Advanced search

Message boards : News : [wildlife] EXACT app credit update 2

Author Message
Travis Desell
Volunteer moderator
Project administrator
Project developer
Project scientist
Send message
Joined: 16 Jan 12
Posts: 1787
Combined Credit: 2,265,607
DNA@Home: 293,563
SubsetSum@Home: 349,212
Wildlife@Home: 1,622,832
Wildlife@Home Watched: 212,926s
Wildlife@Home Events: 51
Climate Tweets: 21
Images Observed: 698

              
Message 6732 - Posted: 1 Feb 2017, 22:22:28 UTC

Looks like I may have been off by a couple orders of magnitude with the new credit, so congratulations to any of you who got some of those lucky workunits. I've toned it down a bit, but I think the new credit should be more in line with how long the workunits are taking, along with a bit of a bonus for the project still being in alpha stages.

As things start rolling in, let me know how the new credit is looking.

Chris Skull
Avatar
Send message
Joined: 11 Apr 15
Posts: 20
Combined Credit: 4,766,966
DNA@Home: 55,861
SubsetSum@Home: 1,272,523
Wildlife@Home: 3,438,582
Wildlife@Home Watched: 1,312,789s
Wildlife@Home Events: 475
Climate Tweets: 0
Images Observed: 46

            
Message 6760 - Posted: 19 Feb 2017, 17:22:37 UTC - in response to Message 6732.

credit still much to high...
____________
Greetz
Chris

Travis Desell
Volunteer moderator
Project administrator
Project developer
Project scientist
Send message
Joined: 16 Jan 12
Posts: 1787
Combined Credit: 2,265,607
DNA@Home: 293,563
SubsetSum@Home: 349,212
Wildlife@Home: 1,622,832
Wildlife@Home Watched: 212,926s
Wildlife@Home Events: 51
Climate Tweets: 21
Images Observed: 698

              
Message 6761 - Posted: 20 Feb 2017, 19:29:59 UTC - in response to Message 6760.

credit still much to high...


If anyone could provide me some info as to about where CSG is comparing to other projects so I can do an adjustment that would be awesome. Not quite sure how much I should be dropping by.

Chris Skull
Avatar
Send message
Joined: 11 Apr 15
Posts: 20
Combined Credit: 4,766,966
DNA@Home: 55,861
SubsetSum@Home: 1,272,523
Wildlife@Home: 3,438,582
Wildlife@Home Watched: 1,312,789s
Wildlife@Home Events: 475
Climate Tweets: 0
Images Observed: 46

            
Message 6765 - Posted: 22 Feb 2017, 11:31:25 UTC

For example at Atlas@Home i get for 35.000 sec.CPU time around 300 credit... Wildlife gives 5.500 credits... but isn't there a formula for Boinc credits ?
____________
Greetz
Chris

Profile JumpinJohnny
Avatar
Send message
Joined: 24 Sep 13
Posts: 235
Combined Credit: 6,927,594
DNA@Home: 192,548
SubsetSum@Home: 201,740
Wildlife@Home: 6,533,307
Wildlife@Home Watched: 55,997,833s
Wildlife@Home Events: 15,584
Climate Tweets: 312
Images Observed: 351

              
Message 6766 - Posted: 22 Feb 2017, 16:26:44 UTC

Both SETI@Home and Atlas@Home are on the very LOW end of the credits spectrum. If you want to maintain users interest and participation, then these examples should be avoided. Both are very long term projects with huge user bases.

It seems to me that the Credits should reflect the work actually done rather than the cpu time. That way when GPU apps are developed, the work they do can be reflected in corresponding credits.

There are also bonuses given by a few projects for various reasons and you were right to give one for a project restart with unproven work. Some projects give bonuses for "long run" work that has WU's that takes many days on an average CPU.

If you are planning to cut the credits to be more in line with other BOINC projects I would suggest caution in cutting more than 50% unless you are fairly certain that you have a LONG TERM need and the necessary number of users to make a substantial base to do the required computing power.
Cutting credits to only 10% of current WILL result in loss of interest and active users, even among those who say they "crunch for science".

