DC Vault 2

Awesome job guys. Since you were doing so well I held down the fort on eOn.
 
Well that does bring the attention of deciding what May's focus project should be. Do we want to hit a large point per position, a small points but lots of positions, or a defensive push? Or does everyone just want to do their own thing for a month?

I haven't watched it for a while...but from the current scores it looks like we could tear through eOn pretty quick taking several positions for 14.67points each.
 
eOn sounds good to me.

After that we should throw some weight to Muon and crunch there for a bit. I believe with the firepower we have now we could gain a position in Muon. Folks just have to remember to run multiple instances of it since it doesn't scale well past about 6 cores.
 
I have brought up an issue in the Admins section in the Vault to possibly discuss for the upcoming changes. CPDN has had a stats export issue for a while and it appears that it isn't on their near future concerns list. Ever since they switched their team up and started making changes to the site it has been nothing but a headache IMO. Since our team isn't real active there for whatever reasons, I wanted to get some of the teams attitudes and opinions on the matter.

CPDN - ongoing credit export issues
http://climateapps2.oerc.ox.ac.uk/cpdnboinc/forum_thread.php?id=7735#48815

I suspect that the previous updates were run manually, as a result of mods asking the project people about credits.
It's still unlikely that anything much will happen with them on a permanent basis for some time, because we're up to our eyebrows trying to push out the new model.
And now with coping with replies from people who have been asked to update their computers off 7.2.39, which as everyone should know, having read the News item about it, causes file transfer failures.
 
Sent an Email to Muon1 Dpad Stephen Brooks asking about a linux compile of muon1 or the source code if doesn't want to do it himself :) , We will see if I get a response ...

Apr 25: nope, still windows only and closed source...
 
Last edited:
Last edited:
I was initially running muon with 24 instances with 1 thread each for max cpu usage... I have been playing around a bit to try and get higher fit models :)

Muon1 is currently processing 2 lattices, 1xc2_nosample and 5xc2...

You have to start the muon client once with network config to download the current lattice files...

I found that setting the muon ini to "no network" config and removing one lattice allow you to have that client/(directory) focus on one lattice, and then you can tune the number of cores you want to run or change how it trying to find the best model (mutate,random, etc)

the 5xc2 generally runs the longest and results in the better mpts overall I believe...
the 1xc2_nosample is more of a test good of a fit a individual client can progress up to without sample data from dpad.

I have been 2 instances of 1xc2 running with 2 threads with mostly random and mutated results and now have taken the best models from those and setup 1 instance running 6 thread trying more interpolated like tests instead, I left the 2nd instance running random/mutate/extreme with 6 threads now.
I was running 2 instances of 5xc2 with 24 threads each with standard fit types so it would use all the cpu remaining (24 cores total), I have change that to 8 threads each now.
so 6+6+8+8 threads on 24 cores = 94% cpu usage.
Best 1xc2_nosample on dpad is 2.85, for me is 1.04
Best 5xc2 on dpad is 4.1, mine is mostly returning 3.4 to 4.06

Biggest advantage of running muon directly instead of thru yoyo is you can play with your results and try to get better model fits. Thru yoyo you just get a couple run of what ever yoyo has decided to sent you.

If wanted I can post a zip file with the best 1xc2_nosample or 5xc2 results (with the whole client directory setup if you want)
 
Last edited:
For those waiting for 64bit ARM processors in phones and possibly tablets, it may be as soon as X-mas. However, nothing is said about 64bit Android yet. Shouldn't be too much longer though.

http://www.cnet.com/news/move-to-64-bit-android-phones-accelerating-says-arm/

I think once the Android OS is 64bit is when DC'ing will see their major advantage. They are still pretty good for the points they produce compared to the power they draw. One user at WCG did some numbers on his U3 device (not a phone) and those numbers look promising for a low power device.

https://secure.worldcommunitygrid.org/forums/wcg/viewthread_thread,35844_offset,20

Short update after a couple of day running the odroid u3 from emmc card: Total runtime of returned results for this device is now 26 Days and 9 hrs giving 52,965 points. Thus every hour is worth ~84 points.
While the odroid u3 running from microSD card returns 61 pts/h the u3 with emmc turns out to be 35% faster. Or compared to my pc:
[email protected] 183 pts/h
[email protected] 84 pts/h

That's really not bad.
 
Wouldn't have helped. Had to power off the electrical panel in the room to add circuits (no hot-add allowed). Good idea, though. :)

You should have just ran an extension chord from Grandpa's house. He probably wouldn't have noticed...lol
 
Mmm hmmm... I'm beginning to think you have either a cold fusion reactor, or your own hydroelectric dam...

