Advanced search

Forums : Technical Support : Comments on Fixed Credit System
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4 · Next

AuthorMessage
Profile Wang Solutions
Avatar

Send message
Joined: 28 Jun 07
Posts: 24
Credit: 1,020,464
RAC: 0
Message 733 - Posted: 3 Jul 2007, 5:22:53 UTC

Regarding the credits, I am finding that the credit is low on a Pentium 4 compared to other projects, but is much better on Core2Duos, though still lower than many other projects (but better than some).

I would agree that about 7 per work unit would put it in line with most projects on most machines.

ID: 733 · Report as offensive     Reply Quote
[B^S] sTrey
Avatar

Send message
Joined: 30 Jun 07
Posts: 2
Credit: 21,257
RAC: 0
Message 740 - Posted: 3 Jul 2007, 8:45:55 UTC
Last modified: 3 Jul 2007, 8:46:42 UTC

Same as many others.. 5 is a bit low for my (Windows) P4s, they average about 10 credit/hour on the majority of the projects I run and on that basis, the wus I'm getting here would average about 7.
ID: 740 · Report as offensive     Reply Quote
STE\/E
Volunteer tester

Send message
Joined: 12 Jun 07
Posts: 375
Credit: 16,539,257
RAC: 0
Message 748 - Posted: 3 Jul 2007, 10:39:46 UTC
Last modified: 3 Jul 2007, 11:24:57 UTC

The problem with a Fixed Credit System is that a 1 WU can take 8 minutes to do & you get 5 Credits, then the next can take 16 minutes & you still only get 5 Credits. Unless the Projects Wu's are all within a few seconds of each other if they are short Wu's like these here at Cosmology or a few minutes of each other for Wu's that take 1 or 2 hours to do then a Fixed Credit system isn't the best way to go.

If you get a lot of the 8 minute ones your in the Gravy, but if you get mostly 16 minute ones then your behind the 8 Ball, and I'm sure it won't take some ingenious people long to figure out which ones are the shorter ones and just start to abort the longer ones.

A Sliding Credit System such as a Project like ABC has implement is a much better way to go for Wu's that can vary in length such as the ones for this Project. That type of System takes into account for the varying length of the Wu's.

Depending on just how much Credit a Project is willing to dole out then you may only get 3 or 4 Credits for the 8 minute Wu but you'll get 6 or 7 for the 16 minute ones.

How much Credit you would get actually depends on the speed of your system, I'm just using the 8 min & 16 min Time Frames as a reference point because that's what most of my PC's seem to run the Wu's in ... :)
ID: 748 · Report as offensive     Reply Quote
Profile Scott
Volunteer moderator
Project administrator
Project developer
Avatar

Send message
Joined: 1 Apr 07
Posts: 662
Credit: 13,742
RAC: 0
Message 752 - Posted: 3 Jul 2007, 13:24:56 UTC

I'm going to leave it at 6 for now; I'd rather over-reward some people that under-reward many.

There's another thing that's been bothering me reading these posts. You tend to think that 10 cr/h is a good average for *any* machine. So are we saying that time spent computing is the important thing here? Should all users be compensated equally for their time, even if the time isn't equally well spent (i.e. Pentium 2 and Pentium 4, both running for 1 hour, get the same credit, even though the P4 churned out 10 times as many results)?

I'm not favoring one over the other, really. What would you prefer, though: effort-based or merit-based rewards?
Scott Kruger
Project Administrator, Cosmology@Home
ID: 752 · Report as offensive     Reply Quote
Profile sysfried

Send message
Joined: 24 Jun 07
Posts: 114
Credit: 5,296,905
RAC: 0
Message 755 - Posted: 3 Jul 2007, 13:31:48 UTC - in response to Message 752.  

I'm going to leave it at 6 for now; I'd rather over-reward some people that under-reward many.

There's another thing that's been bothering me reading these posts. You tend to think that 10 cr/h is a good average for *any* machine. So are we saying that time spent computing is the important thing here? Should all users be compensated equally for their time, even if the time isn't equally well spent (i.e. Pentium 2 and Pentium 4, both running for 1 hour, get the same credit, even though the P4 churned out 10 times as many results)?

I'm not favoring one over the other, really. What would you prefer, though: effort-based or merit-based rewards?


Dear Scott.

Stick with Boinc standards. A faster host will result more credits than a slower host (in terms of cpu power). That's it.

Don't make your life too complicated...

Sysfried
Happy member of Team: Planet 3D Now!