Just my opinion.
____________

StyM
Send message
Joined: 18 Apr 16
Posts: 1
Combined Credit: 18,045,365
DNA@Home: 17,238
SubsetSum@Home: 1,486,000
Wildlife@Home: 16,542,126
Wildlife@Home Watched: 0s
Wildlife@Home Events: 0
Climate Tweets: 0
Images Observed: 0

      
Message 6768 - Posted: 24 Feb 2017, 2:01:45 UTC

credits are fine no need to change. :)

Profile Gibson Praise
Avatar
Send message
Joined: 5 Aug 16
Posts: 52
Combined Credit: 21,950,527
DNA@Home: 0
SubsetSum@Home: 73,256
Wildlife@Home: 21,877,271
Wildlife@Home Watched: 0s
Wildlife@Home Events: 0
Climate Tweets: 767
Images Observed: 23,280

        
Message 6769 - Posted: 25 Feb 2017, 3:49:40 UTC - in response to Message 6761.


If anyone could provide me some info as to about where CSG is comparing to other projects so I can do an adjustment that would be awesome. Not quite sure how much I should be dropping by.


Notes: measurements in terms of # seconds to produce a BOINC point (seconds for unit divided by points granted). Asteroids is noted as a high point project, SRBase is also on the generous side. WCG, like Atlas and seti is noted as being very stingy on points. All measurements based on recently run wu

System Spec: Xeon E5-2650 @2.1 GHz running Win10

EXACT: running in between 17 - 21 sec/pt
Asteroids: pretty steady at 19-20 sec/pt
SRBase: (highly variable) 20-40 sec/pt
tn-grid : 36-40 sec/pt
WCG: 235 sec/pt

I hope this information helps. The EXACT app is currently at the high end of the spectrum for points given (at least for my system).

Personally, I am not in favor of any great modification to the point valuation

g
____________

Travis Desell
Volunteer moderator
Project administrator
Project developer
Project scientist
Send message
Joined: 16 Jan 12
Posts: 1787
Combined Credit: 2,265,607
DNA@Home: 293,563
SubsetSum@Home: 349,212
Wildlife@Home: 1,622,832
Wildlife@Home Watched: 212,926s
Wildlife@Home Events: 51
Climate Tweets: 21
Images Observed: 698

              
Message 6775 - Posted: 1 Mar 2017, 17:09:25 UTC - in response to Message 6766.

Both SETI@Home and Atlas@Home are on the very LOW end of the credits spectrum. If you want to maintain users interest and participation, then these examples should be avoided. Both are very long term projects with huge user bases.

It seems to me that the Credits should reflect the work actually done rather than the cpu time. That way when GPU apps are developed, the work they do can be reflected in corresponding credits.

There are also bonuses given by a few projects for various reasons and you were right to give one for a project restart with unproven work. Some projects give bonuses for "long run" work that has WU's that takes many days on an average CPU.

If you are planning to cut the credits to be more in line with other BOINC projects I would suggest caution in cutting more than 50% unless you are fairly certain that you have a LONG TERM need and the necessary number of users to make a substantial base to do the required computing power.
Cutting credits to only 10% of current WILL result in loss of interest and active users, even among those who say they "crunch for science".

Just my opinion.


So right now I have a major update in the works, but I am waiting for the current 2 week challenge to finish up before implementing them because that would mess with the challenge.

I'm going to hold things steady until the challenge is over, then upload the new app and start tweaking credit rates.

Travis Desell
Volunteer moderator
Project administrator
Project developer
Project scientist
Send message
Joined: 16 Jan 12
Posts: 1787
Combined Credit: 2,265,607
DNA@Home: 293,563
SubsetSum@Home: 349,212
Wildlife@Home: 1,622,832
Wildlife@Home Watched: 212,926s
Wildlife@Home Events: 51
Climate Tweets: 21
Images Observed: 698

              
Message 6776 - Posted: 1 Mar 2017, 17:11:34 UTC - in response to Message 6769.


If anyone could provide me some info as to about where CSG is comparing to other projects so I can do an adjustment that would be awesome. Not quite sure how much I should be dropping by.


