less credit for more powerfull PC's??


log in

Advanced search

Message boards : Number crunching : less credit for more powerfull PC's??

Author Message
Noerrebronx
Send message
Joined: 28 Oct 99
Posts: 10
Credit: 425,728
RAC: 0
Denmark
Message 28268 - Posted: 20 Sep 2004, 7:17:33 UTC

I dont understand. I have 5 machines running BOINC. The most powerfull one is the one that has done FAR more workunits that any of more other PCs (IT is doing 8-10 per day). Still it is of the PCs the one that have been giving the smallest amount of creadits. It is like any credits comming from this PC is dissapearing into the blue air as it has not moved for a while on the statistics. Anybody has the same problem?

Profile Bruno G. Olsen & ESEA @ greenholt
Volunteer tester
Avatar
Send message
Joined: 15 May 99
Posts: 875
Credit: 4,325,371
RAC: 966
Denmark
Message 28271 - Posted: 20 Sep 2004, 7:42:07 UTC

There's actually three issues here ;-)

First of all there are rumours that faster computers get less credit per wu - as you can see in http://setiweb.ssl.berkeley.edu/forum_thread.php?id=4265 there's a bit of a disagreement on this.

The second issue is that you won't get any credit unless two others has done the same wu and have come up with the same result. It would seem a bit odd that one computer gets a lot of wu's for which there's going to be a while until two others have done the work and/or alot of wu's for which hosts disagrees on the result - but in theory it could happen. (I've done a wu on LHC where 4 hosts still don't agree on the result). As far as I know, when hosts disagree, the wu will be sent to more hosts, and when three of them agree on the result, only those three will get credit (maybe a very small amount will be granted for the others).

There are others that have noticed stagnation with the credits http://setiweb.ssl.berkeley.edu/forum_thread.php?id=4472

Personally I've noticed this for a few short periods (not counting downtime on the project) - but I believe it's becourse there hasn't been two others yet that have done the wu's in question and/or my result didn't agree with two others.

Right now, I'm not experiencing a "no credit"-period


S@h Berkeley's Staff Friends Club ©

Noerrebronx
Send message
Joined: 28 Oct 99
Posts: 10
Credit: 425,728
RAC: 0
Denmark
Message 28274 - Posted: 20 Sep 2004, 8:09:05 UTC

I do not think it is related to workunits not having been distributed to others. The other computers I'm running are receiving credits fine. The strange thing here is not that credits are not given at all, but that credits are not given to a specific PC, while it is given to the rest of my PCs. I have tried to reinstall boinc and copy the account.xml file over again from one of the PCs that are working, but it still does not help.

Profile Bruno G. Olsen & ESEA @ greenholt
Volunteer tester
Avatar
Send message
Joined: 15 May 99
Posts: 875
Credit: 4,325,371
RAC: 966
Denmark
Message 28275 - Posted: 20 Sep 2004, 8:18:00 UTC

Did you by any chance notice the last figures on the wu's from that pc? I know it would seem extremely coincidental, but if the pc's processed alot of wu's that end with _3 or above it could mean that it gets alot of wu's where others don't agree on the result and/or not returning results within the deadline and/or not returning results at all. I admit, it seems strange that this would occur in that amount for one pc only


S@h Berkeley's Staff Friends Club ©

Profile Paul D. Buck
Volunteer tester
Send message
Joined: 19 Jul 00
Posts: 3898
Credit: 1,158,042
RAC: 0
United States
Message 28293 - Posted: 20 Sep 2004, 9:30:09 UTC - in response to Message 28274.

> I do not think it is related to workunits not having been distributed to
> others. The other computers I'm running are receiving credits fine. The
> strange thing here is not that credits are not given at all, but that credits
> are not given to a specific PC, while it is given to the rest of my PCs. I
> have tried to reinstall boinc and copy the account.xml file over again from
> one of the PCs that are working, but it still does not help.

As Bruno said ...

There is some belief that the faster computers are, in fact, claiming less credit per result than the other machines. Since this is a cosmetic issue it is not one I would spend time on at this juncture (Personal opinion of Paul, for what that is worth; with it and $100 you can get a cup of coffee at Starbucks).

If the processor is overclocked or unstable you could be losing out because your results are not passing validation. Since we don't have visibility for your results there is no way to tell right now if this is the case.

One suggestion would be to try to get into one of the smaller projects when it opens up again for new participants. Predictor@Home is working on increasing the capacity so they may open up when they come back. LHC@Home is also muttering aboout opening up again at the end of the month...

Then you could look to see if you have a machine issue with the results processed.

