|
cbuchner1
|
 |
January 06, 2014, 06:41:42 PM |
|
Assuming that in cudaMiner debug mode (-D) "hash <= target" entries are indicating stale blocks, do I need to worry about "hash > target (false positive)" entries? I'm having a lot of them in pool mining but obviously there is no wallet debug in that case to figure them out. cudaMiner: [2014-01-06 17:33:59] DEBUG: hash <= target
Wallet debug.log: 2014-01-06 17:33:59 proof-of-work found hash: 000000036f84eb900122146301c4e0a6941ad0da6b8947037b42723d76de49c7 target: 00000003b9e40000000000000000000000000000000000000000000000000000 (...) 2014-01-06 17:33:59 ERROR: generated block is stale
"hash <= target" is a good thing, meaning that your cudaminer completed some work. "hash > target" won't occur a lot with the current github version. It means the hash was close, but not quite acceptable. "ERROR: generated block is stale" is a bad thing meaning that the work that cudaminer has been doing was valid only for an outdated block, and the hash is unsuitable for signing the current block. Either because a new transaction was added to the current block, or because someone else found a block before you did. The longpoll support in cudaminer was designed to reduce the chance of this occuring. I think a short -s parameter in the range of 5s also reduces the chance of such a mishap.
|
|
|
|
|
|
|
|
|
Advertised sites are not endorsed by the Bitcoin Forum. They may be unsafe, untrustworthy, or illegal in your jurisdiction. Advertise here.
|
|
|
|
Ultimist
|
 |
January 06, 2014, 06:45:15 PM |
|
When I use the 12/10 or 12/18 version (Windows 7 x64, 64-bit miner) I get errors "does not validate on CPU" after a few minutes, no matter what settings I try, even auto config. I'm using a Galaxy GTX 670 with 4GB memory. I am using Nvidia inspector to clock to 1137mhz as this is the fastest my card seems to go and still be stable. Unsure about what to set for voltage, etc...
Anyone know what I'm doing wrong?
By the way, the 12/1 version works fine for me, but is slower than the later versions, at least according to the reported hashrates before the errors start.
|
BTC: 1GT4rqdzfRbKW3ZqdH1g5nonG1oyGEsift -- QRK: QZuCnS6aKmk276aFKJCoW46vDGr85aBQr1
|
|
|
bathrobehero
Legendary
Offline
Activity: 980
|
 |
January 06, 2014, 06:52:11 PM |
|
With the current github version in scrypt-jane, I do have a lot of random "hash > target (false positive)" entries. I'm not sure if playing around with -C -H and kernel (K13x1) had some effect on them or if it's just random. Also, I have no idea what the "candidate nounce" line is.   Using -l K13x1 -i 0 -H 1 -C 0 -m 0 -D mostly with GTX 660 clock to 1137mhz as this is the fastest my card seems to go and still be stable.
I've been there, and I believe you went to far, it might be stable (as in not crashing), but it starts to do mistakes.
|
|
|
|
|
|
cbuchner1
|
 |
January 06, 2014, 07:01:17 PM |
|
With the current github version in scrypt-jane, I do have a lot of random "hash > target (false positive)" entries. I'm not sure if playing around with -C -H and kernel (K13x1) had some effect on them or if it's just random. Also, I have no idea what the "candidate nounce" line is.   n there, and I believe you went to far, it might be stable (as in not crashing), but it starts to do mistakes. definitely not the current github version, because I've already removed the candidate nonce print statements...
|
|
|
|
|
|
bigjme
|
 |
January 06, 2014, 07:08:01 PM |
|
cbuchner1 can you please post a link to the current github for windows? i really want to try it but i can not compile. im a willing tester and my 780 needs to be tested on this i just set up a server config using this template off the yacoin admin, hopefully it is right. it does launch, but looks the same as normal. is it suppose to or is it being a pain? rpcuser=<username> *your username here rpcpassword=<pass> *set you password here rpcallowip=127.0.0.1 *this to mine from your computer rpcallowip=192.168.1.* *Put your local address here, is important that you use * at the end, of the ip address to allow inncoming connection from the whole ip range of your network. rpcport=<anyport> *just type here i use 3339 port=7688 daemon=1 server=1 addnode=82.211.30.212 addnode=78.21.9.49 addnode=76.115.8.101 addnode=124.149.56.205 addnode=178.130.36.81 addnode=82.6.77.126 addnode=106.187.55.212 addnode=81.202.104.33 addnode=84.200.17.178 addnode=88.204.169.242
|
Owner of: cudamining.co.uk
|
|
|
bathrobehero
Legendary
Offline
Activity: 980
|
 |
January 06, 2014, 07:36:46 PM |
|
server=1 is there so all you need to do is give it a port and connect through that port (-o http://localhost:<anyport>) If it starts hashing, you're good to go afaik.
|
|
|
|
|
|
bigjme
|
 |
January 06, 2014, 07:39:08 PM |
|
I would test it but I dont have a compiled cudaminer for it :-p
Im waiting for someone to figure out how to compile it and send it my way. i have everything running but the miner now. im leaving the server running to make sure it doesnt crash again. it hasnt crashed since i deleted the entire yacoin folder in appdata and downloaded all the blockchain again
i'm going to give Cygwin a go for compiling it, see if that works
|
Owner of: cudamining.co.uk
|
|
|
|
Ultimist
|
 |
January 06, 2014, 10:58:25 PM |
|
So when will the next new compiled version be added to the original post for those of us who can't or don't know how to compile?
|
BTC: 1GT4rqdzfRbKW3ZqdH1g5nonG1oyGEsift -- QRK: QZuCnS6aKmk276aFKJCoW46vDGr85aBQr1
|
|
|
|
cbuchner1
|
 |
January 06, 2014, 11:45:24 PM |
|
So when will the next new compiled version be added to the original post for those of us who can't or don't know how to compile?