Notes: measurements in terms of # seconds to produce a BOINC point (seconds for unit divided by points granted). Asteroids is noted as a high point project, SRBase is also on the generous side. WCG, like Atlas and seti is noted as being very stingy on points. All measurements based on recently run wu

System Spec: Xeon E5-2650 @2.1 GHz running Win10

EXACT: running in between 17 - 21 sec/pt
Asteroids: pretty steady at 19-20 sec/pt
SRBase: (highly variable) 20-40 sec/pt
tn-grid : 36-40 sec/pt
WCG: 235 sec/pt

I hope this information helps. The EXACT app is currently at the high end of the spectrum for points given (at least for my system).

Personally, I am not in favor of any great modification to the point valuation

g


Thanks for the info! I'm not planning to make major change but I do want to make sure that we're not ridiculously out of line as compared to other projects.

Travis Desell
Volunteer moderator
Project administrator
Project developer
Project scientist
Send message
Joined: 16 Jan 12
Posts: 1787
Combined Credit: 2,265,607
DNA@Home: 293,563
SubsetSum@Home: 349,212
Wildlife@Home: 1,622,832
Wildlife@Home Watched: 212,926s
Wildlife@Home Events: 51
Climate Tweets: 21
Images Observed: 698

              
Message 6777 - Posted: 1 Mar 2017, 17:58:22 UTC - in response to Message 6765.

For example at Atlas@Home i get for 35.000 sec.CPU time around 300 credit... Wildlife gives 5.500 credits... but isn't there a formula for Boinc credits ?


So you can have BOINC try and figure out the credit for a workunit, however it's really not the best algorithm (and tends to award very low). It pretty much becomes based on computation time, as opposed to how much work is being actually done.

For example, I could send out a workunit, and someone may have put in a bunch of compiler optimizations on their version of an application (which can be done as our apps are open source). So that optimized version may run 2x faster than the non-optimized one, however it's not going to get 2x credit even though it's doing the same amount of work.

The way around that is to use fixed credit (which is what I use), as I know how much "work" is being done, but the conversion from "how much work" to how many FLOPS are being done/credit awarded is a very grey area to say the least, as every processor, operating system, etc does things a bit differently.

Anyways, this is probably one of the grand unsolvable problems of BOINC. :) Coming up with a "fair" credit scheme.

Profile Daniel
Send message
Joined: 23 Sep 16
Posts: 27
Combined Credit: 7,338,246
DNA@Home: 0
SubsetSum@Home: 3,381
Wildlife@Home: 7,334,865
Wildlife@Home Watched: 0s
Wildlife@Home Events: 0
Climate Tweets: 156
Images Observed: 18

      
Message 6786 - Posted: 4 Mar 2017, 10:25:21 UTC - in response to Message 6769.


If anyone could provide me some info as to about where CSG is comparing to other projects so I can do an adjustment that would be awesome. Not quite sure how much I should be dropping by.


Notes: measurements in terms of # seconds to produce a BOINC point (seconds for unit divided by points granted). Asteroids is noted as a high point project, SRBase is also on the generous side. WCG, like Atlas and seti is noted as being very stingy on points. All measurements based on recently run wu

System Spec: Xeon E5-2650 @2.1 GHz running Win10

EXACT: running in between 17 - 21 sec/pt
Asteroids: pretty steady at 19-20 sec/pt
SRBase: (highly variable) 20-40 sec/pt
tn-grid : 36-40 sec/pt
WCG: 235 sec/pt

I hope this information helps. The EXACT app is currently at the high end of the spectrum for points given (at least for my system).

Personally, I am not in favor of any great modification to the point valuation

g

Gibson, are these numbers for single app running, or sum for apps running on all your CPU cores? And do you run 16 WUs at once, or 8 with HyperThreading off? I am asking because in my impression Asteroids pays low for their GPU app, and I wonder if it would be better to switch to CPU app there.
____________

Profile Conan
Avatar
Send message
Joined: 13 Apr 12
Posts: 137
Combined Credit: 19,838,171
DNA@Home: 399,792
SubsetSum@Home: 1,448,876
Wildlife@Home: 17,989,503
Wildlife@Home Watched: 70,910s
Wildlife@Home Events: 0
Climate Tweets: 393
Images Observed: 0

          
Message 6803 - Posted: 12 Mar 2017, 11:37:04 UTC - in response to Message 6769.
Last modified: 12 Mar 2017, 12:01:28 UTC


If anyone could provide me some info as to about where CSG is comparing to other projects so I can do an adjustment that would be awesome. Not quite sure how much I should be dropping by.


