senseless, I got the original directions and your new download to run however getting a connection error, tried two different pools "Could not retrieve work from RPC server. CURL return value = 7 "CORRECTION: I think i had standalone server syntax mixed up with other using: rpcminer-cuda.exe -url http://pool.bitclockers.com:8332 -user userhere -password passhere I now get: "Could not retrieve work from RPC server. CURL return value = 3 for example trying to connect to BTC GuildCheck the port, url and user/pass. It works fine when connecting to bitclockers.
|
|
|
Everything is terrible with GTX 680 and newest NVIDIA drivers
Ufasoft is only miner that doesn't crash display driver or crash itself around 120MHash overclocked, uses about 100watts additional power from idle
one of the miner's showed 127Mhash, forget which one maybe Phoenix, but all these crash after closing:
GUIMINER with phoenix GUIMINER with OpenCL GUIMINER with CUDA (rpcminer)
Phoenix 1.7.5 program freezes after a few seconds Phoenix 2 - 123Mhash, freezes on close Diablo Miner - 121Mhash with 98% GPU usage. throws a bunch of Invalid solution errors and says hardware error CGMiner program just crashes before hashing (don't think supports CUDA)
Cgminer suports CUDA: [root@localhost cgminer-2.6.1]# ./cgminer -n [2012-09-15 06:17:35] CL Platform 0 vendor: NVIDIA Corporation [2012-09-15 06:17:35] CL Platform 0 name: NVIDIA CUDA [2012-09-15 06:17:35] CL Platform 0 version: OpenCL 1.1 CUDA 5.0.1 [2012-09-15 06:17:35] Platform 0 devices: 1 [2012-09-15 06:17:35] 0 GeForce GTX 470 [2012-09-15 06:17:35] 1 GPU devices max detected And is the best miner for it (IMO): cgminer version 2.6.1 - Started: [2012-09-15 06:19:06] -------------------------------------------------------------------------------- (5s):103.1 (avg):103.1 Mh/s | Q:5 A:3 R:0 HW:0 E:60% U:2.2/m TQ: 3 ST: 3 SS: 0 DW: 0 NB: 1 LW: 206 GF: 0 RF: 0 Connected to http://pool.coinlab.com:8332 with LP as user protected_mofo Block: 0000042cae5e4ef6ad08463f1ed49ea7... Started: [06:19:06] -------------------------------------------------------------------------------- [P]ool management [G]PU management [S]ettings [D]isplay options [Q]uit GPU 0: | 103.1/103.1Mh/s | A:3 R:0 HW:0 U:2.24/m I: 4 --------------------------------------------------------------------------------
[2012-09-15 06:19:06] Started cgminer 2.6.1 [2012-09-15 06:19:06] Probing for an alive pool [2012-09-15 06:19:06] Long-polling activated for http://pool.coinlab.com:8332/listenChannel [2012-09-15 06:19:06] Pool 0 http://pool.coinlab.com:8332 alive [2012-09-15 06:19:36] Accepted 60cc9cf3.9151d8f4 GPU 0 [2012-09-15 06:19:52] Accepted 0d651d91.13f970de GPU 0 [2012-09-15 06:20:22] Accepted 982b0d2f.a7862ba9 GPU 0 With overclocking I can go as higher as 150Mh/s. That being said, there is no point in mining with an Nvidia card, except maybe if you have free electricity or plan to mine at CoinLab. Check your CPU usage. It doesn't support cuda; it supports opencl which nvidia also supports.
|
|
|
Now do i need a special RPC account or connects to any pool?
Nope, just a normal account. You should just need to set -url -user -pass to get it to run. You may or may not want to change other -gpu and -aggression related options. rpcminer-cuda.exe -url http://pool.bitclockers.com:8332 -user userhere -password passhere btw, was able to find a 680 to test on this AM. so i know it's all working on 6xx devices now. I will test shortly what do you mean AM? at the moment? AM = morning, as opposed to PM
|
|
|
Now do i need a special RPC account or connects to any pool?
