Page 1 of 1
Need help with 80x 7970 cluster
Posted: Mon Aug 05, 2013 1:05 am
by CygnusXI
Does the folding system have a failsafe to stop from giving points if a new user gets points to fast? ie multiple 7970's on core 17
im trying to help him start folding, but all his wu's are getting not credit, it looks like he got credit for only 1
any help would be greatly appreciated.
If we get him up and running, this will be a good one to watch, 80x 7970 on core 17 ....

Re: Need help with 80x 7970 cluster
Posted: Mon Aug 05, 2013 2:02 am
by Joe_H
There is no failsafe for getting points too fast. There are some conditions under which returns will not be accepted, you would have to post logs showing what is going on before we would have an idea why points are not being credited.
This post gives info on how to post logs.
Additional useful information besides the beginning of the log showing FAH version, System information and the configuration would be the version of drivers being used and whether they have been overclocked. That includes factory overclocks.
Re: Need help with 80x 7970 cluster
Posted: Mon Aug 05, 2013 2:12 am
by Zagen30
Did this friend by chance set up FAH on one machine and then clone that drive to all the other ones? If so, that may be the problem. If it was already doing a WU, that one would be copied to all the others. Once the original one was finished, the servers would reject all of the copies since it had already received a successful copy and wasn't expecting any more. There may also be issues with the machine IDs being identical, but I can't remember if that's still an issue with v7.
Re: Need help with 80x 7970 cluster
Posted: Mon Aug 05, 2013 2:26 am
by Joe_H
Zagen30 wrote:There may also be issues with the machine IDs being identical, but I can't remember if that's still an issue with v7.
That is still an issue with V7, ID's are assigned the first time a client connects to the assignment servers. If cloned afterwards, the installs will have identical ID's and that will cause a number of problems.
Re: Need help with 80x 7970 cluster
Posted: Mon Aug 05, 2013 3:40 am
by CygnusXI
this user is a bitcoin miner, that wants to try folding. he is not a coder. he is not cloning wu's. he has one of the most powerful bitcoin rigs there is.
lookup folding extreme stats on team curecoin to see the power of a few bitcoin miners starting up the alpha stage of our folding project. so far only small mining rigs have been used on this project
by rough calculation if we get him running he should be able to get 4 to 5 million ppd by himself, provided he runs core17 for ati performance.
Im runnning the curecoin project, i have a member of my team trying to get this user up and going. Ill ask them to get logs together and direct them to this post.
thanks guys... if you look up my project, youll see that there is a loyal following of bitcoin miners starting to use their power to fold. I plan on increasing this amount to see what kind of folding power bitcoin miners really have. Me and a friend have hosted scrypt ( not the basic sha256 )based mining pools that hit 300 mhash peak ( thats 300 ghash on sha 256 like bitcoin) , which equals about 30 million to 20 million ppd in folding power. A lot of these users are willing to fold with this power. Our most accurate estimate so far for increased folding power for the official launch of the project is between 25 million to 100 million ppd.
Curecoin devs need members to help with all aspects of development. please contact me if you have something to say about the project.
Re: Need help with 80x 7970 cluster
Posted: Mon Aug 05, 2013 3:56 am
by Joe_H
The issue is not just cloning WU's, but cloning the install of F@H. If that cloning is done after the client has contacted the folding servers, then the clones will have identical internal ID numbers and will result in duplicate assignments and other problems. Additional issues can come up in systems with more than one GPU, especially when the cards differ.
Re: Need help with 80x 7970 cluster
Posted: Mon Aug 05, 2013 4:15 am
by 7im
Let's just start by getting one system's config details and work our way up from there.
Re: Need help with 80x 7970 cluster
Posted: Mon Aug 05, 2013 9:25 am
by Napoleon
Joe_H wrote:There is no failsafe for getting points too fast.
I'd say (maintaining) QRB qualification is something of a failsafe (see
http://folding.stanford.edu/home/faq/faq-points/#ntoc7 for details), but it merely discourages gaming the points system and is an incentive for the donors to make sure that their setups are FAH stable and stay that way. Anyway, when
HP Cloud Services decided to do a burn-in with FAH, they certainly racked up points
fast.

Re: Need help with 80x 7970 cluster
Posted: Tue Aug 06, 2013 4:43 am
by CygnusXI
The owner of the cluster as you imagine, keeps his cluster very busy. We are hoping he will make another test run soon. Ill let you know if he fires the cluster up with fah, so we can see what kind of folding power he has. Im still waiting on log files from him.
thanks for the quick answers
Please take a moment to google my team name and see what you find, if you don't know what we are about yet. i think im not allowed to post a url yet on this forum, this is for research though so let this one slide, the real website that just got launched is curecoin dot us .
I could use the help of the experienced people here. The site is brand new not much content, but Im going to assemble my dev team there. More info will be there soon.
In the mean time, the curecoin team produced 472,050 points in the last 24 hours. Friendly competition on the way