Notes: measurements in terms of # seconds to produce a BOINC point (seconds for unit divided by points granted). Asteroids is noted as a high point project, SRBase is also on the generous side. WCG, like Atlas and seti is noted as being very stingy on points. All measurements based on recently run wu

System Spec: Xeon E5-2650 @2.1 GHz running Win10

EXACT: running in between 17 - 21 sec/pt
Asteroids: pretty steady at 19-20 sec/pt
SRBase: (highly variable) 20-40 sec/pt
tn-grid : 36-40 sec/pt
WCG: 235 sec/pt

I hope this information helps. The EXACT app is currently at the high end of the spectrum for points given (at least for my system).

Personally, I am not in favor of any great modification to the point valuation

g


I looked at a number of the projects that I am running and the variation in awarded credit is amazing.
I run older computers such AMD Phenoms, a XEON server @2.53 GHz (with hyperthreading on) and an I5, all running a mixture of Linux and Windows.

CSG--------------------7 to 21 sec/pt
Amicable Numbers--19 to 34 sec/pt
Primegrid--------------19 to 53 sec/pt
SRBase----------------25 to 44 sec/pt
Asteroids--------------35 to 38 sec/pt (only records are for the XEON and hyperthreading)
Universe---------------30 to 50 sec/pt
POGS------------------34 to 57 sec/pt
Einstein----------------43 to 52 sec/pt
SETI--------------------57 to 141 sec/pt
Rosetta----------------71 to 111 sec/pt
VGTU------------------77 to 221 sec/pt
WEP2------------------92 to 201 sec/pt
Leiden Classical-----96 to 209 sec/pt
LHC--------------------94 to 157 sec/pt
Denis@Home--------109 to 205 sec/pt
WCG-------------------122 to 200 sec/pt
MindModeling--------162 to 174 sec/pt (only data is from XEON with hyperthreading on)

On average the lowest scores for SETI and POGS are from the Intel I5.
Mostly the highest scores are from the XEON due to Hyperthreading being on.
The Asteroids score would be a lot lower if I had data from my other computers.

So some big differences in awarded credit across the various projects.

CSGrid is a bit higher than most but is only just in front of Primegrid, SRBase, Asteroids and Amicable Numbers.
Pretty sad awarded credit from some of the projects.

So maybe a bit of a trim to amount of awarded credit but it does not have to be too drastic as some others are not that far behind.

Conan
____________

Profile Gibson Praise
Avatar
Send message
Joined: 5 Aug 16
Posts: 52
Combined Credit: 21,950,527
DNA@Home: 0
SubsetSum@Home: 73,256
Wildlife@Home: 21,877,271
Wildlife@Home Watched: 0s
Wildlife@Home Events: 0
Climate Tweets: 767
Images Observed: 23,280

        
Message 6805 - Posted: 14 Mar 2017, 0:33:24 UTC - in response to Message 6786.
Last modified: 14 Mar 2017, 0:55:52 UTC



System Spec: Xeon E5-2650 @2.1 GHz running Win10

EXACT: running in between 17 - 21 sec/pt
Asteroids: pretty steady at 19-20 sec/pt
SRBase: (highly variable) 20-40 sec/pt
tn-grid : 36-40 sec/pt
WCG: 235 sec/pt

g

Gibson, are these numbers for single app running, or sum for apps running on all your CPU cores? And do you run 16 WUs at once, or 8 with HyperThreading off? I am asking because in my impression Asteroids pays low for their GPU app, and I wonder if it would be better to switch to CPU app there.


Just to update, this is a dual xeon system. HyperThreading is on (32 potential cores). When I took these measurements, 25 cores went to CPU projects, a single GPU project was running (collatz or GPUGrid). System cpu usage from taskmanger was ~90%. 9 ran SRBase, 11 ran EXACT and I used the other five to run the other projects to get a recent number for them. (WCG was the Smash Children's Cancer project).

edit - the figures should be considered rather rough. I gathered 8-10 results and averaged. For EXACT I chose across a range of run times. /edit

As far as Asteroids GPU units, I have not run those in a couple of years so I really can't comment. When GPU units first came out it wasn't the best use of your GPU -- the Asteroid algorithm did not transfer well to GPU architechtures and so had low points compared to other GPU projects. I will note that GPU points given vary almost as widely between projects as CPU points do.

