| 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 |
I was following some posts about BOINC in the Team Room under the MOT thread, and a few of the posts there have caused some concern on my part that I thought ought to be hashed out on neutral territory, hence I'm posting this here.
There have been a number of posts in that thread which suggest that switching over to the BOINC agent results in close to a doubling of points generated by the same device. I naturally assumed that this would be due solely to BOINC being a much more efficient program, and that the increased award of points equated to an increased amount of work being completed. This was very exciting news to me, and I spent time trying to install BOINC as my new agent; many members of MOT are encouraging other members to do the same, and rightfully so. If BOINC is almost twice as efficient, then _EVERY_ member of WCG should switch to it in order to benefit mankind even more. Unfortunately, I have not been able to resolve a software conflict on my primary computer, and so I am not able to use BOINC on it. THEN I read some further posts in the MOT team thread that seems to be suggesting that the extra points awarded by BOINC are not the result of increased work, but rather due to a different award scheme. That is, for the same amount of work, some members appear to be earning more points than the rest of us. If so, then that's bogus and WCG needs to fix it immediately, otherwise there is nothing fair about our little competitions for points. And the longer this issue remains unresolved, the more that the points awarded during the interim become suspect. If we're going to play fair while competing for points, one person shouldn't earn nearly double the points for the same work just because they happen to have BOINC installed. Now, I'm perfectly happy to go along with folks using BOINC so long as it is really speeding up the process; but if it doesn't do anything except alter the amount of points awarded, then there's no sense in playing this game anymore if it's not fixed ... and fixed quickly. SO ... my questions to the Admins and people who run WCG, etc., are this: 1. Is BOINC awarding points based on the same criteria as the other agent? 2. If not, how can the difference be justified or be considered fair? 3. If not fair, will it be fixed ... and how quickly? Thanks. Cheers. Bill Velek, Team Captain of the ... |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I believe I will perform the experiment myself. Can't hurt...
I can answer question (1): BOINC uses an entirely different system to grant points, and you can read about it in the help. It is supposed to be scaled to match UD points, but the conversion is crude at best. I don't know how fair it is. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I believe I will perform the experiment myself. Can't hurt... I can answer question (1): BOINC uses an entirely different system to grant points, and you can read about it in the help. It is supposed to be scaled to match UD points, but the conversion is crude at best. I don't know how fair it is. Crude ... rough ... approximate ... is all good with me. We don't need no dadgummed bean counters for this.  My only concern was that it sounded to me as if people are suddenly receiving nearly TWICE as many points FOR THE SAME AMOUNT OF CRUNCHING. I don't think anyone can argue that that would be fair. Now, if BOINC is managing to do more work because it uses a second 'core' or 'dual processor' or whatever, as has now been mentioned in the MOT thread, then I have no complaints. More work is more work, no matter how it happens. .... just as long as there is more being accomplished.That's all I was concerned about. Cheers. Bill Velek |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Well, all they do is multiply the BOINC points by a fixed factor. If that factor is wrong, the points could be wildly skewed. I have seen oddities in the points, but nothing I can possibly pin down. My current test should yield a result in a week or two.
|
||
|
|
knreed
Former World Community Grid Tech Joined: Nov 8, 2004 Post Count: 4504 Status: Offline Project Badges: 
|
Bill Velek,
----------------------------------------The United Devices software computes points based upon three measurements which make up a computers score: 1) Processor power (as determined by a benchmark test run), 2) RAM and 3) Disk Space. These are normalized to a certain degree and used to determine the number of points that computer will receive for each unit of wall clock time the computer spends processing work. The BOINC software computes credit based upon the processor power (as determined by a benchmark test) which determines how much credit the computer will be awarded for each unit of process execution time. Additionally, all devices that return work for the same workunit are awarded the same amount of credit (see our help section of the website for details). As you can tell by reading this they are based on different measurements. The United Devices software and the BOINC software contain these scoring systems and it is not within World Community Grid's power to change these systems. We have however, tested the same workunits on the same computers running BOINC and UD in order to examine the number of points granted. We also did an examination after BOINC had been running for two weeks at World Community Grid. This resulted in the determination that on average there is a factor of 7 difference between a 'BOINC' credit and a 'UD' point. As such, when the BOINC data is imported into the website database, we uplift the 'BOINC' credit by a factor of 7 to be equivalent to a UD point. There are several key points that you need to look at when comparing your device's scoring between UD and BOINC: 1) The two agents measure time differently. The UD agent measures the amount of clock time that elapsed while the work was being processed. The BOINC agent only includes the time that the thread was actually executing. This means that comparing the points per hour of runtime between the UD agent and the BOINC agent is not a valid comparison. 2) Workunits vary in length. This means that you cannot compare the number of results returned using one agent or the other agent to determine which is faster. All of this means that there are some computers which will earn more points using one agent rather then the other. However, over the entire grid there should not be advantage to one agent or the other. If you still feel that you are not earning your fair share of points using one agent or the other, then we encourage you to switch to the other agent. The scientific work being performed has the same value on either platform. I see that you noted a software conflict when you attempted to install BOINC. Please take advantage of the Contact Us feature on the website and let us know about the conflict and we will work to help you resolve the problem. [Edit 1 times, last edit by knreed at Feb 15, 2006 10:46:49 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hi Bill - I am in agreement with knreed on this - thank you for the excellent explanation - the advantage to me using Boinc is that it allows me to use both cores on my X2 CPU, hence getting double the points, but then again I am processing two work units at once so that is to be expected.
