| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Active Research
Forum: Mapping Cancer Markers Forum
Thread: How do points correlate with work done? |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 29
|
|
| Author |
|
|
David_L6
Senior Cruncher USA Joined: Aug 24, 2006 Post Count: 296 Status: Offline Project Badges: 
|
Not sure how to word this question... how do the points received for a given work unit correlate with the amount of work done for the project? Here's an example that caused me to ask this question.
----------------------------------------13.57 hours run time - 222.5 points 2.58 hours run time - 84 points Same project. More than 5.25 times run time for less than 2.75 times points. Did the longer running work unit really do that much less actual work for the project?  |
|||
|
||||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello David_L6,
Remember that 1 BOINC credit = 7 WCG points. Here is the explanation of the BOINC credit and the 'cobblestone' that it is based on. https://en.wikipedia.org/wiki/BOINC_Credit_System For any more info, look at your Result Status page and click on the Result Name. Lawrence |
||
|
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1684 Status: Offline Project Badges:
|
Lawrence's answer is correct.
----------------------------------------However I notice that MCM is not a "generous" project. Since my hosts work for MCM only, I experience a significant drop down of granted credit: daily average from 17'400 to 14'400. I observe that the WinXP hosts seem to remain on a stable level until the daily average of Ubuntu hosts goes dramatically down, e.g. from 5'400 to 3'700. This effect is visible on all Ubuntu hosts. Maybe the MCM application needs some optimisation? ... Cheers, Yves |
||
|
|
alged
Master Cruncher FRANCE Joined: Jun 12, 2009 Post Count: 2369 Status: Offline Project Badges: 
|
"However I notice that MCM is not a "generous" project."
----------------------------------------As KerSamsom just said i realize why it takes so long to get just a silver badge with an IMac and a Macbook Pro both crunching only for MCM. M Many wu's are labelled not valid.I have not elements of comparison but are Apple machines experience the same drop on daily average? cheers  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello alged,
MCM appears to have a problem with work units that are interrupted and start over again at the last check point. Hopefully this bug will be tracked down and squashed. Until then, it is a good idea to avoid reboots or anything else that will remove the work unit from memory. Lawrence |
||
|
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1684 Status: Offline Project Badges:
|
Just for the record: my hosts are computing 24/7/365 without reboot.
----------------------------------------The reboot/restart problem cannot be the cause for the strong decreasing of granted credit. @alged: the badges depend of CPU time not of granted credit. Yves |
||
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7846 Status: Offline Project Badges:
|
Just for the record: my hosts are computing 24/7/365 without reboot. The reboot/restart problem cannot be the cause for the strong decreasing of granted credit. @alged: the badges depend of CPU time not of granted credit. Yves I also have hosts which run 24/7 and a number of the units show the message: 12:42:38 (17202): No heartbeat from client for 30 sec - exiting 12:42:38 (17202): timer handler: client dead, exiting. Some become valid and some don't. The message has nothing to do with rebooting. Cheers
Sgt. Joe
----------------------------------------*Minnesota Crunchers* [Edit 1 times, last edit by Sgt.Joe at Dec 2, 2013 1:57:48 AM] |
||
|
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1684 Status: Offline Project Badges:
|
After a couple of weeks crunching for MCM, I can report following reality between end of October 2013 (CEP2 and FA@H) and the last 3 weeks (MCM only):
----------------------------------------
There is no problem caused by the reboot issue, since these hosts operate 24/7/365. The strong decreasing performance of the Ubuntu based systems could indicate a serious performance problem with the application; e.g. some missing code optimisation or still activated debugging support ??? In all cases, I would appreciate that this issue could be investigated as well as to receive some feedback. Indeed, the electricity costs are not related to the effective GFlops  .Cheers, Yves |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Think you got it entirely by the wrong end of the stick. FAHV on Linux is hugely faster than the Windows version [That was for HCC1 also], so the credit per unit of time on Linux is substantially higher [coming out of the same comparative application pool]... matching the speed, credit is 60-70 percent higher on Linux. Opposed, faah [AutoDock] and MCM has matching speed on Windows and Linux AFAICT, so their credits are more balanced. Thusly, if you switch on Ubuntu from FAHV [90% of the FA@H work] to MCM, you see a substantial fall in credit [and this is not the first time I post this rationale].
For CEP2... I'll pass up on opining on this one. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
P.S. All VINA based sciences are hugely faster on Linux giving that credit uplift. Have a thread on that discussing this for DSFL/SN2S/GFAM. Maybe the performance problem is on Windows ;P
(calling Lawrence for his compiler spiel) |
||
|
|
|