 |
|
|
21)
Message boards :
Number crunching :
HADAM3P 6.06 on One Core of 4-Core Computers?
(Message 36823)
Posted 27 Apr 2009 by  Pete B
Post: Yeah, if I'm running just one model with other projects, I can get an HadAM3P model done in 48 hours of crunch time. That's a far cry to my old system that took 6 months to run it's first HadCM3. Well, that's better than my original system for this project that took 4 months to run it's first HadSM3, albeit a 4 phase one so 1 month per phase. My current top system, a Phenom 9950 @ stock 2.6GHz is at present running 1 HadCM3 160 yr run (2.03 s/ts), 2 HadAM3P's (3.9 s/ts) & is just about to complete a HadSM3 (1.7 s/ts) in a Linux VirtualBox after which it will begin a HadAM3P in the Virtual machine too. In the current setup, it takes about 75 hrs to do the HadAM3P's but I haven't tried one HadAM3P on its own. No doubt with the AM3P running in the virtual machine, they will all slow down somewhat though. |
|
22)
Message boards :
Number crunching :
HADSM3-MH Crash and Re-Set?
(Message 36662)
Posted 9 Apr 2009 by Pete B
Post: My computer froze and I had to re-boot. I was 20% into my HADSM3-MH model, but then when I re-booted the model went back to less than 1% completed. Nothing will happen when it trickles up, the server will accept it but it won\'t show as a result because it\'s already had it. The trickles will start showing again once the model reaches the trickle point following the last one already sent. Just check with the running HadSM3-MH graphics/timings though make sure it\'s running with normal graphics (not a blue \'iceworld\') and that the timesteps are the same or nearly the same as previously. If all this is OK, then the model should be OK & eventually get past the point it had previuosly got to. It\'s always worth taking an occasional backup of the BOINC directory, this then can be used to restart from that point should there be a loss of one or more WU\'s for any reason. EDIT: Looking at your results, it appears you\'ve aborted them now anyway? |
|
23)
Questions and Answers :
Unix/Linux :
Graphics in Linux HadSM3 BOINC 64 bit Client enabled version
(Message 36659)
Posted 9 Apr 2009 by Pete B
Post: I will deprecate the 64 bit 5.x apps. This should resolve it. Thanks Tolu :-) |
|
24)
Questions and Answers :
Unix/Linux :
Graphics in Linux HadSM3 BOINC 64 bit Client enabled version
(Message 36657)
Posted 9 Apr 2009 by Pete B
Post: Pete, George, this is indeed the case. When I checked the stderr file, I had the same error messages as Cespenar. \'Retrograding\' to the latest of the 5.x.x series of BOINC Client 64 for Linux and the graphics are there so unlike the Windows version of BOINC 6.x.x which could show graphics in ver 5.x WU\'s (but not the other way around) the Linux version of BOINC 6.x.x can\'t. As Cespenar points out, the graphics are not necessary but I often check them once/twice a day as a good way of checking model performance at a point in time for \'cold equators\', \'iceballs\' etc although the latter would soon become evident through resultant changes in the timings. I wonder how many are actually running this combination (BOINC 6.x.x 64 Client in a Linux 64 OS & running \'old\' 5.10 HadSM3\'s, as a result) obviously Cespenar & myself but probably not that many overall hence no previous reference to the problem. I\'ll send a PM to Tolu as to whether the 5.10 WU still has to be kept for 64 bit BOINC client enabled HadSM3\'s now the 6.x versions are compatible. |
|
25)
Questions and Answers :
Unix/Linux :
Graphics in Linux HadSM3 BOINC 64 bit Client enabled version
(Message 36645)
