Granting of credits

Message boards : Number crunching : Granting of credits
Message board moderation

To post messages, you must log in.

1 · 2 · 3 · Next

AuthorMessage
GrafZahl

Send message
Joined: 24 Nov 13
Posts: 3
Credit: 2,839,147
RAC: 0
Germany
Message 623 - Posted: 14 Jan 2014, 18:20:10 UTC

These data were recorded for the same PC.
I am wondering why nearly the same credits were granted while the CPU time was so different. In my opinion the credit assignment is not reasonable and must be reconsidered.

    Runtime CPU time Credits
    10,853.04 53,462.04 293.59
    10,934.43 140,330.00 290.45
    11,262.43 12,121.68 301.15
    11,030.82 51,025.87 307.96
    12,664.74 140,320.30 335.34
    13,221.02 53,934.02 360.64

ID: 623 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile yoyo_rkn
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist

Send message
Joined: 22 Aug 11
Posts: 683
Credit: 10,540,830
RAC: 17,648
Germany
Message 624 - Posted: 14 Jan 2014, 19:45:58 UTC - in response to Message 623.  

This is standard Boinc creditnew. So ask berkeley bout it ;)
ID: 624 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Bryan

Send message
Joined: 7 Oct 11
Posts: 1
Credit: 134,015,353
RAC: 0
United States
Message 626 - Posted: 22 Jan 2014, 13:55:43 UTC
Last modified: 22 Jan 2014, 13:57:01 UTC

Ah yes, credit new the random credit generator!

I think the discussion needs to take place here on the project since this is where the folks are choosing whether to crunch the project or not.

Take a look at 3 of the top four machines under Top Computers:


CPU FP Benchmark Time Credit
I7-2760 2842 10427 2511
I7-3770 4093 10367 474
I7-2600 3316 10413 674

Does it seem equitable to you that a mobile 2.4G processor is making 4 to 6 times the credits of high end 3.4/3.5G desktop processors? Granted the I7 3770 system may be also crunching with the GPU which could affect the results however the I7 2600 has no GPU.

It is easier for you to wash your hands of the problem and refer your crunchers to Berkley. You are however losing folks that would crunch the project if the pay was equitable.

You choose to run credit new and I choose to crunch where mobile laptop processors aren't paid 4-6 times the credits for less work accomplished.
ID: 626 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Werinbert

Send message
Joined: 9 May 13
Posts: 10
Credit: 100,312
RAC: 0
United States
Message 665 - Posted: 24 Apr 2014, 2:16:13 UTC

Comparing with another Credit New project...I seem to be getting from Yafu the same credit/hour as from SetiBeta, however the Yafu WU is using 4 cores whereas the SetiBeta Wu is using only 1 core. Yoyo, are you sure that the credits are correctly calculated for multi-core WUs?
ID: 665 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile yoyo_rkn
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist

Send message
Joined: 22 Aug 11
Posts: 683
Credit: 10,540,830
RAC: 17,648
Germany
Message 672 - Posted: 7 Jul 2014, 6:46:53 UTC - in response to Message 665.  

No I'm not sure. I don't really understand how creditnew is working.
Would be good if somebody can run some tests, e.g. on a 8 core system run one boinc instance with 2 cores and one with 6 over some days.

yoyo
ID: 672 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Steve Hawker*

Send message
Joined: 11 Feb 13
Posts: 7
Credit: 129,498
RAC: 0
United States
Message 674 - Posted: 2 Aug 2014, 8:57:51 UTC - in response to Message 672.  

No I'm not sure. I don't really understand how creditnew is working.


Neither does Berkeley. In fact, no-one does.

I no longer crunch here due to CreditNew. I'm sure I'm not alone. I just popped in because credit is the topic du jour and someone mentioned YAFU.

Yoyo, you do so much for the crunching community, why not do what Slicker has done at Collatz and reject CreditNew?
ID: 674 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Coleslaw
Avatar

Send message
Joined: 6 Sep 11
Posts: 31
Credit: 10,303,007
RAC: 99
United States
Message 686 - Posted: 2 Sep 2014, 2:23:46 UTC - in response to Message 672.  
Last modified: 2 Sep 2014, 3:03:51 UTC

8 thread VM with 8GB ram http://yafu.myfirewall.org/yafu/show_host_detail.php?hostid=9420

4 thread VM with 4GB ram http://yafu.myfirewall.org/yafu/show_host_detail.php?hostid=9417

2 thread VM with 2GB ram http://yafu.myfirewall.org/yafu/show_host_detail.php?hostid=9418

1 thread VM with 1GB ram http://yafu.myfirewall.org/yafu/show_host_detail.php?hostid=9415

All four VM's are running the same version of Ubuntu and the same BOINC version that came in the Ubuntu repository. They are also all on the exact same system that has a total of 32 threads (2 CPU's). I am leaving them running for a few weeks to see how they compare on scoring.

