AP credit, can someone expain this?

Message boards : Number crunching : AP credit, can someone expain this?
Message board moderation

To post messages, you must log in.

1 · 2 · Next

AuthorMessage
Profile Oz
Avatar

Send message
Joined: 6 Jun 99
Posts: 233
Credit: 200,655,462
RAC: 212
United States
Message 1544839 - Posted: 20 Jul 2014, 12:19:18 UTC
Last modified: 20 Jul 2014, 12:33:46 UTC

Wow, take a look at 3636714988 and 3636594228 - returned within 8 hours of each other, both with 0.00% blanking, both ran in a little under 6k seconds with about 2k CPU time but one yielded 312 credits and one yielded 832. How can one possibly do that? A random number generator? Magjic 8-ball? There are many other examples on my machines which also include tasks with similar credit but 200-300% differences in run time...
Flop count is used for verification, right? Why isn't credit equal to flops divided by a constant?
Member of the 20 Year Club



ID: 1544839 · Report as offensive
kittyman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 9 Jul 00
Posts: 51468
Credit: 1,018,363,574
RAC: 1,004
United States
Message 1544859 - Posted: 20 Jul 2014, 13:22:25 UTC

Or as some of us fondly (NOT) call it, CreditScrewed.
"Freedom is just Chaos, with better lighting." Alan Dean Foster

ID: 1544859 · Report as offensive
Darth Beaver Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Avatar

Send message
Joined: 20 Aug 99
Posts: 6728
Credit: 21,443,075
RAC: 3
Australia
Message 1544889 - Posted: 20 Jul 2014, 14:20:26 UTC - in response to Message 1544859.  

Or as some of us fondly (NOT) call it, CreditScrewed.


+1
ID: 1544889 · Report as offensive
Profile Oz
Avatar

Send message
Joined: 6 Jun 99
Posts: 233
Credit: 200,655,462
RAC: 212
United States
Message 1545036 - Posted: 20 Jul 2014, 20:45:54 UTC - in response to Message 1544857.  

Wow, take a look at 3636714988 and 3636594228 - returned within 8 hours of each other, both with 0.00% blanking, both ran in a little under 6k seconds with about 2k CPU time but one yielded 312 credits and one yielded 832. How can one possibly do that? A random number generator? Magjic 8-ball? There are many other examples on my machines which also include tasks with similar credit but 200-300% differences in run time...
Flop count is used for verification, right? Why isn't credit equal to flops divided by a constant?


The answer is indeed a random number generator, called CreditNew.

Read about it here: http://boinc.berkeley.edu/trac/wiki/CreditNew


Ok, I read it, it obviously does not work as stated, if it did, similar jobs would only vary a few percent on the same machine - I could even see a 60-150% range to cover angles and that sort of thing. I could get more consistent credit to work numbers throwing darts blindfolded (slight hyperbole) - I mean it worked more or less okay when they cut the credit in half but now it's just gone mad...
Member of the 20 Year Club



ID: 1545036 · Report as offensive
kittyman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 9 Jul 00
Posts: 51468
Credit: 1,018,363,574
RAC: 1,004
United States
Message 1545040 - Posted: 20 Jul 2014, 20:49:49 UTC - in response to Message 1545036.  



Ok, I read it, it obviously does not work as stated, if it did, similar jobs would only vary a few percent on the same machine - I could even see a 60-150% range to cover angles and that sort of thing. I could get more consistent credit to work numbers throwing darts blindfolded (slight hyperbole) - I mean it worked more or less okay when they cut the credit in half but now it's just gone mad...

The kitties just keep throwing lots of darts.....LOL.
"Freedom is just Chaos, with better lighting." Alan Dean Foster

ID: 1545040 · Report as offensive
Profile Oz
Avatar

Send message
Joined: 6 Jun 99
Posts: 233
Credit: 200,655,462
RAC: 212
United States
Message 1545049 - Posted: 20 Jul 2014, 21:08:02 UTC

Just so I am clear on this, the server guesses the performance of the system and if it guesses your system is slow, gives more credit - unless it can't decide - then it uses actual flops, yes?