@Conan - thanks for those figures!
____________

mmonnin
Send message
Joined: 31 May 16
Posts: 24
Combined Credit: 10,868,351
DNA@Home: 0
SubsetSum@Home: 1,023,200
Wildlife@Home: 9,845,150
Wildlife@Home Watched: 0s
Wildlife@Home Events: 0
Climate Tweets: 54
Images Observed: 0

      
Message 6806 - Posted: 14 Mar 2017, 20:05:33 UTC

Asteroids GPU work is low compared to other GPU projects and a 2P system can keep up with a GPU. That's not usually the case with other projects.

IMO the points should just be compared intra project between apps, not between projects. Too many factors and its never the same work. GPU RAC is all over the place between projects.

Travis Desell
Volunteer moderator
Project administrator
Project developer
Project scientist
Send message
Joined: 16 Jan 12
Posts: 1787
Combined Credit: 2,265,607
DNA@Home: 293,563
SubsetSum@Home: 349,212
Wildlife@Home: 1,622,832
Wildlife@Home Watched: 212,926s
Wildlife@Home Events: 51
Climate Tweets: 21
Images Observed: 698

              
Message 6807 - Posted: 14 Mar 2017, 22:58:14 UTC - in response to Message 6803.


If anyone could provide me some info as to about where CSG is comparing to other projects so I can do an adjustment that would be awesome. Not quite sure how much I should be dropping by.


Notes: measurements in terms of # seconds to produce a BOINC point (seconds for unit divided by points granted). Asteroids is noted as a high point project, SRBase is also on the generous side. WCG, like Atlas and seti is noted as being very stingy on points. All measurements based on recently run wu

System Spec: Xeon E5-2650 @2.1 GHz running Win10

EXACT: running in between 17 - 21 sec/pt
Asteroids: pretty steady at 19-20 sec/pt
SRBase: (highly variable) 20-40 sec/pt
tn-grid : 36-40 sec/pt
WCG: 235 sec/pt

I hope this information helps. The EXACT app is currently at the high end of the spectrum for points given (at least for my system).

Personally, I am not in favor of any great modification to the point valuation

g


I looked at a number of the projects that I am running and the variation in awarded credit is amazing.
I run older computers such AMD Phenoms, a XEON server @2.53 GHz (with hyperthreading on) and an I5, all running a mixture of Linux and Windows.

CSG--------------------7 to 21 sec/pt
Amicable Numbers--19 to 34 sec/pt
Primegrid--------------19 to 53 sec/pt
SRBase----------------25 to 44 sec/pt
Asteroids--------------35 to 38 sec/pt (only records are for the XEON and hyperthreading)
Universe---------------30 to 50 sec/pt
POGS------------------34 to 57 sec/pt
Einstein----------------43 to 52 sec/pt
SETI--------------------57 to 141 sec/pt
Rosetta----------------71 to 111 sec/pt
VGTU------------------77 to 221 sec/pt
WEP2------------------92 to 201 sec/pt
Leiden Classical-----96 to 209 sec/pt
LHC--------------------94 to 157 sec/pt
Denis@Home--------109 to 205 sec/pt
WCG-------------------122 to 200 sec/pt
MindModeling--------162 to 174 sec/pt (only data is from XEON with hyperthreading on)

On average the lowest scores for SETI and POGS are from the Intel I5.
Mostly the highest scores are from the XEON due to Hyperthreading being on.
The Asteroids score would be a lot lower if I had data from my other computers.

So some big differences in awarded credit across the various projects.

CSGrid is a bit higher than most but is only just in front of Primegrid, SRBase, Asteroids and Amicable Numbers.
Pretty sad awarded credit from some of the projects.

So maybe a bit of a trim to amount of awarded credit but it does not have to be too drastic as some others are not that far behind.

Conan


Awesome! Thanks for this. I actually have a large update to roll out, and I think for the time being I'll hold credit fixed as the update will increase the compute time of workunits -- that may be enough to bring us more in line with other projects. If not I can drop a bit more.

Still have a few more things to implement before we have a "stable" implementation, so I'm happy to give people a bit of a bonus while I'm testing things as well...


Post to thread

Message boards : News : [wildlife] EXACT app credit update 2