Grandpa's back yard....

solar_farm1.jpg


And great job guys
We gained another GIMPS position.
 
Majestic-12 had a software update in case anyone was using/interested in it.
 
It's been a while since I did PRP but, I'll try and look tomorrow and see what the registration is like. EDIT found it!

Also just remember for PRP points are only granted if something is found using your computer. Doing the WUs does nothing if you don't find anything.

EDIT!
PRP-PSP
Joining the Team
To join you must go to the forum (http://www.mersenneforum.org/showthread.php?t=2692) and properly post a request in the forum thread and send a PM to ltd. Once you see your name on the list or the PM replied to then you are properly on the team. In the post/PM you must say your username and the team you would like to join.

Running the Project
In the forum there is a thread (http://www.mersenneforum.org/showthread.php?t=12239) which has the newest program for the project. At the link (http://tools.1up.no/psp/prpclient/) download and unzip the proper client. First you must update the master_prpclient.txt with your email address and your username. Enter the username provided on the forum that you posted on. For windows in the folder it give you many options of how many cores you have. Choose the correct amount of cores (I choose 8 cores) and run the install-prpclient once. Then run the start-prpclient. This will create the correct amount of command prompts running for the project. To send stats or update the client manually use the update-prpclient-ini.

And your up and running.


Though I'd add the other harder projects to join since I had written this a bit ago, still have more to write as well.

How to join difficult teams

Muon1
Joining the Team
There is no way to join the team before running the client.

Running the Project
To run the project you must download the zip folder from the main website. To register you must create a user.txt file inside the folder and on the first line of the file it must say [H]Username. For example mine is [H]EricE. This then registers you for the [H]orde once results are sent. Inside the folder is a file called muon1_cmdline. This is the version I use as it shows a command prompt of the progress that you can watch other ones such as muon1_background and muon1startup run behind the scenes. Also, you may use the manualsend to ship out the finished WUs if there is enough to send (it will error if there is not enough). Using the muon1_cmdline will ship out the stats automatically as well.

SoB
To Join the Project:
Go to their website www.seventeenorbust.com and click on the “Sign Up” link on the left side. Create a Username and enter your email address. Login to the site with your new username and select the preferences link to add your name to the [H]ard|OCP team.

To Run the Project:
Go to the link Download on the left hand side and get the proper version of the client software. Also download the Prime.txt file and put it in the directory for SoB. Extract the folder and Run “prime95.exe” Make sure you add your username to the Primenet under Test. Other than that just let the client run and your all set the WUs are pretty long though.


Wieferich@Home
To Join the Project:
Go to their website www.elmath.org. Download the correct version for your needs either regular or the multicore. Install then it will prompt you for registration. Add your member name (Eric E) then a password and your email. To join the team you must start the client and where it says “unknown” find [H]ard|OCP.

To Run the Project:
Find your link and run since this is a program. It will run in the background or you can make the windows pop up. To exit just close out.

And this necro post seems to confirm the same setup... But with outdated exe links...
 
Last edited:
Looks like another spot at GIMPS. Keep up the awesome work guys.
 
Ok...so here is where I feel our team has really failed in regards to giving good advice to new members or to encourage team support. Apparently, by running Moo! you can push our teams position for RC5-72. However, to do it, you not only have to sign up for an account at Moo! like other BOINC projects, but also go to http://stats.distributed.net/projects.php?project_id=8 and associate your name with the team like you would running their client. This is why many Vaulters in the past have had problems with adding Moo! to the DC-Vault. If it were added, it would be a true double dipping project because you would get the same credit added twice as far as Vault points. However, for those who have not taken the time to setup their RC5-72 accounts to reflect the Moo! contributions, please do so now.

To do it, make sure you have either ran some work with the RC5-72 client or through Moo! (BOINC).
Go here http://stats.distributed.net/projects.php?project_id=8 after the stats are updated and search for your email address.

On the stats page for your email address, click the button "Please email me my password", which gets you your password.

After you login, you can join teams (by searching for it and clicking join this team), change the displayed name for the email address, etc.

So, for those wanting to stick mostly to the BOINC client, you can help push our position using the Moo! wrapper. Moo! still uses CAL and therefore many older AMD cards will work fine there. That is one of the 3 projects I attach very old AMD cards to. This project also seems to favor AMD. However, I have heard it doesn't like newer drivers. I think version 11 was one of the last ones to not have some kind of issue, but I haven't really kept up. I have also ran very low end nVidia cards there. I think I had like a 128MB card that ran it just fine and maybe even an integrated one that had even less, but can't recall since it was a borg. I will add this info to the all inclusive list for future reference probably tomorrow some time.

