| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: Website Support
Thread: Where are Stats Explained? |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 9
|
|
| Author |
|
|
i_am_jim
Advanced Cruncher Joined: Nov 1, 2010 Post Count: 62 Status: Offline Project Badges: 
|
If I understand Points the purpose is to normalize how much in crunching/results is achieved independent of computer speed. It helps faster computers look better than simply looking at crunching time. For example, take a supercomputer and a slow PC.
----------------------------------------The supercomputer might produce more results in one hour than the slow PC in five years. Points try to give the supercomputer credit for all its effort by giving the same number of points to the supercomputer for its one hour as given to the PC for it five years. Is this correct? I assume Results Returned is straightforward. I assume it is the number of work units returned and validated. I'm not sure what Run Time is. At first I thought it was simply the number of hours your device/computer devoted to crunching. But, I've seen numbers in my stats that didn't seem to agree what I would have done in a given day. Also there's a member that racks up 60+ years in one day. It's a little hard to visualize how that's done. ---------------------------------------- [Edit 1 times, last edit by i_am_jim at Apr 25, 2014 2:11:58 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
You answered most of your own questions. Run time is two elements, where there is no measurable cpu time, for instance on gpgpu, the elapsed time aka runtime is recorded as base for the points calculation. Where there is a cpu being used, the cpu time is recorded as base. This is fair in that if you set for instance to only use 60 percent of the cpu time, the task is really run much longer than cpu cycles being used. The closer the elapsed/runtime is to the cpu time, the higher the efficiency. Best i've seen is 99.9%. 100% could not be as the operating system does need a little time as does the boinc core agent. At any rate, the result status pages show both times from which you can compute the efficiency achieved.
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Run Time is the amount of CPU time, not elapsed time, taken on a particular WU. On PCs that do little but crunch the CPU and elapsed should be within a few percentage points, if a PC is used for other tasks, elapsed will far exceed CPU time. If you have hundreds of CPUs working then you can get 60 years in one day!
 |
||
|
|
i_am_jim
Advanced Cruncher Joined: Nov 1, 2010 Post Count: 62 Status: Offline Project Badges:
|
I still don't understand. Yesterday one of my computers had 2 days 20 hours of "run time"?? Obviously, that's not possible for the CPU time of one computer unless it separates out multiple processors. If so, how does it know how much was done by each?
----------------------------------------I don't use my graphics card but if a computer does, how is that handled in RunTime ---------------------------------------- [Edit 3 times, last edit by i_am_jim at Apr 25, 2014 2:43:00 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
You have to average out over quite a few days. You can have days when nothing validates and the following day everything validates, including the previous days so day one has zero, but day two has day 1 plus day two. It all depends when wingmen return WUs.
|
||
|
|
i_am_jim
Advanced Cruncher Joined: Nov 1, 2010 Post Count: 62 Status: Offline Project Badges:
|
That makes sense. The problem is, almost every day it's more than a day. Here are the last several days:
----------------------------------------04/25/2014 0:001:03:32:40 3,031 5 04/24/2014 0:002:19:51:16 6,522 8 04/23/2014 0:000:12:40:49 1,308 3 04/22/2014 0:001:18:28:07 3,863 9 04/21/2014 0:001:19:37:18 3,933 7 04/20/2014 0:001:19:16:30 5,180 9 04/19/2014 0:001:05:05:52 3,012 4 04/18/2014 0:002:06:37:09 4,899 8 04/17/2014 0:001:05:50:56 3,075 7 04/16/2014 0:001:13:32:08 3,429 6 04/15/2014 0:002:01:00:35 4,663 8 04/14/2014 0:001:12:58:44 3,503 8 |
||
|
|
jonnieb-uk
Ace Cruncher England Joined: Nov 30, 2011 Post Count: 6105 Status: Offline Project Badges: 
|
Obviously, that's not possible for the CPU time of one computer unless it separates out multiple processors. It does! A quad core computer can therefore complete a (theoretical) maximum 96 hours CPU in one calender day. |
||
|
|
i_am_jim
Advanced Cruncher Joined: Nov 1, 2010 Post Count: 62 Status: Offline Project Badges:
|
Obviously, that's not possible for the CPU time of one computer unless it separates out multiple processors. It does! A quad core computer can therefore complete a (theoretical) maximum 96 hours CPU in one calender day. I see. Ahrite then, as Karl Childers would say. |
||
|
|
twilyth
Master Cruncher US Joined: Mar 30, 2007 Post Count: 2130 Status: Offline Project Badges: 
|
The supercomputer might produce more results in one hour than the slow PC in five years. Points try to give the supercomputer credit for all its effort by giving the same number of points to the supercomputer for its one hour as given to the PC for it five years. Is this correct? Virtually all supercomputers are multiprocessor systems. Many combine both CPUs and GPGPUs to accomplish tasks. From wikipedia Approaches to supercomputer architecture have taken dramatic turns since the earliest systems were introduced in the 1960s. Early supercomputer architectures pioneered by Seymour Cray relied on compact innovative designs and local parallelism to achieve superior computational peak performance.[9] However, in time the demand for increased computational power ushered in the age of massively parallel systems. So a supercomputer would never appear to BOINC as just one CPU but as many cores (or virtual threads in the case of hyperthreading CPUs) as made up the machine.   |
||
|
|
|