It's not done yet. The answer is "when it's done". 
|
|
|
|
|
|
Ultimist
|
 |
January 07, 2014, 12:15:36 AM |
|
It's not done yet. The answer is "when it's done".  That's cool. No disrespect or 'hurry up' intended. 
|
BTC: 1GT4rqdzfRbKW3ZqdH1g5nonG1oyGEsift -- QRK: QZuCnS6aKmk276aFKJCoW46vDGr85aBQr1
|
|
|
bathrobehero
Legendary
Offline
Activity: 980
|
 |
January 07, 2014, 12:24:47 AM |
|
1>LINK : warning LNK4075: ignoring '/INCREMENTAL' due to '/LTCG' specification 1>cpu-miner.obj : error LNK2001: unresolved external symbol curl_easy_cleanup 1>cpu-miner.obj : error LNK2001: unresolved external symbol curl_global_init 1>cpu-miner.obj : error LNK2001: unresolved external symbol curl_version 1>cpu-miner.obj : error LNK2001: unresolved external symbol curl_easy_init 1>util.obj : error LNK2001: unresolved external symbol curl_slist_append 1>util.obj : error LNK2001: unresolved external symbol curl_easy_setopt 1>util.obj : error LNK2001: unresolved external symbol curl_easy_perform 1>util.obj : error LNK2001: unresolved external symbol curl_slist_free_all 1>util.obj : error LNK2001: unresolved external symbol curl_easy_reset 1>D:\Downloads\_coin\[_preregs\cccccccc newest\x64\Release\cudaminer.exe : fatal error LNK1120: 9 unresolved externals Trying to compile x64 release.
|
|
|
|
|
|
bigjme
|
 |
January 07, 2014, 12:37:18 AM |
|
What files are you using for that? I loaded up the github and just downloaded it all
|
Owner of: cudamining.co.uk
|
|
|
|
cbuchner1
|
 |
January 07, 2014, 12:37:59 AM |
|
Trying to compile x64 release.
use these compiled installation prerequisites with Visual C++ 2010 SP1. cudaminer_vc2010_prerequisites.7z (49.3 MB)this should be decompressed using 7-zip right next to the folder "CudaMiner" (or whatever you named your checkout directory). Christian
|
|
|
|
|
|
cbuchner1
|
 |
January 07, 2014, 01:20:09 AM |
|
I've added back interactive mode into the Kepler scrypt-jane kernel. It costs some performance, but the computer remains usable as if nothing was running in the background. I do this by splitting the large N=32768 scrypt computation into small chunks of 1024 iterations each and adding a tiny sleep time of 100us between each chunk. Next I will modify Titan and Fermi kernels accordingly.
The Fermi architecture currently sucks for scrypt-jane. Not sure why. It's performing admirably for regular scrypt-hashing.
Christian
|
|
|
|
|
|
relm9
|
 |
January 07, 2014, 02:10:32 AM |
|
Hmm, I've just tried mining a newer scrypt-jane coin (QQCoin) and I'm only getting around 7 kH/s on my 780. That's really low, my 7970s get around 1.8MH/s currently.
I did change the nTimestamp value to the proper one for this coin in the source, is there anything else I need to change?
Edit: switched to 64-bit build instead, now getting 53kH, but that's still way lower than AMD.
|
|
|
|
|
|
bigjme
|
 |
January 07, 2014, 02:12:15 AM |
|
Relm9, where did you get your cudaminer for scrypt-jane from?
7 is very high
|
Owner of: cudamining.co.uk
|
|
|
|
relm9
|
 |
January 07, 2014, 02:13:45 AM |
|
Relm9, where did you get your cudaminer for scrypt-jane from?
7 is very high
For yacoin it is, but the N factor for QQcoin is still low so hashes should be much higher I compiled it myself from the github repo.
|
|
|
|
|
|
bigjme
|
 |
January 07, 2014, 02:18:22 AM |
|
Is there any chance you could pm me with a link to it if you can upload it? Or even email me it?
I really want to start mining it and I can not for the life of me get it to compile
|
Owner of: cudamining.co.uk
|
|
|
|
relm9
|
 |
January 07, 2014, 02:26:33 AM |
|
Is there any chance you could pm me with a link to it if you can upload it? Or even email me it?
I really want to start mining it and I can not for the life of me get it to compile
I'm not sure if the dev wants unofficial binaries posted publicly, so I've PM'ed you.
|
|
|
|
|
bathrobehero
Legendary
Offline
Activity: 980
|
 |
January 07, 2014, 02:36:00 AM |
|
use these compiled installation prerequisites with Visual C++ 2010 SP1. cudaminer_vc2010_prerequisites.7z (49.3 MB)this should be decompressed using 7-zip right next to the folder "CudaMiner" (or whatever you named your checkout directory). Christian I've spent a few hours gathering and trying different versions and when I was about to give up, a man came down from heaven and gave me everything I needed! 
|
|
|
|
|
|