Posted 8 Apr 2009 by Pete B
Post: I\'m currently running a HadSM3 version 5.10 in Linux Ubuntu 8.10 64 bit using the BOINC 6.4.5 64 bit Client on an AMD Phenom 9950, 2Gb RAM. The model is running fine, now in Phase 2, but there are no and never have been, graphics available. The graphics button is not greyed out and is \'active\', i.e. reacts to the normal mouse action when the mouse pointer is over it but no graphics window at all is displayed. The graphics in other model types (HadAM3P, the beta tests etc) all work perfectly in the same setup. All required 32 libraries are installed & I\'ve tried the \"xhost +local:\" command to no avail. It is not the transarent background issue others have seen as there is no visible window. Anyway, because the graphics element of other experiments work in the same setup, it makes me wonder whether graphics were enabled in this Linux 5.10 version of HadSM3 enabled for the BOINC 64 bit client. Anyone else seen this issue? TIA |
|
26)
Questions and Answers :
Unix/Linux :
Ubuntu 8.10 running in a Sun VirtualBox VM
(Message 36504)
Posted 27 Mar 2009 by Pete B
Post: I have set up Ubuntu 8.10 64bit running in a Sun VirtualBox virtual machine under Windows XP Pro x64 host OS on a Phenomx4 9950, 4Gb RAM PC. The VM has 1.5Gb of the RAM allocated to it. VM\'s can only \'see\' & use 1 processor, even under multiple core physical CPU\'s. I am running 3 models in Windows in BOINC Client 6.2.19 64bit (1 HadCM3 160yr & 2 HadAM3P\'s), with 1 Linux HadSM3 5.10 running under Linux BOINC Client 6.4.5 64bit, a total of 1 model per core overall. Other than the well documented 32bit library issue with running 32bit applications in 64bit Linux BOINC clients, which I soon resolved, everything is running great apart from what seems a minor issue. The 3 Windows models report proper CPU time figures but the VM Linux model reports a time 4 times faster than normal, see here. It looks very impressive on the s/ts figure but the real CPU time is about 1.6 s/ts. It\'s as if a setting somewhere is causing the real CPU time to be divided by 4 giving the seemingly impressive timings at software level due to the only 1 CPU issue in VM. I\'m new to VM\'s & Linux & I\'m not sure whether it\'s a VirtualBox or a \'BOINC in VM\' issue but I\'ve tried setting the specific PC settings in the Linux BOINC Client to either 25% of CPU time or 25% of processors when using multi processors to no avail. I can\'t see any other obvious settings in the Ubuntu system settings, VirtualBox or BOINC. Any ideas? Thanks. |
|
27)
Message boards :
Number crunching :
HadCM3 Performance
(Message 36291)
Posted 4 Mar 2009 by Pete B
Post: AMD Quad core Phenom 9950BE @ stock 2.6GHz, 4GB 800MHz RAM, MSI DKA790GX Motherboard - HTT @ 200MHz, PCI-E GFx card, Windows XP Pro 64bit, BOINC 6.2.19 - Homebrewed Tower system: 2.09 s/ts for 1 HadCM3ist_ model running in conjunction with 3 CPDN Beta HadAM3P runs. The same HadCM3 run rate improved to 1.98 s/TS when the 3 Beta runs had completed and were temporarily replaced with 3 less CPU/RAM intensive, short time to completion runs from the \'World Community Grid\' Project (BOINC version) Therefore, there can be a performance hit when running alongside other more CPU &/or RAM intensive runs. EDIT: It is probably worth also including motherboard (or at least chipset type and GFx type, i.e. onboard or AGP/PCI-E) since these can aslo affect run rate with otherwise identical components. If not known, say on a prebuilt system, the BIOS or CPU-Z will give the information. |
|
28)
Questions and Answers :
Windows :
Windows 7, 64-bit?
