Index  | Recent Threads  | Unanswered Threads  | Who's Active  | Guidelines  | Search
 

Quick Go »
No member browsing this thread
Thread Status: Locked
Total posts in this thread: 277
Posts: 277   Pages: 28   [ Previous Page | 7 8 9 10 11 12 13 14 15 16 | Next Page ]
[ Jump to Last Post ]
Post new Thread
Author
Previous Thread This topic has been viewed 25841 times and has 276 replies Next Thread
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Re: BOINC Points post in Member News

carl.h, may I direct you to this post http://www.worldcommunitygrid.org/forums/wcg/...=9549&offset=20#75943 by a WCG staff member explaining how inflated claims gain extra points despite the quorum system.

As you can see, and as you freely admit, you have gained an unfair advantage over other members during the time you have been using an optimised client.

The reduced credit for over-claims is not designed to punish you for this (although there were extremists who suggested past over-claims should be reduced after the event - fortunately a more sensible approach prevailed). Instead, the reduced credit prevents anyone from profiting further, and also brings the problem instantly to the attention of anyone using an optimised client - far more effective than sending an email and waiting for people to do something about it. Self-interest is a powerful motivator.
[Nov 12, 2006 6:49:31 PM]   Link   Report threatening or abusive post: please login first  Go to top 
David Autumns
Ace Cruncher
UK
Joined: Nov 16, 2004
Post Count: 11062
Status: Offline
Project Badges:
Re: BOINC Points post in Member News

carl.h

if 2 claim 50 points for a work unit and one claims 100 then something is not right with the one claiming 100 something is amiss as all 3 have performed the same amount of calculations on that WU so all should claim around the same score.

This was not happening so IBM have implemented a solution to that issue by discarding the high claim in this case as it is plainly in error.

Optimised clients do not do more efficient work on the science bit - the bit that counts - as the crunching code does not utilise the many SSE MMX 3DNow etc etc that have been developed over the years. All it means with an optimised client is that the benchmark of that client which uses the aforementioned optimisations will try to claim an unfair amount of points for the time it spends on that work unit because it can run the benchmark - not the science bit - quicker.

I'm sorry if this is a bit blunt but us Northerners (in the UK) are reknown for it as they say we call a spade a spade.

No amount of angst is going to change these facts

dave
----------------------------------------

[Nov 12, 2006 6:54:12 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Re: BOINC Points post in Member News

As I previously stated that is fine but there was no FAIR warning, one minute it was o.k. the next I am penalised.

and as you freely admit, you have gained an unfair advantage over other members


Didactylos.....That is untrue, a total fabrication.....where did I state that ?

The reduced credit for over-claims is not designed to punish you for this


No it`s pi$$ed me off, as one minute I`m within the rules the next I`m not by means of a very vague statement.

David I know what the opti clients do but they were o.k. one minute and not the next with no definative ruling.

My client was o.k. for all the work it produced previously then suddenly not and I receive a penalty in my absence.

and also brings the problem instantly to the attention of anyone using an optimised client


Absolute tosh !! It was at least a week until I realised something was wrong.

If someone had posted even here that crunch3rs client was being outlawed and penalised word would have spread quickly instead some vaguery about SUBSTANTIALLY HIGH CLAIMS was posted.

If the new system overcame the opti`s WHY the need for penalties ???? Unthinking, callous power ?

Self-interest is a powerful motivator.


This again is tosh !! It made me want to chuck in DC or find a way to cheat for definate. But I`ll probably stay around to annoy gloaters like Didactylos who most on one side know is gloating and trolling !
----------------------------------------
[Edit 2 times, last edit by Former Member at Nov 12, 2006 7:44:21 PM]
[Nov 12, 2006 7:39:01 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Re: BOINC Points post in Member News

We're just going round in circles here.

I will let WCG's actions speak for themselves.
[Nov 12, 2006 8:04:09 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Re: BOINC Points post in Member News

Giving someone 1/2 credit for equal work is obviously punitive.

And using the mean instead of the median does not help those with low claims, it's hurts them more than anyone.

I see a lot more quorums that are something like 45,45,32 as opposed to 45,32,32, leading me to believe the 32 is the low claim. Now, not all of my quorums have a low claim. So I am only penalised by averaging in the low claim some of the time. But the machine with the low claim is now penalized by averaging in that low claim 100% of the time, whereas before he would get the median.
[Nov 12, 2006 8:32:23 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Re: BOINC Points post in Member News

In a normal (gaussian) distribution, the median is equal to the mean. Therefore, there should be no difference if claims have the expected distribution.

Low and high outliers are discarded before the mean calculation. WCG are keeping the precise algorithm confidential.
[Nov 12, 2006 8:50:36 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Re: BOINC Points post in Member News

The population in the quorum is 3 and it is rarely normally distributed. Plus, if its identical, why change from median to mean.

The only explanation given is that it is to help those poor soles that bench below their crunching capability and thus claim low. But it doesn't. It hurts everyone a little and them most.
[Nov 12, 2006 9:28:01 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Re: BOINC Points post in Member News

Btw, I did notice the lower ones getting thrown out in the beginning. Doesn't seem to happen much anymore.
[Nov 12, 2006 9:29:56 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Re: BOINC Points post in Member News

My Core 2 duo E6600 @ 3300Mhz does a faah in approx 3.3 Hours

Claiming now between 40-45

My Amd 3800+ X2`s taking 6.6-7 hours per faah

Claim 65-75

The claiming appears to be more time based NOW this is not how it should be !

FUBAR

Someone else please bear me out on this or not ?
----------------------------------------
[Edit 4 times, last edit by Former Member at Nov 12, 2006 9:49:04 PM]
[Nov 12, 2006 9:39:57 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Re: BOINC Points post in Member News

It works the same as before. The boinc benchmark is not customised to any given app, so it scores without regard to how well the machine will crunch. If you crunch faster than expected you'll claim lower, slower will claim higher.
[Nov 12, 2006 9:54:08 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Posts: 277   Pages: 28   [ Previous Page | 7 8 9 10 11 12 13 14 15 16 | Next Page ]
[ Jump to Last Post ]
Post new Thread