| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: Need Help |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 17
|
|
| Author |
|
|
[B-S] Gamma-Ray
Cruncher Joined: Feb 27, 2008 Post Count: 24 Status: Offline |
Hello Didactylos: As you suggested I Stopped Boinc and deleted Client_state.exm and Client_state-prev.xml and restarted Boinc. Everything back to normal, completion time reads correctly and crunching two work units again. One question re Sekerob's reply should I ideally only be crunching 1 work unit at a time. I have been crunching 2 units at a time for about a year now and in all that time I have only had 2 work units with a problem. If the recommendation is to only crunch one at a time that's what I will do but I would appreciate any additional input before a change to only one, seems like a waste if I can do two at a time. For me, if a work unit seems to be acting oddly or stuck etc. and checkpointing isnt a issue with that particular work unit, then I will first suspend one or in this case both work units (Or just the whole project) and then just resume them afterwards to see if that fixes the issue. If not, then I will goto the suspend the work units, then exit the client, restart the client and then resume the work and see what happens then. If that still doesn't fix things then I usually go looking on the projects forums for a possible reported problem similar and any solutions to help. Clients can and do occasionally get stuck for various reasons, but usually if its a client issue that effects the work units, then restarting the client is a good first step to resolving it.  |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Thanks for the reply Sekerob I will do that. Another simple fix to that: Go to opening post, hit the Edit button next to title and insert [RESOLVED]I just compared my completion times when I was only doing one unit at a time and they seem to complete faster. Also from retsof's reply as I am only swaping back and forth between two work units sharing a single floating point register it does seem like only doing one unit at a time is the way to go. I would make this thread as complete but I am not sure how to do that, could you please do that for me. The intense swapping can in fact be leading to the 2 parallel jobs needing more 'wallclock' time than 2 sequential jobs, plus the usability improving markedly if just running one job at a time on a P4HT.
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
retsof
Former Community Advisor USA Joined: Jul 31, 2005 Post Count: 6824 Status: Offline Project Badges: 
|
Hyperthread returns for the future....it's aliiiiiiiiveeeee!!!
----------------------------------------I'll wait for a faster phenom, or buy a used clovertown. The successor to Penryn, based on a new post-Core microarchitecture which features the return of hyperthreading, is Nehalem; it was discussed at the September 2007 IDF meeting, though the release date is not until the end of 2008. It looks like the Bloomfield desktop will be a 4 core real or 8 core virtual machine. There's an extended 8 core beckton server 16 core virtual. Thanks for your reply retsof Try it both ways to do a comparison, or just force boinc processors=1 as above for another method to keep the HT otherwise.One question if I turn HT off what are the advantages vs disadvantage of doing so? I assume HT was set up and made available for some reason. If I turn HT off how will it affect other things I may want to do on my computer? HT was a marketing "preempt" to snare some market share for a virtual double computer, ahead of the true Core Duo or X2 machines. It works best when floating and integer applications do the sharing, but was still greatly misunderstood.
SUPPORT ADVISOR
----------------------------------------Work+GPU i7 8700 12threads School i7 4770 8threads Default+GPU Ryzen 7 3700X 16threads Ryzen 7 3800X 16 threads Ryzen 9 3900X 24threads Home i7 3540M 4threads50% [Edit 3 times, last edit by retsof at Mar 30, 2008 6:37:58 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I disagree with the advice to only run one job on a HT processor, especially if you run other projects in addition to WCG. Usually the thouroughput is about 15% higher when using HT. Each task takes nearly double the amount of time but you get more done in a day. When you have different projects running there is usually slightly less contention for the FPU so you get even better thoutoughput.
On the original problem I noticed there is a "reading local preference" line. You may have limited CPU usage there accidentally. In the manager go to advanced -> preferences and select clear. |
||
|
|
retsof
Former Community Advisor USA Joined: Jul 31, 2005 Post Count: 6824 Status: Offline Project Badges:
|
If you have two projects of a different sort, you may have a bit of a gain with HT. One floating point (WCG) and one integer (some mathematical prime computation, maybe) would be a good mix.
----------------------------------------Running two of the same type of workunits here would not be the idea I had in mind. Also check the hard drive for intense activity with HT. Even without it, the disk light was very busy on an X2. I got a second HD to hold the windows swap file and backups. (not hyperthreaded here) That settled it down.
SUPPORT ADVISOR
----------------------------------------Work+GPU i7 8700 12threads School i7 4770 8threads Default+GPU Ryzen 7 3700X 16threads Ryzen 7 3800X 16 threads Ryzen 9 3900X 24threads Home i7 3540M 4threads50% [Edit 4 times, last edit by retsof at Mar 31, 2008 1:24:05 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
.... you may have a bit of a gain with HT. One floating point (WCG) and one integer (some mathematical prime computation, maybe) would be a good mix. Running two of the same type of workunits here would not be the idea I had in mind. .... How about multi wu's in the same machine on multibus/plugin boards ? http://www.worldcommunitygrid.org/forums/wcg/viewthread?thread=10847 "Dedicated Crunching Machines Hardware" Where discussions are mostly about cell broadband engine (BE) processor. The last message on Dec 7, 2007 ends talking about sending " information between multiple cores ...on a chip using pulses of light through silicon, instead of electrical signals on wires. The breakthrough, known in the industry as a silicon Mach-Zehnder electro-optic modulator .... ". I wonder about using the "transputer" engine of days past to improve multi-PC processing grid work units using programable information controlers on chip with exterior hooks. Much cooler HT on a grander scale. This could drive down multiprocessing costs. Can it be built ? |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello PTabolinsky,
This is going off-topic. It is time to let this thread sink into the past. But for the predicted future of computing, look at this recent thread in Chat: http://www.worldcommunitygrid.org/forums/wcg/viewthread?thread=19302 The next decade is going to see major deployment of parallel processors - no doubt with plenty of problems and architectural blind alleys. Lawrence |
||
|
|
|