Type is "Release" in top toolbar 10. Press Play button in top toolbar 11. Build will fail
Now comes the part where it's botched. There is a file called BuildInfo.h that is only created at build time. It ends up in the build directory, but the VC project assumes it is somewhere else. Navigate to the right pane, open the Buildinfo project and look for header file "BuildInfo.h". Right click to remove it from the project and add the correct file Buildinfo.h from the build dir by "Add existing file to project".
I really should fix this but I'm not very good with CMake..
new 1.0.4 beta release is up. supports stratum mining on coinotron. please note it's a beta release. it's looking good, but it would really help if some of you could test it for a while.
any idea why mu Ubuntu compiles athereum but sees no opencl/cuda devices?! I installed latest NVidia drivers - 352... but obviously something more is needed
ethminer -U -M Genoil's ethminer 0.9.41-genoil-1.0.3 ===================================================================== Forked from github.com/ethereum/cpp-ethereum CUDA kernel ported from Tim Hughes' OpenCL kernel With contributions from RoBiK, tpruvot and sp_
Please consider a donation to: ETH: 0xeb9310b185455f863f526dab3d245809f6854b4d
CUDA error in func 'getNumDevices' at line 112 : unknown error. terminate called after throwing an instance of 'std::runtime_error' what(): unknown error Aborted (core dumped)
for opencl it says
[OPENCL]:No OpenCL platforms found No GPU device with sufficient memory was found. Can't GPU mine. Remove the -G argument
@barabeku if it didn't in the previous version, it won't now. If you send me a few ETH I can spend a few hours of debugging on a $2.60/hr AWS console to see why it doesn't work. Yay.
Normally I don't mine myself, but have been running the stratum miner for a 8 hours now without any problems. Hashrate display in the miner will likely show higher values (about 1MH/s extra for my GTX780), but that's just the result of a different way things are handled. @dr_pra of ethpool notified me it also works on ethpool. He also brought to my attention that I still need to implement automatic reconnect on lost connection.
@Genoil — Didn't work with your usage pattern, tried adding http:// and other typical patterns, but no go... what would I be doing wrong?
C:\Users\<...>\Documents\ethereum-mining-windows\ethminer-0.9.41-genoil-10.0.4>ethminer.exe --farm-recheck 200 -U -S coinotron:3344 -O .:
Genoil's ethminer 0.9.41-genoil-1.0.4
=====================================================================
Forked from github.com/ethereum/cpp-ethereum
CUDA kernel ported from Tim Hughes' OpenCL kernel
With contributions from RoBiK, tpruvot and sp_
Please consider a donation to:
ETH: 0xeb9310b185455f863f526dab3d245809f6854b4d
Using grid size 8192, block size 128
Found suitable CUDA device [GeForce GTX 970] with 4294967296 bytes of GPU memory
miner 20:23:39|main Waiting for work package...
miner 20:23:39|main Waiting for work package...
miner 20:23:39|main Waiting for work package...
miner 20:23:39|main Waiting for work package...
miner 20:23:40|main Waiting for work package...
miner 20:23:40|main Waiting for work package...
miner 20:23:40|main Waiting for work package...
miner 20:23:40|main Waiting for work package...
miner 20:23:40|main Waiting for work package...
miner 20:23:41|main Waiting for work package...
miner 20:23:41|main Waiting for work package...
miner 20:23:41|main Waiting for work package...
Could not resolve hostcoinotron:3344, No such host is known i 20:23:41| Disconnecting
C:\Users\<...>\Documents\ethereum-mining-windows\ethminer-0.9.41-genoil-10.0.4>PAUSE
Press any key to continue . . .
@Genoil 6. run cmake -DBUNDLE=cudaminer -G "Visual Studio 12 2013 Win64" .. 7. open ethereum.sln in build directory
well this is how far i got ... with cmake -DBUNDLE=cudaminer -G "Visual Studio 12 2013 Win64" , no ethereum.sln apears in build dir , some kind of errors in cmake logs wich unfortunately means nothing to me
Any hint how to make ethminer see my 750ti card under Ubuntu 15.04? I have all drivers installed, ethminer is compiled, but --list-devices returns nothing. nvidia-settings shows NV drivers up and running, clinfo shows NV opencl active... out of ideas, except starting from scratch
I used https://wiki.tiker.net/OpenCLHowTo#Installing_the_Nvidia_ICD compiled the example, ./print-devices shows my 750Ti as valid opencl platform running ./cl-demo 1000000 10 gives valid results, ie OTHER programs do recognize the device, its only ethminer which doesn't ?!