Also, when I associated my email to the team, this is what it said (and is good for us)

Any submitted blocks that have not been allocated to a team
will be allocated to the team you have now selected during the next stats run.

Any submitted blocks that you have already allocated to a team
will not be affected. Only future blocks will go to the new team.

No changes will be visible until the next stats run.

You have until midnight UTC to change your mind without any problem.
Team joins are tracked on a day-by-day basis, so only the last teamjoin
on a given day will take effect. Once midnight rolls around though, your
team selection for that day turns into a pumpkin, err, becomes permanent.

Stats runs typically occur daily at 00:00 UTC.

So if any of our team that has previously ran Moo! and hasn't associated their points to the RC5-72 HardOCP team, need to do so.
 
Hmm, I think I found PSP setup instructions http://forums.overclockers.co.uk/showpost.php?p=22418577&postcount=88.

Looks like hanluc has been lurking here and not volunteering info on this :p

edit: make that jondi lurking here who is hanluc....

So is it me or do we just need 1M points to gain a position and 333 vault points? :eek:
hsmiwl.png


My SIMAP and Rosetta runs are not going to be as productive going forward. In the past, we were gaining multiple positions a day in SIMAP which equated to 66 points in the last month. Now it's more difficult to overtake folks. Same for Rosetta. We've gained 27 points in the last month, but it's going to take a while to overtake more teams. GIMPS, on the other hand, is still going well. I added another GTX 580 to the mix and I have two more to install. Windows was giving me fits last week (see other thread). Now I feel fairly comfortable using MISFIT on Windows-based GIMPS clients.
 
brilong, that is awesome production and glad someone has been tallying the numbers for some of the lesser scoring positions we have gained.

As far as PSP, that is probably accurate. The problem is that most people don't want to deal with the headache of setting up their account and requesting work. It is rather rediculous the way they do it in todays day and age of computing. Kinda my feelings on how to get work for GIMPS on the GPU's too.

RC5-72 would be one of those projects with small points/position that could have multiple spots gained, but that would be easiest to just run Moo! as instructed in my post above to get your Moo! ID to reflect the team association at RC5-72.

http://stats.distributed.net/team/tmsummary.php?project_id=8&team=24958
 
Edit: Should have be quoted from brilong PSP post ...

They have very long runtimes... looking at Grandpa's Completed Wu in Primegrid..
47 hours intel i7-2700K 3.6ghz, 116 hours amd 6276
There is a cuda app but it only run 2.5 faster than amd on my system (6234/660ti)
From looking at the old stats for our team, it looks like 1,500 points per wu completed...
Still waiting for my post to add me to the team to be approved. Started One WU with already registered teammates name... 25% completed...
 
Last edited:
And Damn, if I only knew that 25M points ago!

But Gilthanis is right - I've had a single 4850 on this project off and on for 3-4 years, and still get about 60K ppd on it. If not for Moo!, this card would likely have been tossed by now...

Ok...so here is where I feel our team has really failed in regards to giving good advice to new members or to encourage team support. Apparently, by running Moo! you can push our teams position for RC5-72. However, to do it, you not only have to sign up for an account at Moo! like other BOINC projects, but also go to http://stats.distributed.net/projects.php?project_id=8 and associate your name with the team like you would running their client. This is why many Vaulters in the past have had problems with adding Moo! to the DC-Vault. If it were added, it would be a true double dipping project because you would get the same credit added twice as far as Vault points. However, for those who have not taken the time to setup their RC5-72 accounts to reflect the Moo! contributions, please do so now.

To do it, make sure you have either ran some work with the RC5-72 client or through Moo! (BOINC).
Go here http://stats.distributed.net/projects.php?project_id=8 after the stats are updated and search for your email address.

On the stats page for your email address, click the button "Please email me my password", which gets you your password.

After you login, you can join teams (by searching for it and clicking join this team), change the displayed name for the email address, etc.

So, for those wanting to stick mostly to the BOINC client, you can help push our position using the Moo! wrapper. Moo! still uses CAL and therefore many older AMD cards will work fine there. That is one of the 3 projects I attach very old AMD cards to. This project also seems to favor AMD. However, I have heard it doesn't like newer drivers. I think version 11 was one of the last ones to not have some kind of issue, but I haven't really kept up. I have also ran very low end nVidia cards there. I think I had like a 128MB card that ran it just fine and maybe even an integrated one that had even less, but can't recall since it was a borg. I will add this info to the all inclusive list for future reference probably tomorrow some time.