My only other thought is to have that machine work cp.net until you can confirm or refute the problem is with that machine. Siince each model is stand-alone you are not competing with, or relying upon, anyone else.
<p>
For BOINC Documentaion: Click Me!


STE\/E [BlackOps]
Volunteer tester
Send message
Joined: 29 Mar 03
Posts: 1137
Credit: 3,233,449
RAC: 160
Message 28302 - Posted: 20 Sep 2004, 10:08:48 UTC

If the processor is overclocked or unstable you could be losing out because your results are not passing validation. Since we don't have visibility for your results there is no way to tell right now if this is the case.
==========

I agree with that Paul, if an Overclocked System is Unstable in any way at all you just going to get a lot of Error's on the WU's and recieve no Credit.

I've been running a P4 3.4 EE @ 3.763 GHZ for the last couple of weeks over at CPDN with no problems so far. But the system is very Stable, right now the CPU is only running at 30c crunching 2 CPDN WU's at once in HT Mode ...

I've also crunched some BOINC Seti WU's with it but have no idea what kind of Credit it is getting for the WU's. I haven't had a single error yet running the Seti Units though... :)

sysfried
Send message
Joined: 3 Apr 99
Posts: 3
Credit: 630,495
RAC: 100
Germany
Message 28305 - Posted: 20 Sep 2004, 10:30:53 UTC - in response to Message 28268.

> I dont understand. I have 5 machines running BOINC. The most powerfull one is
> the one that has done FAR more workunits that any of more other PCs (IT is
> doing 8-10 per day). Still it is of the PCs the one that have been giving the
> smallest amount of creadits. It is like any credits comming from this PC is
> dissapearing into the blue air as it has not moved for a while on the
> statistics. Anybody has the same problem?
>
>

I have the same problem. I have 2 PC's. one Athlon XP 2700+ at 2.17 GHz (not overclocked) and a Dual Opteron 244 System (1.8 GHz).

Both systems need about the same time in min per WU (although the opterons do 2 WU's at a time) but my 2700+ is about head to head with my opterons in terms of score and average credit... I had them run 2 weeks full load ...

Now I removed the SAH project because this seems like a waste of resources. I'm waiting for the CERN Project to start for the open public.

Just my 2 cents here

Saimek
Send message
Joined: 25 Jan 00
Posts: 121
Credit: 454,423
RAC: 0
Poland
Message 28312 - Posted: 20 Sep 2004, 10:52:26 UTC - in response to Message 28293.

> If the processor is overclocked or unstable you could be losing out because
> your results are not passing validation. Since we don't have visibility for
> your results there is no way to tell right now if this is the case.
>

I run my Athlon 2000+XP @ 1900 Mhz. I recieved my credit normally but havent recieved any last week. Is it possible that my comp is returning bad results ? On max load CPU temp is about 45-48 degrees so i think it's not too hot. I Havent got any probs with stability so?..


Profile Paul D. Buck
Volunteer tester
Send message
Joined: 19 Jul 00
Posts: 3898
Credit: 1,158,042
RAC: 0
United States
Message 28358 - Posted: 20 Sep 2004, 14:06:19 UTC

Unfortunately, we are speculating here because of the lack of good information. It is one of those sad but true things.

As work around we were suggesting the other avenues to attempt to see if there might be an alternative way to islolate the problem.
<p>
For BOINC Documentaion: Click Me!


STE\/E [BlackOps]
Volunteer tester
Send message
Joined: 29 Mar 03
Posts: 1137
Credit: 3,233,449
RAC: 160
Message 28368 - Posted: 20 Sep 2004, 14:35:02 UTC
Last modified: 20 Sep 2004, 14:35:27 UTC

It was my understanding when I first started running the Seti Beta Project that the Theory behind the Credit System was to give the slower PC's an equal share of the Credits so as to make them feel like they could compete with the faster PC's...With the old Seti Classic you got 1 Credit for 1 Work Unit so naturally the faster PC's far out paced the slower ones.

I thought with the new credit system you could run a P3 850 & a P4 3.4 24/7 for a month and they would both end up with the same amount of Credit...D'oh was I wrong on that one...

Where I picked that up from I don't know other than a few posts and explanations about the Credit System I guess. Well from running all the different projects I think we can throw that Theory out the window. Maybe that was the way it was supposed to work but by now we all know it doesn't work that way.

Take the CPDN Project for example...My P3 850 will receive credit once a day for 75 Credits, my P4 3.4 EE will receive Credit 7-8 times a day for 525-600 Credits a day...Also I can look at my Computers that I'm running the Seti WU's with and the faster ones are far ahead of the slower ones for Credit recieved...