(Message 36178)
Posted 19 Feb 2009 by Pete B
Post: New \"cruncher\" is underway (i7-920, overclocked). Will BOINC/CPDN work on the OS mentioned? Yes it will, I have been running various models in BOINC 6.2.19 under Windows 7 Beta 64bit without problem. Trickles, final uploads & downloads all have taken place without any issues. Using dual core AMD Athlon 4800+, @ stock speed + 2Gb RAM. |
|
29)
Questions and Answers :
Windows :
move CPDN WU to faster computer
(Message 17565)
Posted 29 Nov 2005 by Pete B
Post: Yes, as long as it is exactly the same OS i.e Win XP > XP or 2k > 2k etc. Install the same version of BOINC on the new PC in the normal way. Don\'t attach to a project or anything, just delete the new entire BOINC directory in \'Program Files (assuming Windows). Copy the entire BOINC directory from \'Program Files\' on the old PC with the WU in it to same directory on the new one. Start the BOINC manager and all should be OK. That\'s all there is to it. The next auto update will record the new PC information with the server. Pete |
|
30)
Message boards :
Number crunching :
RAC (Or lack thereof)
(Message 16080)
Posted 16 Sep 2005 by Pete B
Post: Hi I was taking my information from here, here and here. Taking those and your refs also, it is a very complex issue. No doubt the rate of melting of fresh water ice needs to be accounted for and also over what timescale it may enter the seas. If the melting Arctic Ice and Greenland glaciers slowly trickle into the wider open waters, then the effect may be different to if the fresh water is dammed back until it reaches a \'breakout\' point at which a mass of fresh water suddenly \'floods\' the sea surface. another point, is there even enough water locked up in the Greenland glaciers to put sufficient fresh water into the system? The Younger Dryas with which it is compared, as far as I understand, involved a much wider area of melting ice that Greenland. pete |
|
31)
Message boards :
Number crunching :
RAC (Or lack thereof)
(Message 16046)
Posted 16 Sep 2005 by Pete B
Post: Hi With that containing the ocien currents it would be interesting if some included what would happen if enoughf of the Greenland ice sheet melted and stopped the North Atlantic Convayer (Galf Streem). But that would have to include Currents turn-over, wow would that be a long one. Since, apparently, some (but not all) of the supercomputer runs have predicted this then my guess is that it will be inevitable that some CPDN coupled forecast runs will produce the THC slow/shutdown scenario. It will all depend on the parameters used. What will be interesting though is the proportion of \'good\' (based on the hindcast matching with reality) sets that model it, and having done so, what hypothetical climate changes follow on in any model that does show it. Pete |
|
32)
Message boards :
Number crunching :
RAC (Or lack thereof)
(Message 16023)
Posted 15 Sep 2005 by Pete B
Post: Hi The coupled models will be 50 yrs long, one single phase each. Hindcast - 1950 to 2000, forecast - 2000 to 2050. It\'ll sort out the stable PC\'s OK but I would have thought it will also be a severe test for HadCM3.. Starting in 1950, I wonder if even the best matches of hindcast will pick out the Globally cold years such as 1964, the hot ones such as 1998 or an even harder test, more locally pick out the notables such as the NH winter of 1963 or the European hot summer of 1976 and no doubt similar local notables in other parts of the World. I would have thought that there would have to be some good resemblance to this if the forecast results are going to be sold to the skeptics in business and various Countries Govts. An interesting stage of the experiment, it will most certainly be for for more than one reason. Pete |
|
33)
Message boards :
Number crunching :
RAC (Or lack thereof)
(Message 15769)
Posted 7 Sep 2005 by Pete B
Post: My main PC (3.2GHz P4, H/T on) is running HadSM3 v4.13 at ~2.7 s/TS alongside sulphur at ~4.7s/TS, ie sulphur 70-75% slower than slab when running together. Pete |
|
34)
Message boards :
Number crunching :
Changing from hadsm 4.12 to 4.13
(Message 15222)
Posted 19 Aug 2005 by Pete B
Post: Hi there Just for the record, Amy (3.2GHz P4, H/T) has successfully been running 1 HadSM3 v4.13 model alongside 1 Sulphur Beta test v4.17 model for many trickle points now. Sulphur model well into Ph 2, 4.13 will reach Ph 2 in within the next 36 hrs. Pete |
|
35)
Questions and Answers :
Preferences :
Can I delete this map after the project has finished?
