Message boards :
Number crunching :
LHC still requires Boinc 4.45?
Message board moderation
Previous · 1 · 2 · 3 · Next
| Author | Message |
|---|---|
 Paul D. BuckSend message Joined: 2 Sep 04 Posts: 545 Credit: 148,912 RAC: 0 |
CPDN work will die for a variety of reasons, not all of them are directly caused by making changes to the system, though they may be coincidental to that ... When you are upgrading BOINC there is a good guide in the Wiki on how to do it with minimal risk and it DOES include a process to back-up the folders before you damage things .., oh, and it has pictures ... |
|
Send message Joined: 13 Jul 05 Posts: 82 Credit: 6,336 RAC: 0 |
I am still using 4.68 and have no problems with it. I know that I should upgrade sometime, when I tried before it lost over 1000 hours on a Climet Sulpher model. Had to restore a backup of 4.68 to get all those hours back. John I did that twice, both times lost a CPDN Sulpher model with over 1,000 hours. At least I had a backup to restore of what I am using. With all the problems other were having I was waiting for them to get 5.x more stable, looks like they have it now but well into another Sulpher unit. May try it again but have a good backup of this first. Ray  Pizza@Home - Rays Place - Rays place Forums |
|
Send message Joined: 13 Jul 05 Posts: 82 Credit: 6,336 RAC: 0 |
CPDN work will die for a variety of reasons, not all of them are directly caused by making changes to the system, though they may be coincidental to that ... Yes, when you read the boards at CPDN you find a lot of things that have gone wrong for others. I lost my first model, than had to restore the 2nd one to and it finished up. Looks like 3 and 4 will finish without needing any restores, but have backed them up every 3 to 5 days anyways. If I keep all the backup I could fill a few DVD's with them. But I just backup and delete tha last backup. I always keep the backup file of the other system so if a drive goes I can restore BOINC from the other computer without lousing CPDN hours. Ray Pizza@Home - Rays Place - Rays place Forums |
Fuzzy Hollynoodles Send message Joined: 13 Jul 05 Posts: 51 Credit: 10,626 RAC: 0 |
I don't know if I crunch with the latest BOINC version, but I crunch with 5.3.2 with no errors at all, neither here or in Rosetta, which can be quite tricky also. "I'm trying to maintain a shred of dignity in this world" - Me 
|
|
Send message Joined: 28 Aug 05 Posts: 6 Credit: 1,001,863 RAC: 0 
|
I am running Boinc 5.2.13 on windows (this is the latest recommended version)and I have zero problems, with either lhc, einstein or seti (i.e. all my projects), hence no need to run an outdated version of boinc. |
Jim Baize Send message Joined: 17 Sep 04 Posts: 103 Credit: 38,543 RAC: 0
|
HMMM I have been using BOINC 4.45 since it was first released, and have never had a problem with it. I refuse to upgrade until I am comfortable that all the bugs have been worked out. At this point, I won't be upgrading anytime soon if you get the drift. I look at it this way. An upgrade fixes old problems. It may create new problems and it may not totally fix old problems, but it is always a step forward (at least it is supposed to be.) Now, I know there are times when a newer version has a lot more problems than an older version, but that doesn't seem to happen a lot. Plus, with BOINC, it has gone through several iterations of upgrades. There have been a multitude of problems fixed and processes improved since the 4.xx clients. They are still working, they are adding options, fixing problems, improving processes. If you are waiting for a bug-free release, I think you'll be waiting forever to upgrade. |
|
Send message Joined: 26 Sep 05 Posts: 85 Credit: 421,130 RAC: 0 
|
I'm using the lattest version of the BOINC 5.xx client, and have not had any problems with LHC. Whether or not signup is a problem (specifically with the signup wizzard) not sure, but if it is, one can install the old version, attach, and then upgrade to the new. I did that for some projects when I built my A64, but that was when the 5.xx client was still in beta. LHC now has a password attached to it, so I'm gathering it shouldn't be, as one should be able to attach with both an email and password, rather then through the project ID... As to CPDN, I have tried a few WUs, and have not got a single one to run. I made a good faith effort to get one completed, but I keep comming home after they've run for about 50-100 hours to see they errored out. I've gone onto some other projects for now, to take it's time slice. Perhaps, with a software upgrade I might try it again, but right now, without a sucessful completion, it's just trashing WUs now as CPDN and my comp don't seem to like each other, which based on the CPDN model other comps don't then pick up... All other projects work error free though, and I doubt it's a hardware problem given this and that I have all new RAM now, had a new CPU and mobo last fall, and Prime 95 and other such tests, pass with no errors on the thing... 