ID: 755 · Report as offensive     Reply Quote
Profile [BAT] tutta55
Volunteer moderator
Volunteer tester
Avatar

Send message
Joined: 7 Jun 07
Posts: 217
Credit: 710,406
RAC: 0
Message 756 - Posted: 3 Jul 2007, 13:33:59 UTC - in response to Message 752.  
Last modified: 3 Jul 2007, 13:35:15 UTC

I'm going to leave it at 6 for now; I'd rather over-reward some people that under-reward many.


I think 6 will do nicely, until we have sufficient data to see our averages over a longer period of time. When I look at a few other machines I have, 7 would be a bit much.

There's another thing that's been bothering me reading these posts. You tend to think that 10 cr/h is a good average for *any* machine. So are we saying that time spent computing is the important thing here? Should all users be compensated equally for their time, even if the time isn't equally well spent (i.e. Pentium 2 and Pentium 4, both running for 1 hour, get the same credit, even though the P4 churned out 10 times as many results)?

I'm not favoring one over the other, really. What would you prefer, though: effort-based or merit-based rewards?


I think you misunderstood these people. They were referring to their specific machine (or else I misunderstand them :p ). Rewarding on anything else than actual work done, would be a first in Boinc projects. For the credit hunters it would make this an ideal project to dump Pentium II and III on, and put performant machines on other projects.

BOINC.BE: For Belgians who love the smell of glowing red cpu's in the morning
Tutta55's Lair
ID: 756 · Report as offensive     Reply Quote
Profile [B^S] Acmefrog
Volunteer tester
Avatar

Send message
Joined: 8 Jun 07
Posts: 175
Credit: 446,074
RAC: 0
Message 761 - Posted: 3 Jul 2007, 14:01:41 UTC

Maybe I am in the same world as Tutta (maybe not) but giving a fixed credit per WU should matter the pc. For example:

Comp. #1 (fast pc) does 6 WUs in an hour and gets 30 credits total (6 WU * 5 credit per WU)

Comp #2 (slow pc) does 2 WU per hour would get 10 credits total (2 WU * 5 credit per WU)

Each did the same amount of work on the WU. Each should be rewarded equally. The person with the faster machine will just be able to complete more.

I would say you should just pick a number and use it. No matter what number you pick some people will say it is too much and other will say it is not enough.
ID: 761 · Report as offensive     Reply Quote
STE\/E
Volunteer tester

Send message
Joined: 12 Jun 07
Posts: 375
Credit: 16,539,257
RAC: 0
Message 765 - Posted: 3 Jul 2007, 14:47:51 UTC

I would say you should just pick a number and use it. No matter what number you pick some people will say it is too much and other will say it is not enough.


Exactly, your never going to please everybody unless you give 100 Credits Per Hour, and even that would probably disgruntle a few people ... :)

6 sounds good, leave it at that for awhile, not all Projects give the same amount of Credit anyway, you can always find a Project that gives more or less and start saying well I get this much over here or I don't get this much.

I think as much as a lot of people would like the Credits even at all Projects it's probably never going to happen.
ID: 765 · Report as offensive     Reply Quote
Profile Scott
Volunteer moderator
Project administrator
Project developer
Avatar

Send message
Joined: 1 Apr 07
Posts: 662
Credit: 13,742
RAC: 0
Message 773 - Posted: 3 Jul 2007, 16:10:44 UTC - in response to Message 761.  

Maybe I am in the same world as Tutta (maybe not) but giving a fixed credit per WU should matter the pc. For example:

Comp. #1 (fast pc) does 6 WUs in an hour and gets 30 credits total (6 WU * 5 credit per WU)

Comp #2 (slow pc) does 2 WU per hour would get 10 credits total (2 WU * 5 credit per WU)

Each did the same amount of work on the WU. Each should be rewarded equally. The person with the faster machine will just be able to complete more.

Not true. While both computers spent the same time crunching, the fast one did 3 times as much work.

The problem with giving equal credit to equal time spent is that someone who has 2 of old P2s from 10 years ago would get twice as many credits as the person who has one top-of-the-line machine that does 40 times as much work. I don't think that's fair to the people who invest in newer computers.
Scott Kruger
Project Administrator, Cosmology@Home
ID: 773 · Report as offensive     Reply Quote
Profile Jord
Volunteer moderator
Volunteer tester
Avatar

Send message
Joined: 15 Jun 07
Posts: 345
Credit: 50,500
RAC: 0
Message 775 - Posted: 3 Jul 2007, 16:19:18 UTC - in response to Message 773.  

