TennesseeTony

Elite Member
Aug 2, 2003
4,209
3,634
136
www.google.com
http://www.dc-vault.com/index.php

http://www.dc-vault.com/showteam.php?team=9 We are team 9....we are ancient! :D And we are currently in 15th place, ahead of some HUGE names like Gridcoin and SETI.USA.

I am still trying to wrap my head around this one, but with my current level of knowledge, this seems to be vaguely similar to Formula BOINC, in that what you run (EVERYTHING you run, almost, including Folding and other non-BOINC projects) gets counted for an extended period of time. But that second link provides some awesome data on what we are up to as a team, really digging that!

[H] Coleslaw is very active with the project, and I asked him to explain the scoring, for one. He provided the following information via PM:

Scoring is always confusing for new members. Each project has a maximum of 10,000 points value. However, that doesn't mean 10,000 divided up. Rather each position is based on a formula.

The scoring calculation to determine the team positions within the vault ranking system works as follows:

Overall score = Sum of all category scores
Category scores = Sum of all projects scores in a category
Project score = 10000 - ((Position - 1) * (10000 / Total number of teams in the project))

Now what really throws people from time to time is that the DC-Vault requires teams to sign up. Doesn't really matter whom does it as long as someone does. Now, DC-Vault is more than just BOINC. So, some teams don't participate outside of that. Some only have a few very small contributors outside of it. So, though there are some larger teams out there, they don't always care about the ongoing challenge and position. We just aren't a concern for them. So, if you look through those teams, you can see that they may not even have points for some of the projects. We really haven't added non-BOINC projects for a very long time as there just aren't that many out there that meet requirements. Also, projects like SETI will be worth less points per position change because there are more teams participating. Reference the project score equation above. If there are less teams supporting the project, each position change is worth more points. This is why being 1st place at FAH is worth 10,000 and being 3rd is worth 9,999.791341 . However at SRBase 1st is worth 10,000 and 2nd is worth 9,929.08. So, when calculating where you need to put resources, you also have to calculate how many points those resources are worth at each project. Currently, it would be best for H to focus on SRBase and TN-Grid because a million BOINC points would gain us a lot more ground then at WCG.

You will also find that some teams are actually much larger at non-BOINC projects. Dutch Power Cows have been very distributed.net heavy over the years. When they decide to organize and do their annual stampede, they can be a force to reckon with. However, they just focus on a few things for a small period of the year. They tend to be one of the few teams that still really watch the scoreboard.

Your team for example does not have any points for GIMPS. Nobody has registered your team if you have one. Merely creating a team, turning in 1 valid work unit, and letting the stats update would give you guys some easy points there. http://www.dc-vault.com/showteam.php?team=9 By have zero points at the project guarantees 0 points at DC-Vault.The current last place team for GIMPS at DC-Vault has 1,450.00 points. So, you guys could bump your score by about that many points by merely making a team and turning in a single work unit..... https://www.mersenne.org/

The last important thing to watch out for is loss of interest due to stagnation. One of the largest strengths of DC-Vault is also its biggest down fall. Since it is an ongoing challenge that does not refresh, it is fun at the beginning as you chew threw the other teams. But once you get higher up, there aren't as many easy fruit to pick and you may spend an entire year just bumping up one spot. This loses a lot of the excitement. So be prepared for that. It is always more fun dethroning the champ than it is maintaining that position....
 

[H]Coleslaw

Member
Apr 15, 2014
157
133
116
Until I get time to pull photos off of Photobucket the guide isn't as nice as it was. However, the pictures are there if you want to follow their holders. https://hardforum.com/threads/gimps-installation-guide.1813722/ That would cover CPU crunching. They have a few other clients for GPU's, but I myself had never ran them. But we did briefly mention the GPU applications in the 2 following posts on it.

I did not see your team in the rankings when I last looked. But that doesn't mean I didn't overlook it if it is there.
 
  • Like
Reactions: TennesseeTony

[H]Coleslaw

Member
Apr 15, 2014
157
133
116
And in case anyone was wondering.... yes PrimeGrid does some of the same work (or science) as GIMPS. That is why there are less and less users there. Just as Moo! does work for RC5-72 and YoYo does work for OGR-28. However, the only project just mentioned that does not keep their stats entirely separate for the sake of DC-Vault was Moo! and is the reason why Moo! isn't officially in it. If you want your Moo! work to count for positioning in the DC-Vault, you need to follow these steps: https://hardforum.com/threads/dc-vault-2.1620596/page-49#post-1040794811 It has been a long time since I did mine but others have had hit and miss luck with the process since.
 

Ken g6

Programming Moderator, Elite Member
Moderator
Dec 11, 1999
16,242
3,830
75
It looks like I wound up with a couple of double-check work units. Looks like three days or so on 4 cores. (It won't let me do 6 HT cores! :()
 

TennesseeTony

Elite Member
Aug 2, 2003
4,209
3,634
136
www.google.com
Yeah I'm running one too, Ken. 44% after 22 hours or so, I think mine is using 6 threads...I put 10 in the setup box but it went down to 6, and the CPU usage is....more like 5 and a half. :)
 
  • Like
Reactions: Ken g6

petrusbroder

Elite Member
Nov 28, 2004
13,343
1,138
126
Just checked the DC-vault ... and there is a problem: In World Community Grid the wrong team name is used.
As some of you remember, the old name was Anandtech. We (Team AnandTech Committee = TAC) wanted to change the name to "Team Anandtech", but we could not get hold of the "old" team captain (in spite of many tries during three or four months) so we founded a new team "Team Anandtech". Since Team Anandtech is much more successfull then Anandtech I think it is important that the current and correct team name is used by DC-Vault ...
Can some one contact the admin of DC-vault and help me to get in touch? I have tried the "edit team projects"-link but it "does not take" since there is somewhere a "captcha" which I can not see .... so if someone wants to try ... please feel free.
 

