Posts by old

21) Message boards : Number crunching : Multi-thread models... (Message 7933) Posted 28 Jan 2005 by old_user156 Post: Heh, how do I set big blue text on this messageboard..? ;-) Pleeeeease come back phpBB, all has long since been forgiven. :lol: +++ Now that simultaneous mutlthread capable machines are becoming the norm', eg. Intel Hyperthreading P4s, and the future of maintaining 'Moore's law' for desktop machines appears to be multi-core CPUs - how about looking at how easy it would be to run a single instance of a model with multiple threads..? This modelling software was originally designed to run on a multicore Cray, so it would actually involve changing things backward some way, wouldn't it..? The CP-boinc software needs to evolve in the same direction as standard desktop machines methinks, especially if you desire to increase the computation load........ We will eventually want 64 bit software too but M$ needs to catch up first I guess. ;-)
22) Message boards : Number crunching : Same parameters different results (Message 7926) Posted 28 Jan 2005 by old_user156 Post: > (1) We're intending to put together a "bumper issue" of the cpdn paper (or > maybe an extended mix - not sure how we're supposed to phrase these things...) > that will include the supplementary info as well as the paper, and maybe > Myles's '99 Nature piece when cpdn was just a pipe dream. And it will have a > pretty cover (from the Nature mock-up). I Look forward to reading it &, very probably, printing it out 'for posterity' too. ;-) > (2) I thought we'd already said that we wanted to explore the cold equator > models in the coupled ensemble? My memory of the open day is getting a little hazy now - I do recall one of my quaeries being answered in that these 'cold equator models' wouldn't be arbitrarily ruled out for the coupled model but the comment in 'Nature' is more explicit than that I think. > I think we're all quite keen that we crack on > with a few papers now Us users out in the community are quite keen on reading more 'results' too methinks - nice to see some properley done analysis of our computer's work. ;-) > - some really basic ones are: (a) the duplicates paper > (Sylvia); (b) the quality control paper (Jamie and Stephen at RAL); > observational comparisons (Claudio); the cold equator paper (Nick or me, > probably) and a sort of ic paper, but that might take a bit of time till we > get the big ic ensemble back - Dave S is quite keen on that one, I think. And > the non-linearity paper, which a few of us are keen on. And then there is all > the other research (Nick's THC stuff, Duncan & Ellie's s-cycle, some ocean > spinups...) too. And some of our collaborators have some cool ideas, too. :-D
23) Message boards : Number crunching : BOINC 4.19 (Message 7925) Posted 28 Jan 2005 by old_user156 Post: I upgraded 'Alison' (AMD Athlon, WinXP Home) and 'Caroline Nova' (Dual Intel Xeon, WinXP Pro) to BOINC v4.19 from BOINC v4.13 yesterday without any difficulty and all their models are still running okay. :-)
24) Message boards : Number crunching : BOINC 4.19 (Message 7818) Posted 27 Jan 2005 by old_user156 Post: Since v4.19 is now the version of BOINC available from the <a href="http://boinc.berkeley.edu/download.php">SETI@home BOINC download web page</a> for Windows, I'll give it a try here too...
25) Message boards : climateprediction.net Science : TDAT... (Message 7814) Posted 27 Jan 2005 by old_user156 Post: Just now watched 'the day after tommorrow' on DVD, as was written at the time of it's release on the phpBB; Great special effects but the scenario sux... ;-)
26) Questions and Answers : Windows : HT & CPDN - Let me see if I understand it... (Message 7786) Posted 27 Jan 2005 by old_user156 Post: edited out...