Ethminer --list-devices still doesnt list anything, but forcing via --opencl-device 0 or --cuda-devices 0 correctly finds the card and benchmarks around 8MH/s
@antonio8 the thread that displays hashrate is now doing nothing but count hashrate, so it may actually be more accurate, actual hashing speed is not better.
@Genoil Ha, dumb on my part actually - in a hurry and copy-pasted, should've seen that...
Two 970's getting 40MH/s avg, up from 38 on 1.0.3, seems stable. I'll take 5% anytime, thanks! Don't have much eth, but will send a tip someday soon, might just buy some now that it's down.
Comments
Here and there I've got 4 GTX750Ti cards in different PC all under Win7. I wonder will it do any good if I place them into one Linux farm?
if anyone wants to tinker , here is what i did , i found under devices my hardware ID for gtx950 , looks something like this PCI\VEN_10DE&DEV_1402&SUBSYS_85551043 , 1402 is model hw identificator , since gtx960 was close pick , i checked in driver inf files for 960 ID , it was 1401 , so i crawled through infs and searched for 1401 id. And then made changes to fool installer, that my 1402 card actualy is one that that is supported , bunch of entries like this :%NVIDIA_DEV.1401%= Section034, PCI\VEN_10DE&DEV_1402 so on.
You will have to reboot in recovery with driver signature verification disabled to install this changed driver , also retarded Microsoft driver update prevention doesn't really works as it should, it still updates nvidias drivers even its prohibited via Contr.Panel\system\advanced settings\hardware\device installation settings, so i ended up making local computer group policies to prevent windows from touching those devices. Hit run type gpedit.msc and go to Computer Config\Administrative templ.\System\Device Installation\Device Installation Restriction and open policy named "Prevent installation of devices that match any of these Device IDs" , go to your device manager and look under you GPU`S details\Matching Device Id for your card ID values (something like pci\ven_10de&dev_13c2 ), enable policy and add your card ID strings to policy , windows wont be able to replace driver for those devices. Reboot , enable net , enable updates , recheck that your drivers sits on 347.52 and you are set to mine under Win10 for now at least with --opencl at full hashpower.
There is one thing tho with --cuda , you wont be able to mine with 1.0.3 unfortunately , because miner crashes with "cuda driver version is insufficient for cuda runtime version" , even if i copy cudart64_65.dll into miner dir , it wont pick it up , probably we need someone on linux to recompile 1.0.3 with older imho 1.1 cuda support.
Anyone ?
p.s. there is older release with cuda 1.1 support around like 1.0.1 ?, ask author , tho it performs worse (~-2-3Mhz from 40Mhz) and lack newly implemented functionality unfortunately , but at least you can mine with cuda not cl
If you can do what you did there with Windows, you should try building ethminer ;-)
@Genoil , im not much into linux`es , but for thos 3 additional Mhz , and -R functionality i will probably have to look into , i was just hoping someone with setup for compilation can do that less painfully
edit: am i looking @ correct places ?
https://github.com/ethereum/webthree-umbrella/wiki/Building-on-Windows
https://github.com/Genoil/cpp-ethereum
at the moment still kinda clueless what to do with all that besides , installing VS and trying to feed it something to compile
PS:
Installed a 750Ti on Ubuntu
tried running ethminer and ... even copying lib file into directory of ethminer won't change the errors
Make install runs flawlessly (imho)
[ 32%] Built target devcore
[ 34%] Built target secp256k1
[ 49%] Built target scrypt
[ 60%] Built target devcrypto
[ 67%] Built target ethash
[ 70%] Built target ethash-cl
[ 98%] Built target ethcore
[100%] Built target ethminer
Install the project...