Nope, just a normal account. You should just need to set -url= -user= -pass= to get it to run. You may or may not want to change other -gpu and -aggression related options. rpcminer-cuda.exe -url=http://pool.bitclockers.com:8332 -user=userhere -password=passhere btw, was able to find a 680 to test on this AM. so i know it's all working on 6xx devices now.
|
|
|
how much for you to tell me how to run it under windows? Download and run the exe. I didn't include all of the files. Just what is updated. You will need to download rpcminer-win32 first for all the dlls and crap. 1) download this: http://www.mediafire.com/file/hduqoqdf3onqrod/bitcoin-rpcminer-20110227-win32bin.zip2) unzip #1 3) Download my update: http://www.qfpost.com/file/d?g=oreaN2eXE4) unzip #3 5) move files from #3 overtop of files from #1. I guess maybe I should have just put the dlls and crap in my zip.... ... ALso, make sure to let me know if it works or not. I wasn't able to test it with a 6xx card... I'm curious to know..
|
|
|
Was able to recompile the code on windows for support for revision 1.2, 1.3 and 3.0 devices (2xx and 6xx cards). (Note: It uses cubin 2.0 for 2.1 devices.. I expected a different cubin would be necessary but it looks like they're similar enough that it's not required; You can't even compile a 2.1 specific cubin with nvcc as far as I know.) Contained in zip is rpcminer-cuda.exe along with the necessary cubin files. I only changed the cubin file detection code in bitcoinminercuda.cpp and compiled the 3 new cubin files under their respective archs. if(major==3) { cubin="bitcoinminercuda_30.cubin"; } else if (major==2) { cubin="bitcoinminercuda_20.cubin"; } else if (major==1 && minor==0) { cubin="bitcoinminercuda_10.cubin"; } else if (major==1 && minor==1) { cubin="bitcoinminercuda_11.cubin"; } else if (major==1 && minor==2) { cubin="bitcoinminercuda_12.cubin"; } else if (major==1 && minor==3) { cubin="bitcoinminercuda_13.cubin"; }
Download for Win32: http://www.qfpost.com/file/d?g=mwrLcmBsLInstallation: Unzip and use rpcminer-cuda.exe with the default command params from post #1 of this thread. ... I don't normally do this; but it took me half a day to get the compile environment working for rpcminer. If you'd like to toss me a tip: 136HdE3wxyfxFsuAHYZTLTHDopXYxRJej1
|
|
|
@senseless can you please upload the project somewhere and post the link? In VS2012 it was pretty. Simple you just need to make sure you have the right version of amd app installed.. Then to fix the linker problems for the cuda stuff you need to add an existing file; in this case... gpu-cuda.cpp file to the project. Then just place it with the other gpu- files. Should compile fine.
|
|
|
I've tried compiling it on windows. I downloaded all the libraries and tried using cmake but can't get very far. I am unfamilar with cmake and a few other things. I doubt I would have much trouble compiling this in linux. It's a little beyond me and that's why I was hoping someone else got it to work.
I understand, if You have any output from compilation please paste me here because the aim of this mod is to provide easy to compile version of software. I'm thinking also on providing binary version (for windows too) and to remove OpenCL support, because there are plenty of OpenCL miners but no good (except rpcminer) CUDA miners that are giving large benefits for nvidia users. Hi Ang3lus, I like your miner, and i like that you have not forgotten the CUDA owners like everyone else have. So i will help you make it work on windows, i will use my winxp virtualbox today and work on it. If it work i'll boot my win7 and test it to confirm it work. After, you'll be able to provide an already compiled version to the windows users. I'll say you all i've done to make it work, so you'll be able to make it on the real file. Regards, Mike Any luck?
|
|
|
Switched to AMD APP 2.3 and everything compiling fine on opencl now.
the gpu-cuda.cpp file is not added by default under the "GPU Support" (for VS) with current source tree. If anyone is getting linking errors that they can't figure out where they're coming from. Make sure that gpu-cuda.cpp is included in your tree before building.