|
|
Paul D. Buck Send message Joined: 2 Sep 04 Posts: 545 Credit: 148,912 RAC: 0 |
CPDN models are the hardest probably to complete well. I, for example, have an Apple PowerMac which has not complete on yet. Is it the Mac code, my Mac? who knows. The good news, for me, is that I have no troubles running the models on windows. LHC@Home models are the hardest probably to get validated because of the accuracy requirements. Like everything, each project has pros and cons. Good and easy things, bad and hard ... I was thinking if I do get another G5 PowerMac I will try it with CPDN to see if it will run the models. If not, then maybe it is the Mac version? I know some are getting models done so that does not make much sense. But ... all part of the challenge! |
|
Nightbird Send message Joined: 13 Jul 05 Posts: 55 Credit: 41,230 RAC: 0 |
accuracy requirements ? What are you thinking ? Do you want to get banned for 31 years, your account and credits deleted at a Boinc project ? Predictor@home is your best choice. |
UBT - Halifax--lad Send message Joined: 30 Sep 04 Posts: 23 Credit: 1,372 RAC: 0 |
Ive got the new version of BOINC currently up for testing 5.3.15, no problems here getting or processing LHC WU's on the client Join us in Chat (see the forum) Click the Sig  Join UBT |
|
Paul D. Buck Send message Joined: 2 Sep 04 Posts: 545 Credit: 148,912 RAC: 0 |
The tolerances on error are very low. It is the reason they do not use the PowerMac, the FPU is too different. So, they could not get the results to come back within the tolerances needed. When the PowerMacs go Intel this may no longer be true, though the issue will then be the availability of the FORTRAN compiler on the OS-X (Intel) platform. I am just saying that your chances for failing to validate are slightly higher at LHC@Home because the tolerances are tighter. |
|
Send message Joined: 13 Jul 05 Posts: 4 Credit: 2,442,589 RAC: 269 
|
I upgraded to 5.2.13 and still have loads of Ghost Work Units. Any suggestions on how to get these little bastards under control? LHC definetly has problems in the ghost work unit area. Running 5.2.13 or later on all computers and for the most part rarely see a ghost unit on other projects, except for LHC. I noticed that ghosts seem to like either HT or dual processor machines, don't know if that may help the code folks. I understand that there was some sort of handshake that needs to be implemented at the project to make these go away, understand Einstein did that. Looks like it is needed here. |
|
Send message Joined: 17 Sep 05 Posts: 60 Credit: 4,221 RAC: 0 |
I have NEVER had a problem with an LHC workunit. Rosetta is HORRIBLE and CPDN is alright, but the workunits take to long, and I would rather donate time to another project. I have been wanting to try Ufluids, but rumor has it that there is a possibility that groups may profit from info gained from it. Is this true? I might just detach from Rosetta and CPDN and stick with LHC and Ufluids........ See, I do not have time to backup workunits, and do other funny stuff with BOINC. I simply run the client and Crunch WU that benefit science the most. I would like to donate to a project that is located in the U.S., but my main goal is to benefit the world. LONG LIVE LHC!!!!!!!!!!! By the way, THANK YOU to everyone who have answered my questions. Everyone here is awesome! |
|
Keck_Komputers Send message Joined: 1 Sep 04 Posts: 275 Credit: 2,652,452 RAC: 0 |
While there may be some profits made due to uFluids the main gain will be for spaceflight in general. For example apollo 13 might not have had its problems if they had had that project to test things on beforehand. BOINC WIKI   BOINCing since 2002/12/8 |
Ray Murray Send message Joined: 29 Sep 04 Posts: 281 Credit: 11,859,285 RAC: 0  |
As to CPDN, I have tried a few WUs, and have not got a single one to run. I made a good faith effort to get one completed, but I keep comming home after they've run for about 50-100 hours to see they errored out. I've gone onto some other projects for now, to take it's time slice. Perhaps, with a software upgrade I might try it again, but right now, without a sucessful completion, it's just trashing WUs now as CPDN and my comp don't seem to like each other, CPDN was having problems with their sulphur model for a while but hopefully from this post on their forum it has now been resolved, and even the errored out returns are still useful. I have found that it is very fussy about overclocking. For a while I was trying to squeeze as much out of the CPU as possible but kept getting errors so cranked it back to just slightly more than standard settings. You might have to fiddle with the balance of speed versus stability for your own box. If you can do without the graphics, all the better. Setting your screensaver to "Blank Screen" will ever so slightly up your speed and setting your monitor to go off will also infinitesimally help to save the environment. All the projects I have ever connected to have had difficulties during changes to their application or upgrades of clients. Ufluids, that is mentioned in another reply, had all sorts of bother for a while but seems to have resolved those. Like Halifax Lad I'm trying 5.3.15 without any problem so far. Not that I had any problems with 5.2.15 either but I expect the next general release will be prettier. Oh how I long for the good old progress bars and piecharts of 4.19 