-- Install configuration: "RelWithDebInfo"
-- Up-to-date: /usr/local/lib/libdevcore.so
-- Up-to-date: /usr/local/include/devcore/CommonIO.h
-- Up-to-date: /usr/local/include/devcore/TrieCommon.h
-- Up-to-date: /usr/local/include/devcore/FixedHash.h
-- Up-to-date: /usr/local/include/devcore/CommonJS.h
-- Up-to-date: /usr/local/include/devcore/Guards.h
-- Up-to-date: /usr/local/include/devcore/Assertions.h
-- Up-to-date: /usr/local/include/devcore/Common.h
-- Up-to-date: /usr/local/include/devcore/CommonData.h
-- Up-to-date: /usr/local/include/devcore/StructuredLogger.h
-- Up-to-date: /usr/local/include/devcore/Worker.h
-- Up-to-date: /usr/local/include/devcore/TransientDirectory.h
-- Up-to-date: /usr/local/include/devcore/UndefMacros.h
-- Up-to-date: /usr/local/include/devcore/db.h
-- Up-to-date: /usr/local/include/devcore/Exceptions.h
-- Up-to-date: /usr/local/include/devcore/FileSystem.h
-- Up-to-date: /usr/local/include/devcore/Diff.h
-- Up-to-date: /usr/local/include/devcore/debugbreak.h
-- Up-to-date: /usr/local/include/devcore/picosha2.h
-- Up-to-date: /usr/local/include/devcore/Log.h
-- Up-to-date: /usr/local/include/devcore/TrieDB.h
-- Up-to-date: /usr/local/include/devcore/Terminal.h
-- Up-to-date: /usr/local/include/devcore/Hash.h
-- Up-to-date: /usr/local/include/devcore/SHA3.h
-- Up-to-date: /usr/local/include/devcore/TrieHash.h
-- Up-to-date: /usr/local/include/devcore/vector_ref.h
-- Up-to-date: /usr/local/include/devcore/Base64.h
-- Up-to-date: /usr/local/include/devcore/RLP.h
-- Up-to-date: /usr/local/include/devcore/RangeMask.h
-- Up-to-date: /usr/local/include/devcore/concurrent_queue.h
-- Up-to-date: /usr/local/include/devcore/MemoryDB.h
-- Up-to-date: /usr/local/lib/libsecp256k1.so
-- Up-to-date: /usr/local/include/secp256k1/scalar.h
-- Up-to-date: /usr/local/include/secp256k1/scalar_impl.h
-- Up-to-date: /usr/local/include/secp256k1/ecmult.h
-- Up-to-date: /usr/local/include/secp256k1/ecdsa.h
-- Up-to-date: /usr/local/include/secp256k1/field_impl.h
-- Up-to-date: /usr/local/include/secp256k1/num.h
-- Up-to-date: /usr/local/include/secp256k1/scalar_8x32.h
-- Up-to-date: /usr/local/include/secp256k1/ecmult_gen_impl.h
-- Up-to-date: /usr/local/include/secp256k1/scalar_4x64_impl.h
-- Up-to-date: /usr/local/include/secp256k1/field_5x52.h
-- Up-to-date: /usr/local/include/secp256k1/group.h
-- Up-to-date: /usr/local/include/secp256k1/scalar_4x64.h
-- Up-to-date: /usr/local/include/secp256k1/field_10x26_impl.h
-- Up-to-date: /usr/local/include/secp256k1/ecmult_impl.h
-- Up-to-date: /usr/local/include/secp256k1/group_impl.h
-- Up-to-date: /usr/local/include/secp256k1/ecmult_gen.h
-- Up-to-date: /usr/local/include/secp256k1/num_gmp.h
-- Up-to-date: /usr/local/include/secp256k1/field_5x52_asm_impl.h
-- Up-to-date: /usr/local/include/secp256k1/ecdsa_impl.h
-- Up-to-date: /usr/local/include/secp256k1/hash.h
-- Up-to-date: /usr/local/include/secp256k1/num_gmp_impl.h
-- Up-to-date: /usr/local/include/secp256k1/util.h
-- Up-to-date: /usr/local/include/secp256k1/field.h
-- Up-to-date: /usr/local/include/secp256k1/hash_impl.h
-- Up-to-date: /usr/local/include/secp256k1/field_5x52_impl.h