27) Message boards : Number crunching : Same parameters different results (Message 7781) Posted 27 Jan 2005 by old_user156 Post: ....<b>> However, it will be important not to omit such parameter combinations from > future experiments using models with dynamic oceans since they would not > exhibit the same problem.</b> > Very interesting, thanks Chris, I was wondering how I was going to be able to read the supplements without subscribing to 'Nature'. The final paragraph that I've highlighted above is good to know - we were saying much the same as that at the CPDN open day, and somewheres in the long 'cold equator' thread on the phpBB too but hadn't really heard any response until now. :-)
28) Questions and Answers : Windows : HT & CPDN - Let me see if I understand it... (Message 7660) Posted 26 Jan 2005 by old_user156 Post: > I have a P4 with 2,8 GHz and HT. I am running CPDN on one CPU and the > Seti Classic on the other (emptying the queue). > I am considering to run CPDN only and use HT for that. After reading > through the FAQ, Forum and other sites, I am still a little puzzeled, > so let me see if I got it right: >1) I can run CPDN under BOINC 4.13 on a computer with HT. And (usually) it >does not crash. (There were some earlier messages somewhere where it did.) I have two dual Xeon HT machines running four models each, they're doing fine to date. Many others are running two models on single physical CPU HT machines without any problems. > 2) If I do that, my computer works on 2 WU instead of one, so I need twice > as much disk space (2x 700 MB), too. Yes, although all you <i> have to </i> have is 600MB of <i>free space</i> anytime you resume either model AFAIK. > If the above is right, I have more questions: > a) If I run CPDN only, will HT actually accelerate the calculation? Nope, both models will run slower but your overall throughput will increase, ie. 2 models in 30 days instead of 1 model in 20 days, those sort of figures. > b) If I run CPDN only, will the WUs still be swapped every 60 minutes > (or whatever time intervall I set under preferences)? What/How will > they be swapped, e.g. switch processors? Nope, they just run all the time - I've found I get a slight increase in performance if I use 'Task manager' to set the affinity of each model to a single CPU but it is a very small increase. > c) Is there a way to temporarily (e.g. 30 minutes) to suspend calculation > for just one of the two? Yes, effectively; Set affinity of both models to a single CPU, ie. rightclick on a process in 'task Managaer' (xxx_um_xxx) and set affinity to the same CPU - this will leave you one CPU completely free - I've just now tested this on one of my dual Xeon machines & it works. Both models will still run but at roughly half speed whilst limited to using a single CPU for both.
29) Message boards : Number crunching : CP-boinc model run times. (Message 7642) Posted 26 Jan 2005 by old_user156 Post: Cool models run faster, I have no idea why...
30) Message boards : climateprediction.net Science : Rotating globe in graphics view? (Message 7541) Posted 24 Jan 2005 by old_user156 Post: Les: If you grab a 'sparkly' globe with your mouse pointer, turn it a little in the North~South direction and then let go, it should stop 'sparkling' - you're correct BTW, it uses less CPU whilst stationary but, it also depends on the age of your graphics card, some 'oldies' use close to 100% of the CPU whenever the graphics are running.
31) Message boards : Number crunching : CP-boinc model run times. (Message 7539) Posted 24 Jan 2005 by old_user156 Post: Yeah, I've finished a bunch more models but I'm very busy assembling new machines, modifying my current 'harem' and rebuilding 'oldsters' as hand-me-downs for my younger relatives - I'll get around to posting more up-to-date data eventually. ;-)
32) Questions and Answers : Windows : What is 'isass.exe'..? (Message 7319) Posted 16 Jan 2005 by old_user156 Post: You're correct mavau, lsass.exe it is - it was taking full 50% of the avaialble CPUs but a reboot fixed it, thanks anyways... See <a href="http://www.climateprediction.net/board/viewtopic.php?t=2695&start=0">here for the full setup</a> of that particular machine. :-)
33) Questions and Answers : Windows : What is 'isass.exe'..? (Message 7315) Posted 16 Jan 2005 by old_user156 Post: Hi all, what does 'isass.exe' do..? It's a system process that's taking full 50% of my new dual Xeon system's CPU..?!?!? I had look around at M$.com but they only point you at 'sasser worm' and this is a clean install of WinXP Professional with SP2 laid on top from the M$ CD..! (It searches clean of infection too anyway.)
34) Questions and Answers : Windows : Can\'t find host record (Message 7219) Posted 8 Jan 2005 by old_user156 Post: 'Merge' should end up with only the most recent host (highest number) showing - you don't get any choice in which one is displayed...
35) Questions and Answers : Unix/Linux : Recommend a Linux distro for a complete newbie please..? (Message 7177) Posted 5 Jan 2005 by old_user156 Post: Hi Guys Recommend me a Linux Server distro for a complete newbie please - I need something that will run a Dual Nocona (EM64T) Xeon server / workstation for me with 'proper' server functionality. Eg. Standard WinServ03 (Â£458 OEM, bit 'ouch' for what will be mainly a home LAN server..! :-( ) can handle streaming video, plus a whole shedload of other stuff I'd probably never use but it does tweak my interest simply because it's in there... M$ are now selling a version of WinServ03 called 'web server edition' at a not totally unreasonable price <i>except</i> it is so cut down in functionality one might as well say they pillaged it first. :-(
36) Message boards : Number crunching : Oddities happen (Message 6955) Posted 15 Dec 2004 by old_user156 Post: I happen to agree with you Jim and wish we still had a count of properley completed models displayed in our stats. One can add them up in your 'results' pages by looking for initially 5444.21 & laterly 6805.26 cobblestones credit awards for any particular result. Boinc server errors becsue of restoration from backup or other reasons can mean the 'Outcome' column is occasionally incorrect - much the same as Classic occasionally miscouted a 'full run'- I did something like 120 completed full runs under 'Classic CPDN' but 'only' have 101 reported complete in my Classic stats. If models were only crashing and reporting back when they were actually non viable parameters instead of mostly (IMO) due to machine error, then _any_ run completed at whatever stage of computation, is IMNSHO, effectively a 'full run' but the line between stopped early due to machine error and stopped early due to parameters would be rather difficult to detect whilst we're running these models - post analysis of all the uploaded data ought to be able to pinpoint the exact line between the two cases more reliably.