|
|
UBT - Halifax--lad Send message Joined: 30 Sep 04 Posts: 23 Credit: 1,372 RAC: 0 |
I have been wanting to try Ufluids, but rumor has it that there is a possibility that groups may profit from info gained from it. Every BOINC project currently out will give groups some kind of profit if they find what they are looking for on the projects Join us in Chat (see the forum) Click the Sig Join UBT |
|
Send message Joined: 17 Sep 05 Posts: 60 Credit: 4,221 RAC: 0 |
I have been wanting to try Ufluids, but rumor has it that there is a possibility that groups may profit from info gained from it. I suppose you are correct. LHC may profit monetarily from the results gained here........Thanks Halifax lad |
|
Send message Joined: 13 Jul 05 Posts: 456 Credit: 75,142 RAC: 0 |
My approach is to upgrade rarely, and when I do always to the most recent at the time of the upgrade. Every upgrade introduces some new unexpected effects - that is the nature of software development. Some are intended and only unexpected to me because I did not know the history, others are true bugs and sneak in as a side effect of some other (worthy) change. This means that every release is quirky in a slightly different way - and by leapfrogging releases I don't have to re-learn the quirks so often. Instead of tiny steps forward with a little niggle each time, I go for the occasioanl great leap forward, still with the same size of little niggle. It feels like a good balance to me. I report the bugs in the verisons I use, and let others report the bugs in the intermediate versions. That feels like a fair division of labour to me.
Absolutely so. For all of us the issue is not whether to upgrade ever, it is how often to do it. And none of us can wait for ever - eventually the client drops off the supported list, and a bit later new apps will expose bugs in the old client that nobody is responsible for addressing. I would not give advice on how often to upgrade - but what I would recommend that whenever you do it you jump to the latest recommended version for your platform. River~~ 
|
|
Send message Joined: 13 Jul 05 Posts: 82 Credit: 6,336 RAC: 0 |
----------------------------------------------------------------------- As long as you made a backup of BOINC after you downloaded the unit and before that happened all you have to do is shut down BOINC and restore tha backup. But most only do that for CPDN as they have a lot of crashes. If you have a backup from that time frame b sure not to rerun the ones that were already run, and you will louse others that you downloaded after the backup. Not worth the work for one unit, just accept the fact that it is lost forever from your system. Pizza@Home - Rays Place - Rays place Forums |
|
Send message Joined: 13 Jul 05 Posts: 456 Credit: 75,142 RAC: 0 |
----------------------------------------------------------------------- I disagree, Ray. It *is* worth the effort to save just one CPDN workunit - and furthermore the CPDN prject people positively encourage users to do this. If you did not backup that folder before, it is too late. It you did backup the folder, and if you have any non-CPDN work held locally, proceed as follows: 1) set no more work on all projects, incuding CPDN 2) abort any new CPDN wu that was downloaded automatically and has hardly started (reason it will be sitting around for a long time while your reurrescted one crunches, it is better for the project to allow another user tog et started on it) 3) let the work on all other projects complete and upload and report 4) when you have a client running with no work, stop BOINC 5) copy the backed up folder onto the BOINC folder 6) restart BOINC 7) reset all projects except CPDN (this makes sure you do not re-run work from the other projects) If you only have CPDN running on that box, steps 5 & 6 should be enough. To backup the BOINC folder on windows there are two methods A) right-click the folder, copy; go somewhere else and paste - all the files under the folder will get copied, this may take some time B) if you have winzip, or if you know how to use the zipping facility built into XP, you can create a zipped archive of the entire BOINC folder. To backup from Linux GUI, you can proeceed as for windows (A). To backup from Linux command line, use tar - man tar will tell you exactly which switches to use to create a zipped archive. Hope that helps, River~~
|
©2024 CERN