(Message 14894)
Posted 3 Aug 2005 by Pete B
Post: Apart from archiving for the reasons given in the "CP Data Retention" thread, having the archives of one's previously completed models may occasionally help in other ways. Having a personal archive of my 17 completed BOINC runs (and numerous Classic ones as well) on my hard drive went some way in helping the administrators to sort out broken links in the database between completed runs and machines that had done them a few days ago. Pete |
|
36)
Message boards :
Number crunching :
Massive amount of credit lost - What's up?
(Message 14787)
Posted 30 Jul 2005 by Pete B
Post: > > Bugger, so I've lost nearly 2500 credits (if I've read it correctly)? > <b> Hopefully not, mines gone from 123,000 originally (all made up of properly allocated models when I checked my own data record here at home) first, down to 102,000 and then right down to 28,000 but I've PM'd Carl with the "lost" ones of mine which hopefully should be a bit of info help him to track down the problem for everyone and get things back. Pete |
|
37)
Message boards :
Number crunching :
I Love this Project...!!!
(Message 14723)
Posted 29 Jul 2005 by Pete B
Post: There does seem to have been some random problem with recalculated credits!! Between 6:30pm on Fri 29/07 and now (Sat 30/07) my credits have inexpicably dropped from ~123,000 to ~102,000!!! If I was still in the "United Kindom" team, I would have dropped 5 places compared to the surrounding team members as a result of this!! All those I had, as far as I worked out during the whole time on the project were correct on the basis of what had been run and completed and nothing appeared "wrong". Perhaps with this continual credit problem over the last few months with the project, it is high time to scrap this idea of credits and just count the number of runs completed as some kind of token of work done for CPDN. What's the point of "credits" when they are so fickle and meaningless as a result of various "recalculations" Pete |
|
38)
Questions and Answers :
Windows :
Transition from hadsm 4.12 to 4.13 failed
(Message 14635)
Posted 24 Jul 2005 by Pete B
Post: Hi there crandles asked: "Also has anyone successfully downloaded Windows hadsm3*_4.13_* files?" Yes, unintentionally so in my case. When I re-set Amy back to "use 2 CPU's for BOINC" on Friday, ready to set up the Sulphur Beta test, I successfully downloaded and started a Sulphur model, then suspended the already running HadSM3 4.12 model on the std project just to see how fast the Sulphur Beta would run on its own. The response was for BOINC to think it was short of work and download a new HadSM3 4.13 and start running it. I immediately suspended it and so down came another one the same. I could only stop by setting both projects to download no more work and suspend BOINC network access so I could suspend the std model without new ones coming down. I aborted the susbequent downloads to prevent having too many waiting to run in future but the first to download is held in suspension for the time being until the existing std run is finished. Although the files show the deliberate aborting via the GUI "error", the download info is there as well if required by anyone. Pete |
|
39)
Message boards :
Number crunching :
Incorrect credit allocation
(Message 13862)
Posted 26 Jun 2005 by Pete B
Post: Hi there If you are doing a run that appears to be allocating credit to another machine, it looks as if you possibly have a new variant of the problem in this <a href="http://climateapps2.oucs.ox.ac.uk/cpdnboinc/forum_thread.php?id=2577">thread</a> I see from the results table what you mean, up to Ph2, T/S 10802 on 19/06, it's allocated to your machine 321, threafter from Ph2, T/S 21604 on 20/06, the same Result#ID is now being allocated to <a href="http://climateapps2.oucs.ox.ac.uk/cpdnboinc/trickle.php?resultid=872375">machine ID5336</a> belonging to <a href="http://climateapps2.oucs.ox.ac.uk/cpdnboinc/show_user.php?userid=2790">BakerB</a> The correct interpretation and solution can only come from the Project IT guys Pete |
|
40)
Message boards :
Cafe CPDN :
CPView for BOINC
(Message 13849)
Posted 25 Jun 2005 by Pete B
Post: MartinSykes31 said: "Hi there, haven't been around for a while but I'm back now..." Hi there Martin, nice to see you back again in the group and hope all is well with you :-) I will look forward to "playing" with the new version of CPView later on this evening. Pete |
©2024 climateprediction.net