Also, when I associated my email to the team, this is what it said (and is good for us)



So if any of our team that has previously ran Moo! and hasn't associated their points to the RC5-72 HardOCP team, need to do so.
 
I have a few borgs running 4350's and one with a 4550 not to mention the old laptop integrated GPU's that were even older than those that can do CAL...lol. I have now moved the priority level up to match everything else instead of at 0 for a backup project...
 
It took 24 hours but my Moo! contribution just showed up on RC5-72's stats for our team. My little contribution is currently worth 4% of the teams over all blocks total.
 
Okay, went and logged in, but don't see what to change to do this... I also think the machine I have running my 4850 also has the distributed.net command line client running in the background, which is how I've been currently adding points - so that may be doing weird things, too... But all I see on the dnet participant configuration page is this:
Participant: jathanis@**hoo.com
Team: 24958: HardOCP

Looking at my dnet.com RC5-72 page, I have 3.2M blocks toward our HardOCP team, but I have 25M Moo! points toward our [H]ard|OCP team in BOINC... Any thoughts, Gilthanis?
 
Did you use the same email address for both Moo! and RC5-72 accounts?
 
If you take your total points at Moo! and divide them by the number of blocks you have completed, they are just slightly less points/block than what I have done. I don't know which client is more efficient or if your command line is even doing much. You will have to trace your steps there. However, what I did was take my Moo! points and divide them by number of blocks and got ~8.5 points per block. I then looked at your Moo! points and divided them by your blocks and it was ~7.9 points per block. All I could suggest is maybe contacted the admin to see what is up. Are you running BOINC on the GPU at the same time as the command line client? If so, you may not be doing much with one of them or may be bogging the card down to lesser efficiency.
 
I'll try some testing this week by shutting down one client for a few days, then the other, and see what comes out of it...
 
They have very long runtimes... looking at Grandpa's Completed Wu in Primegrid..
47 hours intel i7-2700K 3.6ghz, 116 hours amd 6276
There is a cuda app but it only run 2.5 faster than amd on my system (6234/660ti)
From looking at the old stats for our team, it looks like 1,500 points per wu completed...
Still waiting for my post to add me to the team to be approved. Started One WU with already registered teammates name... 25% completed...

44% Completed...

Some notes on how to setup if anybody else wants to try ...
unzip program
edit the master_prpclient.ini file
set userid and machineid to your username (bug in prpclient)
uncomment or add servers
server=PSPfp:100:1:www.psp-project.de:8100
server=PSPdc:0:1:www.psp-project.de:8101
comment out any other servers.
comment/uncomment between 32 bit and 64bit exe's or cuda (I had to add cufft32_55.dll to directory for cuda to run)

run ?-install-prpclient.bat (creates X subdirectories)
run ?-update-prpclient-ini.bat (copies ini file to directories with instance id updated)
run ?-start-prpclient.bat (starts prpclient in each directory)

you can see what the prpclient reserved at tools.1up.no/psp/stats
The 1 which was my machine id before, hopefully on submit it will fix itself.
I started a 2nd instance with my username in machineid...
 
44% Completed...

Some notes on how to setup if anybody else wants to try ...
unzip program
edit the master_prpclient.ini file
set userid and machineid to your username (bug in prpclient)
uncomment or add servers
server=PSPfp:100:1:www.psp-project.de:8100
server=PSPdc:0:1:www.psp-project.de:8101
comment out any other servers.
comment/uncomment between 32 bit and 64bit exe's or cuda (I had to add cufft32_55.dll to directory for cuda to run)

run ?-install-prpclient.bat (creates X subdirectories)
run ?-update-prpclient-ini.bat (copies ini file to directories with instance id updated)
run ?-start-prpclient.bat (starts prpclient in each directory)

you can see what the prpclient reserved at tools.1up.no/psp/stats
The 1 which was my machine id before, hopefully on submit it will fix itself.
I started a 2nd instance with my username in machineid...

server=PSPfp:100:1:www.psp-project.de:8100
server=PSPdc:0:1:www.psp-project.de:8101

You can change the set up of these servers, each server gives out different work, the top one is fp (first pass) work, these take the longest to complete, around 3-4 days on my 3770K and currently give around 11,000 pts per WU. The bottom one is dc (double check) these take half the time of fp pass work, and give 5,500 pts per work unit, they scale the same, I've never found any advantage pts wise between the two but the dc work maybe preferable to some on slower machines.

100:1 means that you will recieve 100% of work from this server, 0:1 means no work, you can change them around, do both types of work evenly 50:1, 50:1 or whatever percentage you prefer.
 
Last edited:
Back
Top