-- Up-to-date: /usr/local/include/secp256k1/field_5x52_int128_impl.h
-- Up-to-date: /usr/local/include/secp256k1/eckey.h
-- Up-to-date: /usr/local/include/secp256k1/num_impl.h
-- Up-to-date: /usr/local/include/secp256k1/eckey_impl.h
-- Up-to-date: /usr/local/include/secp256k1/libsecp256k1-config.h
-- Up-to-date: /usr/local/include/secp256k1/scalar_8x32_impl.h
-- Up-to-date: /usr/local/include/secp256k1/field_10x26.h
-- Up-to-date: /usr/local/lib/libscrypt.so
-- Up-to-date: /usr/local/include/scrypt/slowequals.h
-- Up-to-date: /usr/local/include/scrypt/libscrypt.h
-- Up-to-date: /usr/local/include/scrypt/crypto_scrypt-hexconvert.h
-- Up-to-date: /usr/local/include/scrypt/b64.h
-- Up-to-date: /usr/local/include/scrypt/sysendian.h
-- Up-to-date: /usr/local/include/scrypt/sha256.h
-- Up-to-date: /usr/local/lib/libdevcrypto.so
-- Up-to-date: /usr/local/include/devcrypto/CryptoPP.h
-- Up-to-date: /usr/local/include/devcrypto/Common.h
-- Up-to-date: /usr/local/include/devcrypto/AES.h
-- Up-to-date: /usr/local/include/devcrypto/Exceptions.h
-- Up-to-date: /usr/local/include/devcrypto/OverlayDB.h
-- Up-to-date: /usr/local/include/devcrypto/WordList.h
-- Up-to-date: /usr/local/include/devcrypto/SecretStore.h
-- Up-to-date: /usr/local/include/devcrypto/ECDHE.h
-- Up-to-date: /usr/local/lib/libethash.so
-- Up-to-date: /usr/local/lib/libethash-cl.so
-- Up-to-date: /usr/local/include/ethash-cl/ethash_cl_miner.h
-- Up-to-date: /usr/local/include/ethash-cl/ethash_cl_miner_kernel.h
-- Up-to-date: /usr/local/lib/libethcore.so
-- Up-to-date: /usr/local/include/ethcore/Params.h
-- Up-to-date: /usr/local/include/ethcore/Ethash.h
-- Up-to-date: /usr/local/include/ethcore/EthashSealEngine.h
-- Up-to-date: /usr/local/include/ethcore/CommonJS.h
-- Up-to-date: /usr/local/include/ethcore/EthashAux.h
-- Up-to-date: /usr/local/include/ethcore/EthashCUDAMiner.h
-- Up-to-date: /usr/local/include/ethcore/KeyManager.h
-- Up-to-date: /usr/local/include/ethcore/Common.h
-- Up-to-date: /usr/local/include/ethcore/Farm.h
-- Up-to-date: /usr/local/include/ethcore/ICAP.h
-- Up-to-date: /usr/local/include/ethcore/EthashGPUMiner.h
-- Up-to-date: /usr/local/include/ethcore/Sealer.h
-- Up-to-date: /usr/local/include/ethcore/Exceptions.h
-- Up-to-date: /usr/local/include/ethcore/Miner.h
-- Up-to-date: /usr/local/include/ethcore/ABI.h
-- Up-to-date: /usr/local/include/ethcore/EthashCPUMiner.h
-- Up-to-date: /usr/local/include/ethcore/BasicAuthority.h
-- Up-to-date: /usr/local/include/ethcore/BlockInfo.h
-- Up-to-date: /usr/local/include/ethcore/Transaction.h
-- Up-to-date: /usr/local/bin/ethminer
@Mo35
1. Download and install CMake and Visual Studio 12 2013 (community edition is free and fine)
2. Download and install CUDA Toolkit 6.5 to make sure you don't get the error message (not 100% sure if this is the cause, but there's no performance hit anyway for using an older CUDA toolkit). Make sure to only install the toolkit and not the bundled graphics driver
3. Download or check out the source code from github
4. run getstuff.bat in cpp-ethereum/extdep. This will take a while.