37) Message boards : Cafe CPDN : Thread Closed (Message 6952) Posted 15 Dec 2004 by old_user156 Post: You're welcome. :D
38) Message boards : Cafe CPDN : Thread Closed (Message 6949) Posted 15 Dec 2004 by old_user156 Post: A1) About an hour flying around gently - I'm using relatively high compression ~10cc engines of European manufacture that aren't bad for fuel consumption. A2) Micro-video cams have been put aboard micro-electric helicopters that you can fly in your living room. Even the smallest Methanol fuelled I/C engine helicopters currently readily avaialable ie. 0.3 cubic inch or a '30' powered helicopter could easily lift a mini handycam - the major difficulty is vibration supression. Any of my bigger machines could very probably pick up a Professional VidCam but they would be almost unflyably sluggish and run out of fuel very quickly. Other folks with a personal interest in photography from the air have worked all this out already - try posting the same quaestion in the rec.models.rc.helicopter newsgroup on usenet - they're a friendly bunch and you'll get much firmer more up-to-date information on the current 'state of the art'; <a href="http://www.quickuk.co.uk//">QuickUK</a>, my sponsors for quite a while in the 90s, also do aerial photography professionally.
39) Message boards : Number crunching : sec/tstep increase (Message 6878) Posted 12 Dec 2004 by old_user156 Post: I'm currently following my own advice with 'Dilly' - she's upchucked a whole bunch of models these last few days - so she's running her latest model along with Prime95 at priority 4 thus they're both using 50% of the CPU...
40) Message boards : Number crunching : Weird that this result got credit for a full run..? (Message 6877) Posted 12 Dec 2004 by old_user156 Post: I have one machine, <a href="http://climateapps2.oucs.ox.ac.uk/cpdnboinc/show_host_detail.php?hostid=76562">'Dilly'</a> that has upchucked on the majority of her 8 results in her seond incarnation, (merge is not available to Dilly Mk2. :lol: ) She upchucked <a href="http://climateapps2.oucs.ox.ac.uk/cpdnboinc/result.php?resultid=427444">one result late last night in only 4.5 minutes..!</a> I guess I was more than a little too enthusiastic overclocking her back to much the same settings she had handled okay whilst watercooled... Dilly's only hobbling along at the moment, she's fitted with a lousy overclocking Tbred B 2400+ that is only air-cooled but on yet another new Abit NF7 whilst I sort through some watercooling gear plus assembling & 'tweaking' my new dual Xeon - 'Caroline' MkX will be the #1 priority for quite some time methinks... :D The Barton 2500+ that Dilly was running refused to reboot after a shutdown yesterday morning - I dunno what's wrong with it - maybe Barton's suffer from 'Northwood Sudden Death Syndrome' too but after a longer period of time overVolted. I don't push OverVolting toooo far anyways - it increases the power you have to dissipate through the cooling system rather a lot for very little extra speed anything much above 10% OverVolt for the particular core, or thereabouts. (I OverVolt my spec 1.45Vcore XP-M CPUs by up to 28% (1.85Vcore) but that is okay for a CPU that is underVolted at spec' IMNSHO - I was running one 1.65V spec Barton at 1.9Vcore but I've backed her off to 1.85V with Dilly's Barton going 'tits-up' for no apparent reason. 1.85V is only 12% above the 1.65Vcore spec' for many higher-end Barton & Tbred XPs. That darn Barton took an NF7's BIOS flash chip with it too when it died. :-( I've got a few spare lower-end CPUs but no spare BIOS chips - I'm running Tracy with her BIOS chip extracted and fitted to Marion at the moment. :lol: (I always setup the BIOS to shadow copy into RAM at boot 'cause RAM's faster than flash memory. ;) I had noticed that the 'Top Computers' page with my quickest machine on (Emma, Host #140, is currently in 77th place.) ie.61~80 <b>weren't</b> displaying properley - I wonder what <b>was</b> up with that..? (It's fixed as of now - thanks CP website guys. :-) )

Previous 20 · Next 20

Posts by old_user156