| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: Computing for Clean Water Forum
Thread: 64 Bit Computing for Clean Water application deployed |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 32
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I tell by looking at the Version number in the WU Result. If it is 640 then it is the old 32 bit, My i7 shows Version 641 on the llatest ones....
----------------------------------------[Edit 1 times, last edit by Former Member at Mar 19, 2011 5:12:13 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
The Version 641 seems about 20 minutes quicker on a WU....
 |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
...the number of credits were slightly more, about one credit higher...
 |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I tell by looking at the Version number in the WU Result. If it is 640 then it is the old 32 bit, My i7 shows Version 641 on the llatest ones.... Good way, for the moment, until a new version is released for all platforms. Anyway, I've inserted new columns in the Matrix FAQ so folks can see what's known. Not looked at resource use differences at all, if there at all are any. The Version 641 seems about 20 minutes quicker on a WU.... Did not do a Beta average as just done for HCC, but my own devices did about 6% faster and those that who participated did ballpark 4.33 hours per task mean. Think by Monday we'll be able to see the project wide impact, for now the previous near flat 5.00 hours dropped to 4.9 hours as average. The good thing is, that long as it's a newer client than the host will be auto-assigned either 32 or 64 bit, according to ''skill''. ...the number of credits were slightly more, about one credit higher... Think the credits tap into the same platform pool, so yes it's like getting 32 bit credit total per task for a faster 64 bit. Could be that over time an equalizer kicks in. Can't care too much... more work done in the same amount of time. --//-- edit: Added Matrix link [Edit 1 times, last edit by Former Member at Mar 19, 2011 7:59:58 PM] |
||
|
|
martin64
Senior Cruncher Germany Joined: May 11, 2009 Post Count: 445 Status: Offline Project Badges: 
|
The Version 641 seems about 20 minutes quicker on a WU.... Not on my computer. While 640 WUs took less than 3:40 in average, 641 WUs are well above an average of 3:50. Credits granted didn't change noticably. Regards, Martin  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
That's what we were fearing for. Wish/Wonder if there is a way to force to keep the 32 bit coming other than client downgrading. Think Ingleside (who's present momentarily) posted somewhere from what version it's automatic. Probably step back to 6.2.28 if you're not involved in GPU computing.
--//-- |
||
|
|
Ingleside
Veteran Cruncher Norway Joined: Nov 19, 2005 Post Count: 974 Status: Offline Project Badges: 
|
Not on my computer. While 640 WUs took less than 3:40 in average, 641 WUs are well above an average of 3:50. Credits granted didn't change noticably. To accurately compare the speed, all the tasks should be the same size, or you'll need to use the exact same tasks for benchmarking. A quick look on my own c4cw reveals over 20% variation in cpu-time between them, so any statements about 64-bit being faster or slower compared to these old results wouldn't make much sence... If you really wants to only use the 32-bit application, a downgrade to 32-bit BOINC v6.6.xx (or lower) should solve this.  "I make so many mistakes. But then just think of all the mistakes I don't make, although I might." [Edit 1 times, last edit by Ingleside at Mar 19, 2011 10:20:39 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Were you looking at CPU or Elapsed? Less quick look, all the 6.40 from my not so special but very constant cycle stable duo, ALL target03:
6.40 c4cw c4cw_target03_059480847_0 07:14:24 (06:21:39) 2-3-2011 1:14 2-3-2011 1:16 6.40 c4cw c4cw_target03_061446836_0 06:54:47 (06:23:19) 5-3-2011 21:53 5-3-2011 21:55 6.40 c4cw c4cw_target03_062270441_0 07:18:48 (06:22:16) 7-3-2011 0:16 7-3-2011 0:18 6.40 c4cw c4cw_target03_062330178_0 06:31:11 (06:13:39) 7-3-2011 6:22 7-3-2011 6:24 6.40 c4cw c4cw_target03_062389132_0 07:03:33 (06:26:00) 7-3-2011 13:25 7-3-2011 13:28 6.40 c4cw c4cw_target03_062414726_0 07:22:32 (06:27:39) 7-3-2011 17:50 7-3-2011 17:52 6.40 c4cw c4cw_target03_063384779_0 06:50:28 (06:25:05) 8-3-2011 4:24 8-3-2011 4:26 6.40 c4cw c4cw_target03_063549901_0 06:44:19 (06:27:30) 8-3-2011 9:46 8-3-2011 9:48 6.40 c4cw c4cw_target03_063609873_1 06:49:17 (06:27:39) 8-3-2011 11:15 8-3-2011 11:15 6.40 c4cw c4cw_target03_064089360_0 07:17:10 (06:35:17) 8-3-2011 18:31 8-3-2011 18:33 6.40 c4cw c4cw_target03_064364536_1 06:59:05 (06:28:20) 8-3-2011 19:40 8-3-2011 19:42 6.40 c4cw c4cw_target03_064520250_0 06:53:22 (06:24:59) 9-3-2011 4:09 9-3-2011 4:11 6.40 c4cw c4cw_target03_064527103_0 06:34:21 (06:14:58) 9-3-2011 9:09 9-3-2011 9:11 6.40 c4cw c4cw_target03_064705293_0 06:52:28 (06:21:31) 9-3-2011 11:01 9-3-2011 11:03 6.40 c4cw c4cw_target03_064706420_0 07:15:32 (06:27:14) 9-3-2011 16:40 9-3-2011 16:42 6.40 c4cw c4cw_target03_064875420_0 07:23:35 (06:26:46) 9-3-2011 19:05 9-3-2011 19:07 6.40 c4cw c4cw_target03_065084067_0 06:39:20 (06:22:03) 10-3-2011 6:49 10-3-2011 6:51 6.40 c4cw c4cw_target03_065672281_1 06:48:11 (06:22:00) 11-3-2011 6:38 11-3-2011 6:40 A high/low range variance of 5.4%. Before the upgrade to 6.41 64 bits, the flatter than flattest Kansas project mean was 5.01 hours, suggesting the variation is very small and in fact smaller than with HCC when looking at the WCGART chart. --//-- |
||
|
|
martin64
Senior Cruncher Germany Joined: May 11, 2009 Post Count: 445 Status: Offline Project Badges:
|
Were you looking at CPU or Elapsed? CPU time, as per results page. Unfortunately the 640 results are gone from the results page now, so I cannot post any details. System is Win7/64, Pentium E5400@3.1GHz. |
||
|
|
Ingleside
Veteran Cruncher Norway Joined: Nov 19, 2005 Post Count: 974 Status: Offline Project Badges:
|
Were you looking at CPU or Elapsed? Less quick look, all the 6.40 from my not so special but very constant cycle stable duo, ALL target03: Well, I looked at both cpu and elapsed, but quoted the info for cpu-time. Apparently it's some time since last c4cw since still has v6.15. In any case: type - low - high - %difference cpu - 16474 - 20065 - 21.8% run - 16515 - 22379 - 35.5% %efficiency - 77.307 - 99.885 - 29.2% fpops_est - 3.32E13 - 4.12E13 - 23.9% Total: 83 results on i7-920. Looking only on target03: cpu - 18276 - 20065 - 9.8% run - 18523 - 21745 - 17.4% %efficiency - 88.18 - 99.33 - 12.6% fpops_est - 4.099E13 - 4.119E13 - 0.47% Total: 44 results on i7-920. Target03 on another computer: cpu - 13071 - 14959 - 14.4% run - 13130 - 15092 - 14.9% %eff - 97.471 - 99.779 - 2.4% fpops_est - 4.108E13 - 4.119E13 - 0.25% Total: 47 results So, based on fpops_est, the variations between wu's in target03 was less than 0.5%, meaning the wu's most likely is the same. But, despite of this, the variations in cpu-time for me was 9.8% on one computer, and 14.4% on another computer... "I make so many mistakes. But then just think of all the mistakes I don't make, although I might." [Edit 1 times, last edit by Ingleside at Mar 20, 2011 12:38:06 PM] |
||
|
|
|