First few work units showed:
2 core scoring at 1 point/60 seconds, 4 core scoring 2 points/60 seconds, and 8 core scoring at 3 points/60 seconds. That was based on run time. However, time may adjust that.

Edit: If the single core VM doesn't get any work, I may bump it up to 16 just to see how the points scale beyond 8.
ID: 686 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile yoyo_rkn
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist

Send message
Joined: 22 Aug 11
Posts: 683
Credit: 10,540,830
RAC: 17,648
Germany
Message 687 - Posted: 2 Sep 2014, 7:03:56 UTC

Very good test, appreciate it.
Let it run for a while and we will see how RAC adjusts.
ID: 687 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Coleslaw
Avatar

Send message
Joined: 6 Sep 11
Posts: 31
Credit: 10,303,007
RAC: 99
United States
Message 688 - Posted: 3 Sep 2014, 13:21:05 UTC

Recent work units are showing:
8 core - 1 point every 20 seconds
4 core - 1 point every 10 seconds
2 core - 1 point every 54 seconds

Quad cores are still the sweet spot. I still haven't received work on the single core VM, so I will probably go ahead and bump it up to 16 sometime today. More testing to come.
ID: 688 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Coleslaw
Avatar

Send message
Joined: 6 Sep 11
Posts: 31
Credit: 10,303,007
RAC: 99
United States
Message 689 - Posted: 4 Sep 2014, 0:30:44 UTC

Due to complications I swapped a few things around.

This is now 16 thread: http://yafu.myfirewall.org/yafu/show_host_detail.php?hostid=9418
This is now the 2 thread: http://yafu.myfirewall.org/yafu/show_host_detail.php?hostid=9435

I had to clone the client due to issues with installing a new one. But, this should give us some good numbers for higher than 8.
ID: 689 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Coleslaw
Avatar

Send message
Joined: 6 Sep 11
Posts: 31
Credit: 10,303,007
RAC: 99
United States
Message 690 - Posted: 4 Sep 2014, 13:29:31 UTC

Results today appear to be:

16 thread: 1 point every 3 seconds
2 thread: still working on a work unit for ~8 hours
4 thread: 1 point every 30 seconds
8 thread: 1 point every 17-18 seconds

Will wait and see how the numbers adjust since the 16 thread has only ran over night.
ID: 690 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Coleslaw
Avatar

Send message
Joined: 6 Sep 11
Posts: 31
Credit: 10,303,007
RAC: 99
United States
Message 691 - Posted: 5 Sep 2014, 13:21:43 UTC

Todays results showed:
2 core: 1 point ~ every 60 seconds
4 core: 1 point ~ every 32 seconds
8 core: 1 point ~ every 17 seconds
16 core: 1 point ~ every 13 seconds
ID: 691 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Coleslaw
Avatar

Send message
Joined: 6 Sep 11
Posts: 31
Credit: 10,303,007
RAC: 99
United States
Message 692 - Posted: 7 Sep 2014, 3:18:09 UTC

Todays results showed:
2 core: 1 point ~ every 60 seconds
4 core: 1 point ~ every 30 seconds
8 core: 1 point ~ every 16-17 seconds
16 core: 1 point ~ every 16-17 seconds
ID: 692 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile yoyo_rkn
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist

Send message
Joined: 22 Aug 11
Posts: 683
Credit: 10,540,830
RAC: 17,648
Germany
Message 694 - Posted: 8 Sep 2014, 7:39:42 UTC - in response to Message 692.  

Regarding the credits it looks good for the 2,4 and 8 core systems.
ID: 694 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Werinbert

Send message
Joined: 9 May 13
Posts: 10
Credit: 100,312
RAC: 0
United States
Message 695 - Posted: 8 Sep 2014, 16:19:01 UTC

I am running some on a i5-2410M CPU @ 2.30GHz http://yafu.myfirewall.org/yafu/show_host_detail.php?hostid=6380. I have set the Boinc to 75% and I am running prpnet on the last core. For my returns I am getting about 1cr far every 55 seconds.

