Advanced search

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

To post messages, you must log in.

1 · 2 · 3 · 4 · Next

AuthorMessage
Profile Jayargh
Volunteer moderator
Volunteer tester
Avatar

Send message
Joined: 25 Jun 07
Posts: 508
Credit: 2,282,158
RAC: 0
Message 688 - Posted: 2 Jul 2007, 16:25:49 UTC

I see as of this posting fixed credits are now a reality ...Thanks Scott

I started this thread here in technical ...not sure if it belongs here but if anyone has any concerns(including myself) about the new system they can voice them here.

So far so good.
ID: 688 · Report as offensive     Reply Quote
Profile Cori
Volunteer tester
Avatar

Send message
Joined: 22 May 07
Posts: 52
Credit: 34,836
RAC: 0
Message 690 - Posted: 2 Jul 2007, 17:17:29 UTC

I finished two of the new WUs with fixed credits. They took about 52 and 56 minutes and got both 5 credits. A bit low even for an old P4 3.6 GHz HT. ;-)

---> http://cosmos.astro.uiuc.edu/cosmohome/results.php?hostid=290
Lovely greetings, Cori

ID: 690 · 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 692 - Posted: 2 Jul 2007, 17:47:29 UTC
Last modified: 2 Jul 2007, 18:12:45 UTC

I think the current fix is a bit low especially due to the variability(up to 100%) of how long a wu takes on any given host.eg: Some take 9mins others take 18min. But to quantify the issue I found this on the Einstein message boards done by Bruce Allen project Scientist/admin from a few months ago:

You can find a cross-project credit comparison here. You will see that almost all projects are in good agreement with SAH. EAH was out of line. We're now fixing that.

(The credit comparison chart was created at my request by James Drews. It provides a solid basis for cross-project credit comparisons.)

Cheers,
Bruce


Scott I hope this can help clarify credit here in Boinc for you a bit more...note;not all projects were included in that study(some major projects missing), and you can see credit granting issues are still encountered by mature projects when they change or modify applications.

ID: 692 · Report as offensive     Reply Quote
Profile Rebirther
Volunteer tester
Avatar

Send message
Joined: 22 May 07
Posts: 23
Credit: 277,369
RAC: 0
Message 694 - Posted: 2 Jul 2007, 18:07:02 UTC - in response to Message 690.  

I finished two of the new WUs with fixed credits. They took about 52 and 56 minutes and got both 5 credits. A bit low even for an old P4 3.6 GHz HT. ;-)

---> http://cosmos.astro.uiuc.edu/cosmohome/results.php?hostid=290


This is very low. 10cr/h is a good cut.
ID: 694 · Report as offensive     Reply Quote
Profile vfrey

Send message
Joined: 24 Jun 07
Posts: 4
Credit: 39,046
RAC: 0
Message 703 - Posted: 2 Jul 2007, 19:53:03 UTC

could this be related to the type of processor? With my Athlon 3800+ the WU was finished in 22 minutes and the second computer on this WU was also an AMD in about the same time. I would think, 5 credits for 22 minutes is a fair value.

http://cosmos.astro.uiuc.edu/cosmohome/workunit.php?wuid=95676
ID: 703 · 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 704 - Posted: 2 Jul 2007, 20:06:34 UTC
Last modified: 2 Jul 2007, 20:13:55 UTC

Just wanted to add based on my sample of wu's run it would seem there is distinctly 2 basic runtimes of the wu's of CAMB 1.18....75-85% take x-time per cpu ...15-25% take about 1/2of x.I am finding less than 5% falling either side of 1 standard deviation from either the x or the y units. Can you see that at your end Scott? My sample is a bit higher than many others can have but still may not be accurate project wide...(this may be why people are getting conflicting opinions based on a 1 or even 10 sample results ...I had about 800 to look at on 4 hosts.

If your findings are similar and can be determined before you send the wu out perhaps you need a multi-tier system as Einstein,Seti, and other fixed credit projects do. Thanks JR

[I'm on vacation,the rains forced me inside so I have too much time on my hands today ;)]
ID: 704 · 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 705 - Posted: 2 Jul 2007, 20:46:23 UTC
Last modified: 2 Jul 2007, 20:47:40 UTC

My 3800x2 usually runs a work unit for around 25-30 minutes each or (1400-1600 cpu seconds). Although there are some that are less at around 600-700 cpu seconds. Perhaps going to a timestep crediting process would work better ? Something like every X amount of cpu time, The user is credited X amount of credit. I would think this would be fair as far as amount of time a person or cpu runs each work unit ?

