My last statement about getting less short Wu's dropping granted credit has come true.

Your completion times seem unusually long, more in line with an Opteron at 1.8 ghz vs stock of 2.6 on your Opteron 285. I would take a close look at other running processes to see what's going on unless you're deliberately underclocking.

edit: perhaps there's some cpu throttling enabled like C&Q in bios.

My last statement about getting less short Wu's dropping granted credit has come true.

Your completion times seem unusually long, more in line with an Opteron at 1.8 ghz vs stock of 2.6 on your Opteron 285. I would take a close look at other running processes to see what's going on unless you're deliberately underclocking.

edit: perhaps there's some cpu throttling enabled like C&Q in bios.

Computer is OK and no throttling in place. I believe it is the Work Units themselves, as others are now saying how long Wu's are taking (over 4 hours on Core 2 Duo in another thread).

It is causing my RAC to reverse direction. Credit on a per WU basis for the length of time and amount of work done would be better, but I think Scott said there was problems with doing this.

---

I am curious to know if those people that say they are getting more credits than "x"."y",or"z" projects are taking into consideration that their times crunching are being lowered and their credit claims appear higher on reboots and restarts.If you reboot or don't leave in memory you may have crunched 5 hours rebooted took another hour and reported only 1 hour of work. Credit per time can only be calculated here taking into account that when reported a wu might say it took 1 hour when it really took 6.You have to run ALL work to completion every time due to the wrapper.Averages taken off the website don't mean didly because almost all of them are lower credit claimed,longer crunch times than reality. So when people give "averages" I just roll my eyes ;-)~

---

I am curious to know if those people that say they are getting more credits than "x"."y",or"z" projects are taking into consideration that their times crunching are being lowered and their credit claims appear higher on reboots and restarts.

>>> I monitor my results fairly closely as I only have a few computers and a bit of spare time. I have not seen my "crunching being lowered and credit claims appear higher" due to any reboots or restarts (I run Linux and Windows for my projects and don't restart very often). Besides if it is happening as you say to all projects then credit claims can be compared to all projects as they all have the same problem.


If you reboot or don't leave in memory you may have crunched 5 hours rebooted took another hour and reported only 1 hour of work. Credit per time can only be calculated here taking into account that when reported a wu might say it took 1 hour when it really took 6.

>>> I leave all projects in memory. I have only noticed twice in the last year, where a WU in Boinc Manager reverts back zero in all stats and begins again. One was in CPDN where a running WU wiped its data in BM but the server kept all the Time Steps and the WU has kept going without error.
The second was the other day on Cosmology@home where I did happen to restart Boinc as it would no longer update completed WU's. Resetting Boinc fixed this but a Cosmology WU lost over 2 hours crunch time and reverted back to the start as if the checkpointing did not work. It completed after another 3+ hours and this is the time reported on the Web Site.
I have not noticed this behaviour before even after restarting Boinc or rebooting the computer.


You have to run ALL work to completion every time due to the wrapper.

>>> Not sure what you mean there, is not completing a WU also completing ALL work on that WU? Plus I can't report a WU if it has not completed.


Averages taken off the website don't mean didly because almost all of them are lower credit claimed,longer crunch times than reality. So when people give "averages" I just roll my eyes ;-)~

>>> In some instances this may be the case. As far as I have seen the times on the Web Site have matched pretty well with the time the WU took to process.
I am in a number of active projects and have very rarely seen what you have described, I have also read the forums for issues like this, so would not say it was as common as you imply.
If I had 100 completed Work Units of a project, containing both short length and long length work units, I am not going to list them all for people to see the differences. I will do an overall average of these WU's to see the effective output of my computer for that particular project, so "average" it must be.
If I have a problem with a certain WU then I will list that WU.

There is a list of projects comparing one to the other and the respective cobblestone per second of each project (I have forgotten its name and have lost the link), is in effect just "averaging" the work units from each project to get those results.

Sorry but you will have to keep rolling your eyes when I say average, I don't see another way to describe it.

---

P4HT, Windows XP, 5.10.13, multi-project: Roughly 15% of all the CaH Wus I've ran claim more than 50 credits. Conversely it means 85% claim less than 50 credits, so I'm happy.

Recent longer WUs have changed my stats shown above: Instead of 15% of completed tasks claiming more than 50, it has risen to 30%. My last returned result took 9.3 hours.

The second was the other day on Cosmology@home where I did happen to restart Boinc as it would no longer update completed WU's. Resetting Boinc fixed this but a Cosmology WU lost over 2 hours crunch time and reverted back to the start as if the checkpointing did not work. It completed after another 3+ hours and this is the time reported on the Web Site.
I have not noticed this behaviour before even after restarting Boinc or rebooting the computer.


You have to run ALL work to completion every time due to the wrapper.

>>> Not sure what you mean there, is not completing a WU also completing ALL work on that WU? Plus I can't report a WU if it has not completed.

Hey Conan the CAMB application at present works off of a windows wrapper,in other words it doesn't communicate with Boinc except through the wrapper therefore Boinc doesn't see CAMB's checkpoints or progress even though it is saved.When restarted cosmo picks up where it left off even though boinc says it restarts at 0 ...that is why you say the cp's failed but didn't.This is the only way cosmo works at the moment see here

Now if you understand how cp's work here now you will understand what I mean especially for someone who turns off their machine or reboots a lot. Their average times will be skewed.

