| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: Website Support
Thread: Why only BOINC version 6.2.28 on this website and not the latest version? |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 19
|
|
| Author |
|
|
doive1231
Cruncher Joined: Jun 9, 2009 Post Count: 10 Status: Offline Project Badges: 
|
Hi,
----------------------------------------I've downloaded the World Community Grid BOINC Manager from this website and it is version 6.2.28. If you go to the BOINC website, the version there is 6.10.18, which would appear to be much newer. Surely this website should be updated to include the latest Manager version. I say this as I wish to use my GPU to do work units with other projects but with version 6.2.28, my GPU is not recognised in the messages tab. I believe that the latest version supports GPU detection and projects. So I hope I'm being helpful to all to suggest the latest versions. Thanks. |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
WCG is preparing to test 6.10 for Windows. The 6.10 offered for Mac an Linux had technical reasons. Older versions would not work with newer OS releases.
----------------------------------------Now, given that WCG has no GPU projects, there's no pressure from their side to jump on the bandwagon. The Berkeley development process and twice a retraction of the recommended releases, first 6.10.36, then 6.10.43 is kind of a sign of the stability which has not been reached yet before WCG will offer this as as a stable release. There is an alpha WCG build, the latest 6.10.56 which has been running on my duo, but since it's WCG exclusive computing, not much I can say about the CPU/GPU scheduling. There are still reports of work not being fetched for the CPU when the GPU has been engaged for certain projects... and that I think is a major problem.
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
doive1231
Cruncher Joined: Jun 9, 2009 Post Count: 10 Status: Offline Project Badges:
|
So I should beware using 6.10.18 and World Community Grid. I was hoping to use my CPU for WCG and my GPU for another project but if there could be stability issues then I will try it but accept it has not been tested thoroughly.
---------------------------------------- |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Many people run newer versions of BOINC and it processes WCG just fine. I have been using different versions for more than a year and have personally not encountered any issues processing WCG with them.
|
||
|
|
knreed
Former World Community Grid Tech Joined: Nov 8, 2004 Post Count: 4504 Status: Offline Project Badges: 
|
The BOINC 6.10.56+ version is getting close to being marked stable. We are testing the version that includes the WCG skin and auto-attach mechanism. Once the BOINC version has been public for a couple of weeks and if no new issues are identified, then we will adopt the 6.10.56+ version as well.
I include the (+) because it is possible that there might be another dot release or two before it is declared stable. If you wish to get it a try, you can download it here: http://boinc.berkeley.edu/dl/wcg_boinc_6.10.56_windows_intelx86.exe Please note that this is a TEST version. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Many people run newer versions of BOINC and it processes WCG just fine and many people run versions other than those tested by WCG, have many problems and create many support requests for the CAs; and many of these people end up going back to the WCG approved version. |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
The BOINC 6.10.56+ version is getting close to being marked stable. We are testing the version that includes the WCG skin and auto-attach mechanism. Once the BOINC version has been public for a couple of weeks and if no new issues are identified, then we will adopt the 6.10.56+ version as well. I include the (+) because it is possible that there might be another dot release or two before it is declared stable. If you wish to get it a try, you can download it here: http://boinc.berkeley.edu/dl/wcg_boinc_6.10.56_windows_intelx86.exe Please note that this is a TEST version. Kevin, feel free to peruse the Berkeley forums on issues I've posted there, still living today in 6.10.56 that have been brought to the attention there, and ignored and for this one even given a suggestion how to mitigate, met with inaction [as seen from my corner on the planet]: World Community Grid 6.11 hfcc HFCC_s2_02718272_s2_0001 02:31:08 (07:43:33) 24-5-2010 10:59:43 24-5-2010 12:42:46 Reported: Ok World Community Grid 6.11 hfcc HFCC_s2_02710083_s2_0000 01:49:59 (06:57:14) 24-5-2010 9:23:08 24-5-2010 12:42:22 Reported: Ok Elapsed time gets lost on system crash or power outs, cache and scheduling [such as DCF] build upon this value instead of the CPU time (bracket time) for CPU based science application. The mitigation rule proposed for the client to test if Elapsed [wallclock] is less than CPU time during resume and have it take that time as minimum elapsed. PS: Don't want to know how credit is computed on GPU tasks when time is lost during crashes if they depend on the elapsed time value????
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
still living today in 6.10.56 that have been brought to the attention there, and ignored and for this one even given a suggestion how to mitigate, met with inaction How can anyone yet suggest a new version is close to being ready? Newer is NOT better; only better is better |
||
|
|
Ingleside
Veteran Cruncher Norway Joined: Nov 19, 2005 Post Count: 974 Status: Offline Project Badges: 
|
The mitigation rule proposed for the client to test if Elapsed [wallclock] is less than CPU time during resume and have it take that time as minimum elapsed. The problem with your suggested rule is, for SMP-applications the wallclock-time is (nearly) always less than cpu-time... PS: Don't want to know how credit is computed on GPU tasks when time is lost during crashes if they depend on the elapsed time value???? Well, I've not looked too closely on the new credit-system, but I wouldn't expect it's using wallclock-time for this... BTW, since many of the GPU-applications uses maybe only 1% of a core, even with your rule of setting wallclock = cpu-time, they'll still only get a couple of seconds wallclock-time for the crash...  "I make so many mistakes. But then just think of all the mistakes I don't make, although I might." |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
The mitigation rule proposed for the client to test if Elapsed [wallclock] is less than CPU time during resume and have it take that time as minimum elapsed. The problem with your suggested rule is, for SMP-applications the wallclock-time is (nearly) always less than cpu-time... PS: Don't want to know how credit is computed on GPU tasks when time is lost during crashes if they depend on the elapsed time value???? Well, I've not looked too closely on the new credit-system, but I wouldn't expect it's using wallclock-time for this... BTW, since many of the GPU-applications uses maybe only 1% of a core, even with your rule of setting wallclock = cpu-time, they'll still only get a couple of seconds wallclock-time for the crash... We were talking in first line solution for CPU projects !!! If there are exceptions, this can be coded in, but MOST IMPORTANTLY, the DC world is CPU ruled, not subordinate to GPU or SMP quirks which has been dominating all of 6.4 and upwards. Crunch WCG only, you're all fine with 6.2.28 from W2K, XP through W7.
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
|