[H]Coleslaw

Member
Apr 15, 2014
157
133
116
Just checked the DC-vault ... and there is a problem: In World Community Grid the wrong team name is used.
As some of you remember, the old name was Anandtech. We (Team AnandTech Committee = TAC) wanted to change the name to "Team Anandtech", but we could not get hold of the "old" team captain (in spite of many tries during three or four months) so we founded a new team "Team Anandtech". Since Team Anandtech is much more successfull then Anandtech I think it is important that the current and correct team name is used by DC-Vault ...
Can some one contact the admin of DC-vault and help me to get in touch? I have tried the "edit team projects"-link but it "does not take" since there is somewhere a "captcha" which I can not see .... so if someone wants to try ... please feel free.

I had/have that issue with our team's name too. I went as far as contacting WCG to try and intervene. However, they said if they got a reply from the other team captain then they would do nothing. However, since he was inactive at the time if he did not disagree and/or did not reply they might help me out. Unfortunately, he started crunching minimally again and declined it preventing us from unifying the name choice we wanted. So, we continue to have one BOINC project different than the rest. We have no idea why he continues to squat on the name as nobody else joins him intentionally. My understanding is that he had a fallout with the team back before I joined. Some people just like to hold grudges I guess.
 

petrusbroder

Elite Member
Nov 28, 2004
13,343
1,138
126
Done, and now correct:

oqw4l2.jpg


And up 357 positions in WCG and 105 points in the score, and up 1 position in the misc. score in the total.
Better, somewhat better. ;)
 

StefanR5R

Elite Member
Dec 10, 2016
5,498
7,786
136
Last time several of us ran it, which was at the September 2017 sprint of Formula Boinc and a few times after, the "fu" in "Yafu" described how the Credit New system worked for Yafu. (The gflops estimation for a given host never converged, hence credit/work kept drifting randomly, even over much longer periods than that sprint.)
 

[H]Coleslaw

Member
Apr 15, 2014
157
133
116
The points issue should be a complaint to the BOINC devs to fix how credit New operates. YoYo uses the default credit system provided as he tests the server software. And yes I have seen it and had done extensive testing a few years back showing it. However, again that is for the devs to fix if they really cared all that much... I think at this point, we all pretty well know that Credit New is garbage.
 

StefanR5R

Elite Member
Dec 10, 2016
5,498
7,786
136
@crashtech, sites which pull stats from LHC@Home probably receive stats summarily over all subprojects nowadays.

@[H]Coleslaw, IMO:
  • Projects which distribute work units of deterministic computational effort should of course give deterministic credits per WU.
  • Projects with non-determinsitic WUs need to look for an algorithm which estimates credits somewhat fairly, based on the particular characteristics of the specific WUs. Credit New could be such an algorithm, under circumstances.
But AFAIU, Credit New is designed upon the assumption that FLOPS stay reasonably constant during the duration of each task, regardless if the WU turns out short or long. (This requirement for some uniformity is only needed per sub-project/ per application version, not globally for the project.)

And this is precisely what YAFU lacks. In YAFU, there is not only no predictability of FLOP/WU, it also has widely different FLOPS during distinct sections of a task. Each task is a sequence of sub-applications, each sub-application having rather different performance profiles. (E.g. multi-threaded vs. single-threaded portions; perhaps FPU intensive vs. memory intensive vs. I/O intensive portions...) The catch is: A task may terminate early, as soon as one of those sub-applications got a conclusive result. I.e. not all of the sub-applications may have been executed. Whether or not such an early return will happen is not known beforehand.

This is apparently a scenario which is not covered by the current design of Credit New. What the crediting facility at YAFU needs to do is
either
  • to examine for each returned result which sub-applications were performed for which duration during the task, and compute FLOP proportionally. This requires that the returned result includes a record of the runtimes of its parts, and that YAFU maintains FLOPS estimations separately for each of the sub-applications. If mainline boinc-server code doesn't have the necessary hooks for that, but YAFU wants to run only mainline code, somebody needs to make an effort to get the features into the mainline eventually.
Or
  • to split the WUs out into different sub-projects. Then apply stock Credit New to each subproject. This requires that the work generator takes results from subproject A to build new work for subproject B, and so on. (A beneficial side effect of this solution would be that contributors' hosts would be utilized better, by running single-threaded and multi-threaded applications in separate jobs.)
Oh, I forgot a 3rd potential option: Maybe higher fairness in crediting is not a priority at the YAFU project, hence the status quo remains.

(Sorry for veering off topic for a bit. Maybe I should repost at the YAFU forum.)
 

phoenicis

Member
Nov 26, 2017
56
63
91
Looking at our DC Vault stats, we seem to be in reasonable shape other than GIMPS where we are languishing in 517th place. I've dusted off the old guide and will see if I can nibble away at this.
 
  • Like
Reactions: TennesseeTony