G^R
Windows-XP-Pro, AMD 3800X2, 5.10.28
ID: 705 · Report as offensive     Reply Quote
Profile Saenger
Volunteer tester
Avatar

Send message
Joined: 22 May 07
Posts: 110
Credit: 353,577
RAC: 0
Message 707 - Posted: 2 Jul 2007, 21:09:41 UTC

For me it's fine, I get double the credit I ask for ;)
I think the WUs are better suited for Linux, so I'm simply faster with them.

I usually ask for 2 - 2.5 credits, now I get 5 all the time.
Grüße vom Sänger
ID: 707 · 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 713 - Posted: 2 Jul 2007, 21:45:11 UTC
Last modified: 2 Jul 2007, 21:51:13 UTC

These are the results for a quad Xeon running 2.88GHz

- Linux 64-bit: 5 credits per WU. WU takes 360-400 seconds. That makes 4300-4800/day on 4 cores. Not as much as RS, but not bad at all.
- Windows: WU takes 800-900 seconds. 1900-2300/day. Ok for Windows, I guess. Not overly generous though.

I'd need to see the average over a longer period to have more accurate results.


EDIT:

The results for my Pentium D 3.0 are a bit disappointing. Running Vista. 5 credits per 2600-2700 seconds, makes 310-330 per day for 2 cores. The machine is worth 450 on average on most projects.

BOINC.BE: For Belgians who love the smell of glowing red cpu's in the morning
Tutta55's Lair
ID: 713 · 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 716 - Posted: 2 Jul 2007, 22:18:19 UTC

I am in the process of turning our legacy app CAMB into a real, bona-fide BOINC app. So, in the end, we should have all the nice features like checkpointing, progress reporting, and accurate credit calculation working.

However, the difficulty is that CAMB is written in FORTRAN 90 and is just a complete pain to work with. Therefore, it will take me a while (i.e on the order of weeks) to finish.

Is there a compromise that would make everybody happy in the mean time or is the fixed 5 credit system OK for now?
Scott Kruger
Project Administrator, Cosmology@Home
ID: 716 · 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 717 - Posted: 2 Jul 2007, 22:19:59 UTC
Last modified: 2 Jul 2007, 22:21:19 UTC

[off-topic]OMG, I am starting to stack messages again. The ATA virus is hitting hard[/off-topic]


Now that there is a fixed credit system, do you still need the quorum of 2? If there is a failsafe way of validating a result, quorum of 1 would do.


EDIT: 7 or 7.5 per WU seems more accurate for my Pentium D.

BOINC.BE: For Belgians who love the smell of glowing red cpu's in the morning
Tutta55's Lair
ID: 717 · 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 718 - Posted: 2 Jul 2007, 22:23:31 UTC - in response to Message 717.  

[off-topic]OMG, I am starting to stack messages again. The ATA virus is hitting hard[/off-topic]


Now that there is a fixed credit system, do you still need the quorum of 2? If there is a failsafe way of validating a result, quorum of 1 would do.

The 2-quorum is in place because the science requires it. If we find some way to validate results without the 2-quorum, I'll implement it right away. For now, though, it's going to have to stay in place. Sorry.
Scott Kruger
Project Administrator, Cosmology@Home
ID: 718 · 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 719 - Posted: 2 Jul 2007, 22:28:35 UTC - in response to Message 716.  
Last modified: 2 Jul 2007, 22:45:00 UTC

I am in the process of turning our legacy app CAMB into a real, bona-fide BOINC app. So, in the end, we should have all the nice features like checkpointing, progress reporting, and accurate credit calculation working.

However, the difficulty is that CAMB is written in FORTRAN 90 and is just a complete pain to work with. Therefore, it will take me a while (i.e on the order of weeks) to finish.

Is there a compromise that would make everybody happy in the mean time or is the fixed 5 credit system OK for now?