Computer is OK and no throttling in place.

The facts aren't adding up. Your system should run 2.6 ghz yet your primary wingman, host 4357, continues to complete 25% - 30% quicker even on the new ones. This could be explained by an extreme overclock on his system which is hidden,

but,

let's look at other wingman.

Here's a Sempron 3000 (stock 1.7 ghz approx) that nearly beats you:

http://www.cosmologyathome.org//workunit.php?wuid=444590

Here's one that runs 2.0 ghz stock:

http://www.cosmologyathome.org//workunit.php?wuid=568391

2 running at 2.4 ghz stock:

http://www.cosmologyathome.org//workunit.php?wuid=559990

http://www.cosmologyathome.org//workunit.php?wuid=567619

Finally one running at 2.6 ghz stock:

http://www.cosmologyathome.org//workunit.php?wuid=577979


All complete more quickly and by a substantial margin. While there's a chance that all wingmen examined here are overclocked I believe this to be unlikely.

Any explanation for this?

I am curious to know if those people that say they are getting more credits than "x"."y",or"z" projects are taking into consideration that their times crunching are being lowered and their credit claims appear higher on reboots and restarts.

A reboot is about once a week at most, so no need to take this into consideration. Usually I look what's running and wait that little bit if a reboot seems necessary.

I agree 100% with Conan
Now do this. It will make your eyes roll!
Open BOINC manager and look at a cached “ready to start” wu at the bottom and note the “To completion” time.
Come back in a few hours and look at the “To completion” time again for the same wu. ???
On my old PC (2.8 Celeron) this time increased from 03:28 to 06:42 in one day and the average time to complete a wu from 3hours to 6 – 9 hours. (A Pentuim II can produce more than 50 points in 9 hours crunching any other project)

I’m enabeling “No new tasks”, wait for cache to run dry, revert back to my old project and “Roll my eyes”.
Maybe Scott can give the real explaination for this. I understand this is an Alpha project but JRenkar rolling his eyes won’t solve this bug
P.S. I also reboot about once a week at most.

---

I agree 100% with Conan
Now do this. It will make your eyes roll!
Open BOINC manager and look at a cached “ready to start” wu at the bottom and note the “To completion” time.
Come back in a few hours and look at the “To completion” time again for the same wu. ???
On my old PC (2.8 Celeron) this time increased from 03:28 to 06:42 in one day and the average time to complete a wu from 3hours to 6 – 9 hours. (A Pentuim II can produce more than 50 points in 9 hours crunching any other project)

I’m enabeling “No new tasks”, wait for cache to run dry, revert back to my old project and “Roll my eyes”.
Maybe Scott can give the real explaination for this. I understand this is an Alpha project but JRenkar rolling his eyes won’t solve this bug
P.S. I also reboot about once a week at most.

I think you misunderstood me...I 'roll my eyes' when people say they get too much credit on average, because restarts cut down on what you claim versus how long it really took.So 'averages' don't work here unless you never reboot in the middle of a unit. :)

http://www.cosmologyathome.org/forum_thread.php?id=231&nowrap=true#3288

---

Scott Kruger
Project Administrator, Cosmology@Home

I'm increasing the credits for the new WUs to 100, since run times seem to have doubled in most cases. I'd rather err on the side of too many credits than too few.

Please tell me any concerns.

---

Scott Kruger
Project Administrator, Cosmology@Home

I'm increasing the credits for the new WUs to 100, since run times seem to have doubled in most cases. I'd rather err on the side of too many credits than too few.

Please tell me any concerns.

That's great news Scott,
Thanks for that, even with you giving out 70 per job, the longer ones were only giving low/medium returns.

With your great reponse to your crunchers concerns, (and even pre-empting them as in this example), is the reason I added another 2 computers to the project.

Thanks again.

---

Hi! Can someone check wu's 629084 and 627646? I would think credit would be given anyway, not zero credit?

Thanks in advance,
Scott

Just a quick comparison on my 3.0GHz P4 with HT on running XP. So far on only 2 Cosmology wu which have been granted 100c each, this machine is getting basically the same amount of credit as the optimised S@H app per day. Which is just under twice as much as I get on MalariaControl.net, (i.e MC only give 50% of the credit as I get here). Is the Cosmology app an optimised app based on processor capabilities or just a basic \'doze app?

Live long and BOINC!

Paul.

500 credit in just over 24 hours....LOVE it.

2.6GHz with 1.5GB of RAM

---

Here\'s a credit comparison: http://boincstats.com/stats/project_cpcs.php

Cosmology@Home is #3 behind TPS(?) and RieselSieve, but is still comfortably ahead of the remaining projects. I\'d say a good trade-off, especially since it\'s still in beta. ;)

Is the Cosmology app an optimised app based on processor capabilities or just a basic \'doze app?

Just a standard application, Paul. No CPU specific optimizations.

---

The Credit Uniformity Train Members dont give up easy. Yeez.
PS: Hi Saenger

Honestly: I do not understand how credits are granted here. I find that the claimed credits are granted fairly enough so there is no need to add more to reach the magic (and not understandable) and uniform 100.