|
|
|
OP, I feel your pain. I have a number of nvidia machines that we would like to mine on when idle. But can't because the opencl gpu bugs make the machines more or less unusable. I'm willing to offer a reward for someone to code a cuda kernel for bfgminer https://github.com/luke-jr/bfgminer/issues/132I'd be willing to put up 10BTC reward for it. rpcminer-cuda already has working cuda miner; but I have no idea how hard it would be to add the cubin files from rpcminer-cuda to bfgminer. Edit: Found this in my google crawl. The puddinpop rpcminer-cuda works pefectly. No cpu usage bug; but it lacks many features. I'm not sure how hard it would be for someone to add the cuda kernel to phoenix, cgminer or bfgminer? http://luke.dashjr.org/programs/bitcoin/w/puddinpop-bitcoin-pool.git/tree/HEAD:/src/cuda
|
|
|
+5 Bitcoins for a Cuda Kernel in Phoenix...
That cpu bug in opencl is driving me nuts. The rpcminer-cuda does not have the same problem.
|
|
|
Any ideas what may be causing this? Do I need a certain version of amd app sdk to compile? 7>el\comp\ext-opencl.cpp(111): error C2039: 'clEnqueueMarker' : is not a member of '`global namespace'' 7>el\comp\ext-opencl.cpp(111): error C3861: 'clEnqueueMarker': identifier not found 7>el\comp\ext-opencl.cpp(116): error C2039: 'clEnqueueBarrier' : is not a member of '`global namespace'' 7>el\comp\ext-opencl.cpp(116): error C3861: 'clEnqueueBarrier': identifier not found
|
|
|
Why on earth would AMD put an inefficient CPU usage bug based on openCL global worksizes in there on purpose for BFL to release ASICs?
I'm saying that, if a card was pushing 7Ghash/s by itself, it would probably need 30-100% cpu to keep the work coming in with no delay, or, it's a bug either in the engine or drivers. But, I didn't see anything in the change log (or im blind) stating that anything above 9 intensity would consume 100% of my cores as a future proof for ASICs. So, I'm guessing it's a bug. The weird thing is, if i switch to an older cgminer (2.6.1 for instance) the problem does not occur at intensity's higher than 9. So, I'm not sure what to make of it.
|
|
|
Intensity of 9. Intensity 11 does peg a core, is that a bug though or part of what intensity means?
I'm thinking it's a bug. I would think there would be a more gradual step up. I use to run at 14 with older versions of the miner. When i run at 10 now it will consume 30-40% for a single card (rest disabled); which is a pretty huge jump from 0% usage at 9. Unless they did this intentionally for the ASICs that are coming.
|
|
|
I have no CPU issue with 12.8 and 7000 series.
What intensity are you using?
|
|
|
6xxx and 7xxx cards don't work with old drivers.
How are people dealing with the 100% cpu issue? Seems pretty damaging. Using intensity 9 as others have suggested, my rates have fallen significantly, though cpu usage does drop to 0. Others have suggested limiting the miner to a single core (affinity), which also works, but doesn't seem like a suitable fix for the problem.
|
|
|
I've been looking for some info about what happened to the drivers and why everything is using 100% cpu. I've not yet been able to find anything relevant in google searches. Only that I should install older versions of the drivers.
I've tried to install the 12.1 drivers with the 2.7 amd sdk, but now the miners can't find the cards at all. Seeing elsewhere on this site that people are using 12.1 with 2.5-2.7 just fine. So why can't my miner find the cards?
I've done a complete uninstall, reboot, reinstall, reboot, install sdk, reboot.
|
|
|
I found the issue, was a bug in my miner. Moved to an older version, problem went away. Doh.
All of the latest miners use 100% cpu, seems like some sort of bug in the engine or mining code. If I swap to slightly older versions the usage drops to < 2%. Any idea what might be going on there? I'm running the latest drivers as of today.
For aggression/intensity, I"m using 14 under BFGMiner, and 12 aggression on phoenix.
|
|
|
|