As regards comparing points, unfortunately its like comparing apples and pears, as both use different methods of benchmarking - some types of machine will benefit from the UD agent and some from using Boinc, you just need to pick the right agent for your system. Overall across the board I believe the system is fair.  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Bill,
----------------------------------------Three of my devices are P4 2.4 GHz single core machines. One running XP Boinc and the other two running XP UD. I do not see a real difference in points or results on my WCG device stats page (my one linux box is too slow for comparison). Now, my dual core P4 3.2 machines running Boinc are a different story. With 1 G of ram, these beasts are ripping through two work units at once and are out performing everything else in my stable. Makes me want to order a dual processor Xeon or Opteron setup :) [Edit 1 times, last edit by Former Member at Feb 15, 2006 11:38:21 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I use Linux as my OS ... I have no choice but to use Boinc, and I hate it.
However, I'm sure you'll agree that points do not really matter ... it is results that count.  Drop the points system altogether, and use results ... you then would probably loose quite a few members.   |
||
|
|
David Autumns
Ace Cruncher UK Joined: Nov 16, 2004 Post Count: 11062 Status: Offline Project Badges: 
|
I've been scolded away from the MOT thread apparantly me discussing this is bad for business
---------------------------------------- So as requested I'll post here I'm running an experiment like Didactylos on 6 PC's at the moment I was scoring around the 3900 mark per day using UD so we'll see what boinc will bring. Preliminary results from my prime cruncher currently show 2.5 times the UD score  (all my PC's are single core and no Intel HT to confuse matters)- time will tell Dave   [Edit 3 times, last edit by David Autumns at Feb 17, 2006 9:37:32 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
As I've tried to say a couple of times now, as long as reasonable steps are taken to award points fairly, I don't really care if they are precisely equal. But a factor of 2.5 times seems very unreasonable and unfair to me IF ITS FOR THE SAME AMOUNT OF CRUNCHING. Now, it seems to me that I read somewhere in the rules (too lazy to look it up right now), that no device can be rated at more than double the comparison device; if true, I don't know why that limit was adopted, and it doesn't seem fair to me either. If you have a super fast computer that is four or five times as fast as the comparison device, ... and it is actually doing that much more work ..., then I don't see any reason to not award points accordingly. SO ... could it be that BOINC is not imposing that artificial limit, and the folks who are seeing a large increase in points are merely receiving what they would receive with UD without that limit? If that's the case, then again I have no problem. Points for work is okay with me. On the other hand, if I were to somehow successfully install BOINC on my computer, and began receiving twice as many points, then that would not be fair IF MY COMPUTER WERE NOT ACCOMPLISHING ANY MORE WORK as a result of a more efficient program or a program making use of unused capacity/abilities within my computer. I don't know much about computer technology, so I don't know dual core from single core, DDT (or whatever it was), fancy-schmancy, etc.; however, if there are computer resources unused by any agent when the user is willing to donate them, then I'd say that the agent needs to be fixed in order to start making better use of what is available to it. That might be easier said than done, but if it's possible then it ought to be worth it, considering the number of computers out there and the goals we all share. I still haven't gotten around to trying to install BOINC on my other computer because my son has been using it when I've been availabe to do it; he's back in his room with it right now, too.
Cheers. Bill Velek |
||
|
|
|