If that's it, it doesn't work - the RAC on my 2.0 GHz dual core, GTX 660 (2GB ram) is constantly higher than my 2.2 GHz dual core, GTX 770 (4GB ram) and as Yosemite Sam so succinctly puts it, "that just ain't right..."
Member of the 20 Year Club



ID: 1545049 · Report as offensive
Profile Oz
Avatar

Send message
Joined: 6 Jun 99
Posts: 233
Credit: 200,655,462
RAC: 212
United States
Message 1545051 - Posted: 20 Jul 2014, 21:15:39 UTC

AAAaaaaaaaaaaaaaaaaaarrrrrrrrrrrrrrrrrrgggggggg!!!!

Thanx for letting me whine...
Member of the 20 Year Club



ID: 1545051 · Report as offensive
Profile BilBg
Volunteer tester
Avatar

Send message
Joined: 27 May 07
Posts: 3720
Credit: 9,385,827
RAC: 0
Bulgaria
Message 1545068 - Posted: 20 Jul 2014, 22:35:23 UTC - in response to Message 1545036.  

Read about it here: http://boinc.berkeley.edu/trac/wiki/CreditNew

Ok, I read it, it obviously does not work as stated, if it did, similar jobs would only vary a few percent on the same machine ...


jason_gee seems to know how to fix it (in the code), the hard part is to convince Dr. A. that there exist a problem with CreditNew

Observation of CreditNew Impact (4)
http://setiathome.berkeley.edu/forum_thread.php?id=73306

"the four known credit System design flaws currently under detailed analysis":
http://setiathome.berkeley.edu/forum_thread.php?id=73935&postid=1498621#1498621
 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 
ID: 1545068 · Report as offensive
juan BFP Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 16 Mar 07
Posts: 9786
Credit: 572,710,851
RAC: 3,799
Panama
Message 1545074 - Posted: 20 Jul 2014, 22:57:05 UTC
Last modified: 20 Jul 2014, 23:01:18 UTC

Actualy they are testing a new builds at albert@home to try to fix that but they are in the beginng of the hard task.
ID: 1545074 · Report as offensive
Price Family - Brighton

Send message
Joined: 4 Nov 00
Posts: 42
Credit: 48,039,168
RAC: 233
United Kingdom
Message 1545231 - Posted: 21 Jul 2014, 9:38:41 UTC

The title of this thread seems to refer to Astropulse credit. I didn't think that was affected by credit new.
ID: 1545231 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1545241 - Posted: 21 Jul 2014, 10:00:52 UTC - in response to Message 1545231.  

The title of this thread seems to refer to Astropulse credit. I didn't think that was affected by credit new.

All credit awarded by the SETI project - both for Multibeam tasks and for Astropulse tasks - is determined by CreditNew, and has been for about four years.
ID: 1545241 · Report as offensive
Darth Beaver Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Avatar

Send message
Joined: 20 Aug 99
Posts: 6728
Credit: 21,443,075
RAC: 3
Australia
Message 1545242 - Posted: 21 Jul 2014, 10:07:26 UTC - in response to Message 1545040.  

The kitties just keep throwing lots of darts.....LOL.


me and the Kitty man are throwing darts big 1's with explosive tips
ID: 1545242 · Report as offensive
Profile Spiked Punch
Avatar

Send message
Joined: 8 Dec 01
Posts: 14
Credit: 48,927,213
RAC: 0
United States
Message 1545254 - Posted: 21 Jul 2014, 11:40:14 UTC - in response to Message 1544889.  

Or as some of us fondly (NOT) call it, CreditScrewed.


+1


+ Rndm#

Just keeping in the spirit of things.
ID: 1545254 · Report as offensive
Profile HAL9000
Volunteer tester
Avatar

Send message
Joined: 11 Sep 99
Posts: 6534
Credit: 196,805,888
RAC: 57
United States
Message 1545288 - Posted: 21 Jul 2014, 13:22:00 UTC - in response to Message 1545231.  