Profile ThePhantom86
Volunteer tester
Avatar
Send message
Joined: 30 Jan 02
Posts: 268
Credit: 1,728,001
RAC: 0
United States
Message 28383 - Posted: 20 Sep 2004, 15:02:58 UTC

One thing I've heard is that while you may recieve less credit per wu for a faster machine, the faster machine will do more wu's than the slower one so in theory, the faster cpu would recieve more credit in the end.

And I'm pretty sure it's the same here, but in Predictor a wu was processed by the same type of processor. I can't remember if it was the same way in Beta or not seeing as how I can't see my results since Berkeley turned them off.



Honie
Send message
Joined: 22 Jan 04
Posts: 141
Credit: 29,681,066
RAC: 0
Germany
Message 28390 - Posted: 20 Sep 2004, 15:27:02 UTC - in response to Message 28368.

> It was my understanding when I first started running the Seti Beta Project
> that the Theory behind the Credit System was to give the slower PC's an equal
> share of the Credits so as to make them feel like they could compete with the
> faster PC's...With the old Seti Classic you got 1 Credit for 1 Work Unit so
> naturally the faster PC's far out paced the slower ones.


I could not agree with this statement. The credit system will NOT equalize
fast and slow pcs. It will only equalize the processing time of different WUs.
If you get 1 Credit for a WU which takes 5 min to resolve and you get 1 Credit
for a WU which takes 4 hours to resolve on the SAME System, thats the
difference the credits system should solve. And when in the future the seti client 4.03 will be changed and different calculations will be done, you can still compare the work of your Clients which they did before , with this Credit system.

But naturally a fast CPU earn more Credits than a slow CPU.

There is still a Bug for calculating Credits for HT CPU's - as far as I know -
But if you look how a Cobblestone is defined, it should be clear, that there
are and will be differences between slow and fast CPU because in fact they are
doing different amount of work. And I think it is the way it should be!

Arm
Send message
Joined: 12 Sep 03
Posts: 308
Credit: 15,494,254
RAC: 2,237
Message 28409 - Posted: 20 Sep 2004, 16:53:07 UTC - in response to Message 28302.

>> If the processor is overclocked or unstable you could be losing out because
>> your results are not passing validation. Since we don't have visibility for
>> your results there is no way to tell right now if this is the case.
> ==========
>
> I agree with that Paul, if an Overclocked System is Unstable in any way at all
> you just going to get a lot of Error's on the WU's and recieve no Credit.
>
> I've been running a P4 3.4 EE @ 3.763 GHZ for the last couple of weeks over at
> CPDN with no problems so far. But the system is very Stable, right now the CPU
> is only running at 30c crunching 2 CPDN WU's at once in HT Mode ...
>
If somebody wants to have an ultimate proof of system stability he'd use Prime95. A hard test for, at least 8-9 hrs, is enough in most cases.



S@h Berkeley's Staff Friends Club ©

Noerrebronx
Send message
Joined: 28 Oct 99
Posts: 10
Credit: 425,728
RAC: 0
Denmark
Message 28559 - Posted: 21 Sep 2004, 6:55:21 UTC - in response to Message 28409.

I came in this morning and checked my stats. The computer in question had updated this morning. Guess it must have been a long string of results that had not been processed by anybody else.
Thanks for your help though. I did gain a little bit more knowledge about BOINC.

Dr Jacks
Avatar
Send message
Joined: 23 Feb 03
Posts: 1
Credit: 5,715
RAC: 0
United States
Message 28587 - Posted: 21 Sep 2004, 9:38:18 UTC

I have been running this for almost 2 weeks now and have receive 0 credit... I think the "classic" version is much more stable and should not be "decommissioned" until they have worked ALL of the bugs out of this new version.
Hoping I can clear the way
By stepping through my shadow
Coming out the other side
Step into the shadow
Forty six and two are just ahead of me
~Tool

Saimek
Send message
Joined: 25 Jan 00
Posts: 121
Credit: 454,423
RAC: 0
Poland
Message 28589 - Posted: 21 Sep 2004, 10:13:12 UTC - in response to Message 28368.

> I thought with the new credit system you could run a P3 850 & a P4 3.4
> 24/7 for a month and they would both end up with the same amount of
> Credit...D'oh was I wrong on that one...

It's not what way. Both computers recieve the same amount of credit for one WU but the faster comp is doing more WU's so it's getting more credits. It would be unfair if P3 850 and a P4 get the same amount of credit, dont u think?



Message boards : Number crunching : less credit for more powerfull PC's??

Copyright © 2014 University of California