| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Community
Forum: Chat Room
Thread: Are points being awarded fairly? |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 103
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
The point conversion factor was developed by comparing global results. Meaning that about half the processors do better with UD than BOINC and vice-versa. More interesting would be to develop a method of predicting which platform produces more points for a particular system. For example, what are the geeneral characteristics of a computer system that produces more points using UD rather than BOINC?
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
The point conversion factor was developed by comparing global results. Meaning that about half the processors do better with UD than BOINC and vice-versa. More interesting would be to develop a method of predicting which platform produces more points for a particular system. For example, what are the geeneral characteristics of a computer system that produces more points using UD rather than BOINC? "More interesting ..." -- I'll concede that point, but from a humanitarian perspective, and ignoring "points" which have no value at all except for the competitive aspect, it would be more useful to develop a method of predicting which program operating on a given platform/computer accomplishes the most work for the good of mankind. Assuming, that is, that there is a difference in the efficiency of one program over the other (BOINC vs. UD). In other words, if I took two identical computers and had them do nothing else except crunch the same work unit, with one using BOINC and the other UD, whichever one finished fastest would indicate which program is best for that computer (platform, hardware, configuration, etc., or whatever). What would be super fantastic is if WCG could make those determinations, compile a database, and then analyze a user's computer and recommend the best program for it. Or would it depend upon other things, like interruptions of the program when user's do other things with their computers? Cheers. Bill Velek |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
The variation in points is due to the different benchmarking systems, not any actual difference in performance.
There may be some performance difference, but nobody has measured it, and I consider it to be negligible. However, the totally different benchmarking systems are going to significantly benefit certain setups on one client or the other. Only detailed knowledge of both systems OR comprehensive testing will tell you which client to run. And, as you have figured out - it isn't worth it, since the actual crunching is largely unaffected. |
||
|
|
David Autumns
Ace Cruncher UK Joined: Nov 16, 2004 Post Count: 11062 Status: Offline Project Badges: 
|
Hi Bill
----------------------------------------My machines were nowhere near the buffers on the UD Benchmarking scores so they haven't suddenly been unleashed as a result of BOINC being able to cope with their speed tomorrow night we'll know for sure I'll have all six on for the full 24 5 athlons and an intel mobile processor all single core no hyperthreading all now running just one work unit at a time on BOINC remember nothings changed but as a result of the new boinc agent it's about 0.3% more efficient as there's no waiting around for the uploads and downloads The UD 6 days worth in a day score was 3900 there or thereabouts. If nothing is amiss then I should get 3912  <GRIN>  |
||
|
|
David Autumns
Ace Cruncher UK Joined: Nov 16, 2004 Post Count: 11062 Status: Offline Project Badges:
|
the UD waitings are "alledgedly" See thread 1505 (Oh and who wrote it
----------------------------------------)Processor Score * 0.4 + Memory Score * 0.25 + Storage Score * 0.05 + Network Score * 0.3 = Overall Score rounded down to the nearest full number and capped to a maximum of 200 As storage and network score are fundementally fixed if you set the max storage you always get 196 for storage and 100 for network Only memory and processor come into this equation and it's heavily waited in favour of processor. It should be possible to introduce a correction factor that is within a reasonable margin of error |
||
|
|
David Autumns
Ace Cruncher UK Joined: Nov 16, 2004 Post Count: 11062 Status: Offline Project Badges:
|
with only a partial cut over to boinc about 50:50 UD:BOINC today it's not looking too promising for tomorrow night
----------------------------------------with just 4 days and 16 hours of run time posted I got my highest points score ever of 5118 and I've gained 10 places in the last 2 days and previously I was running as fast as I could to maintain that 469th position. Roll on another 24hrs of pure BOINC I now suspect it will be at least twice as much as with UD  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
In the MOT team thread (this post now transferred here at MOT's request), Adywebb said:
The number of units download to your machine depends on how often you have it set up to communicate with the server in your default profile (the default is 0.1 days) - the longer the gap, the more units it downloads to make sure you have enough work until the next time it communicates. I presume that the above statement pertains to the BOINC agent, and not to UD, correct? Because I've just carefully gone through all the pages I can find in the UD agent re my profile and device settings, and I can't find such a setting. Also, a statement was made elsewhere that BOINC does not award points until validation is completed, and that UD awards points immediately. I don't know if that is true about UD or not; I thought that points are validated first, which would explain how my computer has, on a couple of occasions, reported as many as a hundred run time hours in a single 24-hour day. I can't think of what else would explain that except that I am picking up credits for work done on other days. Cheers. Bill Velek |
||
|
|
David Autumns
Ace Cruncher UK Joined: Nov 16, 2004 Post Count: 11062 Status: Offline Project Badges:
|
Hi Bill
----------------------------------------Points for UD are credited immediately and will appear in the next 6 hourly stats update for your personal stats and appear in your teams stats at the 00:30GMT'ish daily update. The reason why some days you get more points than others is down to the fact that some work units take longer to crunch than others and you get points for run time so if you get a work unit lasting 36 hours say you won't get any points for that device one day and a day and a halves worth the following day. You get the same amount of points for 10 1 hour WU's as you do for 1 10hour Work Unit. The BOINC points are only allocated to you once that same workunit has been crunched on a number of crunchers PC's and the results have been checked against each other and proved to be accurate (this function also happens with the WU's that are crunched through UD - we think it might be crunched in 5 different places to verify the accuracy of each work units) This checking function before allocating your points with BOINC causes a slight delay to the arrival of your points not so with UD. However this all said 6 days worth of crunching with UD produces as much scientific info as 6 days worth of crunching using BOINC so ideally as you point out in your first post the number of points in both cases should be there or thereabout the same with UD as with BOINC. First results show that this may be miles out with about 2.5 times as many points being doled out by BOINC as UD This although not being the end of the world or anything nearly as important in the grand scheme of things it would be nice if our history of UD crunching could be aligned with our probable BOINC future. Otherwise derivations such as the ongoing improvement in the crunching power of the grid would be impossible to infer from the figures we have before us as they could be before with purely UD - this is just one example of what would be lost if reasonable parity could not be maintained going forward. So it's worth a bit of an experiment in my book Dave |
||
|
|
David Autumns
Ace Cruncher UK Joined: Nov 16, 2004 Post Count: 11062 Status: Offline Project Badges:
|
here's the 6:30 update
----------------------------------------So far today I've been credited with 3 days 11 hours and 39 minutes of crunch time and already with 6 hours to go I have once again my greatest score ever I've bagged 5,714 points already which extrapolating to a 6 days in day figure which remember used to net me ~3900 points with UD results in a score of 9836 (give or take )Whichever way you look at it you get 2.5 * the points UD would have allocated for no extra effort/electric bill/pc building expertise or crunching being provided to the World Community Grid project knreed please could we have a divide by 2.5 introduced into the Windows BOINC script PDQ or those figures won't be worth the webpage they've been written on Roll on half past midnight (here in the UK) Dave |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
You may be getting that sort of point factor, but remember that other people are not.
If "fair on average" is the best that the developers can do, then we will have to live with that - and enjoy the points advantage. It is equally possible that the UD system was penalising you for some ill-defined reason. Hey, I'm a Senior Cruncher! |
||
|
|
|