The title of this thread seems to refer to Astropulse credit. I didn't think that was affected by credit new.

When Astropulse v7 is released the screams will begin a new I expect.
SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[
ID: 1545288 · Report as offensive
Profile Oz
Avatar

Send message
Joined: 6 Jun 99
Posts: 233
Credit: 200,655,462
RAC: 212
United States
Message 1545295 - Posted: 21 Jul 2014, 13:40:48 UTC - in response to Message 1545288.  
Last modified: 21 Jul 2014, 13:43:11 UTC

When Astropulse v7 is released the screams will begin a new I expect.


Astropulse V.... 7... ???!!!!!

Ooooooooooooooh noooooooooooooo!
Member of the 20 Year Club



ID: 1545295 · Report as offensive
juan BFP Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 16 Mar 07
Posts: 9786
Credit: 572,710,851
RAC: 3,799
Panama
Message 1545303 - Posted: 21 Jul 2014, 13:54:59 UTC - in response to Message 1545288.  
Last modified: 21 Jul 2014, 13:56:54 UTC

When Astropulse v7 is released the screams will begin a new I expect.

That´s one of the paradox of creditscrew, more optimized your apps is, less credit we gain. We humans expect exactly the opositive.

Unless Jason´s and his team fix creditscrew before the lounch of AP 7, we all could expect a big fall in the credit paid by the AP work. But at least we could have a gain, the balance AP vs MB paid could be restored or reduced to a more friendly numbers.

Fingers crossed. BTW I still hate creditscrew.
ID: 1545303 · Report as offensive
Cosmic_Ocean
Avatar

Send message
Joined: 23 Dec 00
Posts: 3027
Credit: 13,516,867
RAC: 13
United States
Message 1545409 - Posted: 21 Jul 2014, 18:09:11 UTC - in response to Message 1545231.  

The title of this thread seems to refer to Astropulse credit. I didn't think that was affected by credit new.

It is. I kept a spreadsheet of all the AP tasks I have been given since CreditNew began (I remember the days of APs being awarded 1294.84 no matter what).

My observations of it is that it seems to generally hover in the 650-750 range about 90% of the time, but then that other 10% of the time.. it can range from as low as 10 to as high as several thousand with no rhyme or reason behind it.

Interestingly, maybe it was just a total coincidence, but occasionally, I would be awarded 3-400 on one task, and then a few tasks later I would be awarded 900-1100 to offset the low one.

That was all on CPU crunching and had a couple thousand tasks of consistency to help smooth things out. From my understanding, the large variations in granted credit seem to happen more frequently on GPUs where the run times can vary in much larger margins than on a CPU alone.
Linux laptop:
record uptime: 1511d 20h 19m (ended due to the power brick giving-up)
ID: 1545409 · Report as offensive
Profile James Sotherden
Avatar

Send message
Joined: 16 May 99
Posts: 10436
Credit: 110,373,059
RAC: 54
United States
Message 1545688 - Posted: 22 Jul 2014, 6:32:07 UTC

I had an AP pay me 9 credits one time.
[/quote]

Old James
ID: 1545688 · Report as offensive
Darth Beaver Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Avatar

Send message
Joined: 20 Aug 99
Posts: 6728
Credit: 21,443,075
RAC: 3
Australia
Message 1545726 - Posted: 22 Jul 2014, 7:34:53 UTC - in response to Message 1545688.  

James ask Mark if he has any spear darts for you i live to far away sorry
ID: 1545726 · Report as offensive
kittyman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 9 Jul 00
Posts: 51468
Credit: 1,018,363,574
RAC: 1,004
United States
Message 1545729 - Posted: 22 Jul 2014, 7:46:57 UTC - in response to Message 1545688.  

I had an AP pay me 9 credits one time.

I don't recall if it was MB or AP, but I do recall stories about NEGATIVE credits at one time....it was long ago.
"Freedom is just Chaos, with better lighting." Alan Dean Foster

ID: 1545729 · Report as offensive
1 · 2 · Next

Message boards : Number crunching : AP credit, can someone expain this?


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.