The problem with giving equal credit to equal time spent is that someone who has 2 of old P2s from 10 years ago would get twice as many credits as the person who has one top-of-the-line machine that does 40 times as much work. I don't think that's fair to the people who invest in newer computers.

I don't even see a P2 do one result in one hour. Perhaps you want to test with older CPUs, to see how much hours or days they take. :-)
ID: 775 · Report as offensive     Reply Quote
Profile [B^S] Acmefrog
Volunteer tester
Avatar

Send message
Joined: 8 Jun 07
Posts: 175
Credit: 446,074
RAC: 0
Message 777 - Posted: 3 Jul 2007, 17:43:29 UTC - in response to Message 773.  

Maybe I am in the same world as Tutta (maybe not) but giving a fixed credit per WU should matter the pc. For example:

Comp. #1 (fast pc) does 6 WUs in an hour and gets 30 credits total (6 WU * 5 credit per WU)

Comp #2 (slow pc) does 2 WU per hour would get 10 credits total (2 WU * 5 credit per WU)

Each did the same amount of work on the WU. Each should be rewarded equally. The person with the faster machine will just be able to complete more.

Not true. While both computers spent the same time crunching, the fast one did 3 times as much work.

The problem with giving equal credit to equal time spent is that someone who has 2 of old P2s from 10 years ago would get twice as many credits as the person who has one top-of-the-line machine that does 40 times as much work. I don't think that's fair to the people who invest in newer computers.


No maybe I wrote wrong but I was saying per WU not time. In my example the faster machine will be able to do more WUs than the slower one during the same amount of time. Thus the faster machine will obtain more credit per hour ONLY because their machine can handle more WUs per hour.
ID: 777 · Report as offensive     Reply Quote
Profile [B^S] Gamma^Ray
Volunteer tester
Avatar

Send message
Joined: 7 Jun 07
Posts: 47
Credit: 70,587
RAC: 0
Message 783 - Posted: 3 Jul 2007, 19:29:13 UTC

For now, I would just suggest leaving it at 6 as you have it, As this is in Alpha stage at the moment, Im sure there are other issues of more importance to deal with. And later on when there are more data to compare too, You can decide whether to up the c/wu, Or completely change it to something else. You still have Beta to work things out after Alpha. ;) My original issue was the Linux os was running so much faster per work unit, That it was causing me to get get less than one credit per work unit done on some of them, And not the faster for sure, A 3800x2 should be worthy of a little more than that regardless of its OS. HeHe But being at 5,and now 6, Issue solved on that one. :)

Regards
G^R
Windows-XP-Pro, AMD 3800X2, 5.10.28
ID: 783 · Report as offensive     Reply Quote
Profile [AF>Futura Sciences>Linux] Thrr-Gilag
Volunteer tester
Avatar

Send message
Joined: 22 May 07
Posts: 32
Credit: 145,576
RAC: 0
Message 792 - Posted: 3 Jul 2007, 23:42:55 UTC - in response to Message 775.  

The problem with giving equal credit to equal time spent is that someone who has 2 of old P2s from 10 years ago would get twice as many credits as the person who has one top-of-the-line machine that does 40 times as much work. I don't think that's fair to the people who invest in newer computers.

I don't even see a P2 do one result in one hour. Perhaps you want to test with older CPUs, to see how much hours or days they take. :-)

You're answer tommorow ^^

But I bet rather on 1.5 hours
------
Thrr-Gilag Kee'rr

L'Alliance Francophone
ID: 792 · Report as offensive     Reply Quote
Profile Jayargh
Volunteer moderator
Volunteer tester
Avatar

Send message
Joined: 25 Jun 07
Posts: 508
Credit: 2,282,158
RAC: 0
Message 793 - Posted: 4 Jul 2007, 0:06:15 UTC - in response to Message 792.  

The problem with giving equal credit to equal time spent is that someone who has 2 of old P2s from 10 years ago would get twice as many credits as the person who has one top-of-the-line machine that does 40 times as much work. I don't think that's fair to the people who invest in newer computers.

I don't even see a P2 do one result in one hour. Perhaps you want to test with older CPUs, to see how much hours or days they take. :-)

You're answer tommorow ^^

But I bet rather on 1.5 hours


Linux or windows? Big difference

ID: 793 · Report as offensive     Reply Quote
Profile Wang Solutions
Avatar

Send message
Joined: 28 Jun 07
Posts: 24
Credit: 1,020,464
RAC: 0
Message 797 - Posted: 4 Jul 2007, 5:44:32 UTC

I am happy to go with 6 also.

Regarding the 10cr/hr scenario, that seems to be an average of what most projects offer on a typical Pentium 4 computer (per thread with HT on). At 6 credits per WU, that is about what a P4 will now earn.