Coleslaw, what cpu are you using? It says 2.5MHz in the host details, can you be more specific? For my 3 cores, I am getting about as much credits as you are getting on 2 cores. Is this a difference between the two cpus or is it something troubling with CreditNew...?

* * * * *

I do not use my faster computer (i7-3770k) for any projects using CreditNew as credits per hour is nearly the same as the slower computers (sometimes worse). The limited number of WUs that I did run for this project resulted in about 21.5 cr/hour/cpu (which is 60-70% what SETI gives for CreditNew).
ID: 695 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Coleslaw
Avatar

Send message
Joined: 6 Sep 11
Posts: 31
Credit: 10,303,007
RAC: 99
United States
Message 696 - Posted: 8 Sep 2014, 19:04:46 UTC - in response to Message 695.  

Werinbert, my CPU's on that rig are ES chips and thus I can't give you an exact model. It is somewhere between an e5 2665 and an e5 2670. That rig has 2 of them for a total of 32 threads. Base clock is 2.5, but they boost up to 2.9GHz.

Right now I am just testing to see if the scoring is proportional to the number of cores/threads used. I have no idea how they compare beyond that as far as difference in clock speeds, manufacturer, etc...

Today's numbers are:

2 core: 1 point ~ every 107 seconds
4 core: 1 point ~ every 29-30 seconds
8 core: 1 point ~ every 17 seconds
16 core: 1 point ~ every 17.9-112 seconds
ID: 696 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Werinbert

Send message
Joined: 9 May 13
Posts: 10
Credit: 100,312
RAC: 0
United States
Message 697 - Posted: 9 Sep 2014, 4:21:18 UTC

Cool, I wish I had some ES chips to play around with.

Just ran one task on two cores (the other cores again running prpnet). The result is an ugly 1 point per 230 seconds. And at 106K seconds of runtime, I think this is a no go for my computer. Probably need a larger sample size...maybe later when I am not in a hurry on other projects.
ID: 697 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Coleslaw
Avatar

Send message
Joined: 6 Sep 11
Posts: 31
Credit: 10,303,007
RAC: 99
United States
Message 698 - Posted: 9 Sep 2014, 12:53:19 UTC

Werinbert, how long have you let this system run? CreditNew is designed to adjust itself over time. That is why my numbers keep shifting. My numbers are less than a week running. Many projects don't level their RAC out until you run them a few weeks. CreditNew especially. Make sure your test isn't over only a few days.

On that note, todays numbers are:

2 core: 1 point ~ every 160 seconds
4 core: 1 point ~ every 52-54 seconds
8 core: 1 point ~ every 31 seconds
16 core: 1 point ~ every 117 seconds
ID: 698 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Werinbert

Send message
Joined: 9 May 13
Posts: 10
Credit: 100,312
RAC: 0
United States
Message 699 - Posted: 9 Sep 2014, 19:55:25 UTC

Perhaps that is my problem with Yafu. When I finish my my PRPnet goal of 100K (in about 5 days) I will try and run 3 cores eventually going to 4 cores. Around the end of last year I was crunching WUs just fine. Then after a month or so hiatus I started crunching more WUs but these had a dramatic decrease in credits. That and the longer run times sort of put me off the project. I hope that CreditNew doesn't keep resetting itself whenever someone takes a break from a project.
ID: 699 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Coleslaw
Avatar

Send message
Joined: 6 Sep 11
Posts: 31
Credit: 10,303,007
RAC: 99
United States
Message 700 - Posted: 10 Sep 2014, 13:27:24 UTC
Last modified: 10 Sep 2014, 13:28:57 UTC

todays numbers are:

2 core: 1 point ~ every 135-137 seconds
4 core: 1 point ~ every 58-59 seconds
8 core: 1 point ~ every 39 seconds
16 core: 1 point ~ every 137 seconds

So, over a weeks time things still have not leveled out steady. Here is to week two...
ID: 700 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
1 · 2 · 3 · Next

Message boards : Number crunching : Granting of credits




Datenschutz / Privacy Copyright © 2011-2022 Rechenkraft.net e.V. & yoyo