That is great news! A few weeks ( er could be longer as we all know because we are all pawns of Murphys'Law) won't make much difference then, however if you look at the stats table in the link on earlier posts here for windows credit (Seti standard) a 6-7 would be more accurate than a 5 ..my 2 cents...but 5 beats .081 anyday :) Normal credit lets us Boincers comparing Boinc totals still be competetive in world/team rankings imho.I think I would rather you putting your time into your own native application rather than fiddling with credits at this point ...progress and checkpoints are highly desired for your empire to grow :)
ID: 719 · Report as offensive     Reply Quote
Profile Cori
Volunteer tester
Avatar

Send message
Joined: 22 May 07
Posts: 52
Credit: 34,836
RAC: 0
Message 723 - Posted: 2 Jul 2007, 22:48:35 UTC - in response to Message 717.  

... EDIT: 7 or 7.5 per WU seems more accurate for my Pentium D.

My P4's would like that as well. ;-)
Lovely greetings, Cori

ID: 723 · Report as offensive     Reply Quote
Profile Saenger
Volunteer tester
Avatar

Send message
Joined: 22 May 07
Posts: 110
Credit: 353,577
RAC: 0
Message 725 - Posted: 2 Jul 2007, 23:01:05 UTC - in response to Message 723.  

... EDIT: 7 or 7.5 per WU seems more accurate for my Pentium D.

My P4's would like that as well. ;-)

I would like that even more, would give me more then 20 C/h, while I usually get about 10 C/h.
Grüße vom Sänger
ID: 725 · 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 726 - Posted: 2 Jul 2007, 23:16:46 UTC

I'm fine with what ever, 5 or 6 or 7..
ID: 726 · 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 728 - Posted: 3 Jul 2007, 0:14:12 UTC

5 is alot better than the occasional .90 I was getting for some wu's. I guess at the hopes that the 5.5.0 doesnt make an apperence again, You could always go with the higher of the two quorum instead of the lower, At least the slower pc's would get their full credit, And the faster one's would get the bonus. But waiting a few weeks is fine by me eitherway. Thanks !

G^R
Windows-XP-Pro, AMD 3800X2, 5.10.28
ID: 728 · Report as offensive     Reply Quote
Profile Misfit
Volunteer tester
Avatar

Send message
Joined: 9 Jun 07
Posts: 150
Credit: 237,789
RAC: 0
Message 729 - Posted: 3 Jul 2007, 0:53:10 UTC - in response to Message 728.  

I guess at the hopes that the 5.5.0 doesnt make an apperence again,

G^R

It can't. It's below the project minimum version. Of course there is the 5.9.0 replacement.
me@rescam.org
ID: 729 · 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 730 - Posted: 3 Jul 2007, 2:42:29 UTC - in response to Message 729.  

I guess at the hopes that the 5.5.0 doesnt make an apperence again,

G^R

It can't. It's below the project minimum version. Of course there is the 5.9.0 replacement.


Does it also overclaim as 5.5.0 does ?
Windows-XP-Pro, AMD 3800X2, 5.10.28
ID: 730 · Report as offensive     Reply Quote
Profile Saenger
Volunteer tester
Avatar

Send message
Joined: 22 May 07
Posts: 110
Credit: 353,577
RAC: 0
Message 732 - Posted: 3 Jul 2007, 4:41:04 UTC
Last modified: 3 Jul 2007, 4:42:17 UTC

You are giving far too much credits per WU for me. I usually get about 10 credits per houir, here I have on average 17.76 for the last 26 WUs, the best one yet has even 41.43 c/h, 4 times of what I "deserve".

I see on the other hand that the majotity seems to complain even about this high grants, so the application seems to be either extremely well suited for Linux or rather sloppy for Windows.

I don't know what's better for the future: to give the majority the fitting credits (i.e. what's granted yet, perhaps even more), although the application doesn't seem to run well on it or to give the well runing Linux the fitting credits (i.e. half of what's granted yet), and improve the Windows application to run as good as Linux and make clear that a lot of cycles seem to be wasted on Windows for now by granting less.
Grüße vom Sänger
ID: 732 · Report as offensive     Reply Quote
1 · 2 · 3 · 4 · Next

Forums : Technical Support : Comments on Fixed Credit System