But faster more modern computers will appropriately earn much more - a core 2 duo with 64-bit linux app will earn a similar amount of credit to that offered by ABC, Riesel Sieve etc - ie 4-5 times that earned by the P4.

ID: 797 · Report as offensive     Reply Quote
Profile Saenger
Volunteer tester
Avatar

Send message
Joined: 22 May 07
Posts: 110
Credit: 353,577
RAC: 0
Message 798 - Posted: 4 Jul 2007, 6:14:47 UTC
Last modified: 4 Jul 2007, 6:15:29 UTC

I think that the current Linux c/h is more appropriate once the better Windows compilation is available.
Now I'm getting for my AthlonXP2200+, not the newest race horse, about 20 c/h with 6 c/WU.

For the moment the majority of crunchers is still using Windows and simply doesn't use the whole possible crunch power of it's machines for this project, or better can't utilize it because of the bad compilation. Once the compilation is good for Windows as well 2.5 - 3 c/WU should be the right amount.

As long as the credit situation is as it is I think Linux crunchers will be attracted, as they get far more credits then usually, even more then with SIMAP who seem to have a similar pro-Linux bias.
Grüße vom Sänger
ID: 798 · Report as offensive     Reply Quote
Profile [AF>Futura Sciences>Linux] Thrr-Gilag
Volunteer tester
Avatar

Send message
Joined: 22 May 07
Posts: 32
Credit: 145,576
RAC: 0
Message 810 - Posted: 4 Jul 2007, 11:31:37 UTC - in response to Message 793.  

The problem with giving equal credit to equal time spent is that someone who has 2 of old P2s from 10 years ago would get twice as many credits as the person who has one top-of-the-line machine that does 40 times as much work. I don't think that's fair to the people who invest in newer computers.

I don't even see a P2 do one result in one hour. Perhaps you want to test with older CPUs, to see how much hours or days they take. :-)

You're answer tommorow ^^

But I bet rather on 1.5 hours


Linux or windows? Big difference

http://cosmos.astro.uiuc.edu/cosmohome/results.php?hostid=66
2 hours and 7 minutes on a linux PII server.

Nota : A FTP server was quite intensively used during the calculation, so calculations could take a little bit less time.
------
Thrr-Gilag Kee'rr

L'Alliance Francophone
ID: 810 · Report as offensive     Reply Quote
Profile Kokomiko
Volunteer tester
Avatar

Send message
Joined: 22 May 07
Posts: 7
Credit: 1,014,912
RAC: 0
Message 923 - Posted: 11 Jul 2007, 12:56:29 UTC

Why I get mostly 0 credits for my WUs on a LINUX-System with AMD64 X2 4800?

http://cosmos.astro.uiuc.edu/cosmohome/results.php?hostid=68

Under this conditions it's make no sense to crunch for Cosmology with this machine.

Also I've get WUs with 41 minutes and get 6 credits per WU on a AMD X2 4200.
http://cosmos.astro.uiuc.edu/cosmohome/workunit.php?wuid=117336
Here it's make no sense to work with this bad windows compilation. I've canceld a dozen of this WUs, while a other machine with 25% of the energy does the same.

So I must say: Don't use Windows for this application, it's a to bad efficiency factor. Also it seems, my Linux machine don't work correct for Cosmology.
ID: 923 · Report as offensive     Reply Quote
Desti

Send message
Joined: 9 Jul 07
Posts: 12
Credit: 279,092
RAC: 0
Message 924 - Posted: 11 Jul 2007, 13:34:30 UTC - in response to Message 923.  

Why I get mostly 0 credits for my WUs on a LINUX-System with AMD64 X2 4800?

http://cosmos.astro.uiuc.edu/cosmohome/results.php?hostid=68

[...]



Most of your results are invalid.

Linux Users Everywhere @ BOINC
ID: 924 · Report as offensive     Reply Quote
Svenie25

Send message
Joined: 21 Jun 07
Posts: 68
Credit: 34,092
RAC: 0
Message 925 - Posted: 11 Jul 2007, 13:35:31 UTC

Can´t agree with your opinion. My E6300 works just fine under Win XP. No WU is longer then 16 minutes. But I think the 6 credits per WU are on the higher limit of credits. I mostly get the credits with my machine. And the few WU´s where I don´t get credits I count to the alphastatus of the project.
ID: 925 · Report as offensive     Reply Quote
Previous · 1 · 2 · 3 · 4 · Next

Forums : Technical Support : Comments on Fixed Credit System