5. Create a directory "build" inside "cpp-ethereum" and go there with a command prompt
6. run
cmake -DBUNDLE=cudaminer -G "Visual Studio 12 2013 Win64" ..
7. open ethereum.sln in build directory
8. go to right pane, right click ethminer and choose "Set as startup project"
9. Make sure build Type is "Release" in top toolbar
10. Press Play button in top toolbar
11. Build will fail
Now comes the part where it's botched. There is a file called BuildInfo.h that is only created at build time. It ends up in the build directory, but the VC project assumes it is somewhere else. Navigate to the right pane, open the Buildinfo project and look for header file "BuildInfo.h". Right click to remove it from the project and add the correct file Buildinfo.h from the build dir by "Add existing file to project".
I really should fix this but I'm not very good with CMake..
Build again and it should work.
https://github.com/Genoil/cpp-ethereum/tree/stratum/releases
usage ethminer.exe -U (or -G) -S coinotron.com:3344 -O Username.Workername:workerpass
I installed latest NVidia drivers - 352... but obviously something more is needed
Genoil's ethminer 0.9.41-genoil-1.0.3
=====================================================================
Forked from github.com/ethereum/cpp-ethereum
CUDA kernel ported from Tim Hughes' OpenCL kernel
With contributions from RoBiK, tpruvot and sp_
Please consider a donation to:
ETH: 0xeb9310b185455f863f526dab3d245809f6854b4d
CUDA error in func 'getNumDevices' at line 112 : unknown error.
terminate called after throwing an instance of 'std::runtime_error'
what(): unknown error
Aborted (core dumped)
for opencl it says
No GPU device with sufficient memory was found. Can't GPU mine. Remove the -G argument
PS: downloading cuda toolkit from https://developer.nvidia.com/cuda-downloads
And still ethminer doesn't see opencl/cuda devices
refreshing experience why windows won the desktop
In windows it didnt submit shares.
Can you test in EC2 g2 linux instance please, does it submit shares?
6. run
cmake -DBUNDLE=cudaminer -G "Visual Studio 12 2013 Win64" ..
7. open ethereum.sln in build directory
well this is how far i got ... with cmake -DBUNDLE=cudaminer -G "Visual Studio 12 2013 Win64" , no ethereum.sln apears in build dir , some kind of errors in cmake logs wich unfortunately means nothing to me
i uploaded small zip with cmakeerror and cmakeoutput logs , in case anyone willing to look at http://expirebox.com/download/78133cd106b51407cf63afc3fc93867a.html
, im clueless whats wrong , as i said i know jack sh1t about coding and compiling stuff
Not mining ETH much anymore but still lurking and just tried out the latest release. Went from about 48-49 Mh/s to 54-55 Mh/s.
The other miner again was an old one from Sept of last year.
I have all drivers installed, ethminer is compiled, but --list-devices returns nothing.
nvidia-settings shows NV drivers up and running, clinfo shows NV opencl active... out of ideas, except starting from scratch
I used https://wiki.tiker.net/OpenCLHowTo#Installing_the_Nvidia_ICD
compiled the example, ./print-devices shows my 750Ti as valid opencl platform
running ./cl-demo 1000000 10 gives valid results, ie OTHER programs do recognize the device, its only ethminer which doesn't ?!
I think it was http://askubuntu.com/questions/541114/how-to-make-opencl-work-on-14-10-nvidia-331-89-drivers that helped - activating nvidia-352-uvm via modprobe
Ethminer --list-devices still doesnt list anything, but forcing via --opencl-device 0 or --cuda-devices 0 correctly finds the card and benchmarks around 8MH/s
real-world "mining" is ~ 5MH/s
@bluebox should be coinotron.com, sorry
Two 970's getting 40MH/s avg, up from 38 on 1.0.3, seems stable. I'll take 5% anytime, thanks! Don't have much eth, but will send a tip someday soon, might just buy some now that it's down.