Message boards :
News :
Theory application reaches 5 TRILLION events !!
Message board moderation
Author | Message |
---|---|
Send message Joined: 1 Sep 04 Posts: 139 Credit: 2,579 RAC: 0 |
LHC@home's Theory application will tomorrow pass the milestone of 5 TRILLION simulated events. This project, under its earlier name "Test4Theory", began production in 2011 and was the first BOINC project anywhere to use Virtual Machine technology (based on CERN's CernVM system). Over the coming weeks we plan to publish some more details about all this on the LHC@home and CERN websites. Our timetables have of course been affected by the Coronavirus disruptions, but we absolutely could not miss announcing and celebrating such a milestone as this. The whole LHC@home team sends our sincerest thanks to all our volunteers for enabling this achievement !! |
Send message Joined: 24 Oct 04 Posts: 1127 Credit: 49,750,905 RAC: 9,376 |
No problem Ben, And to think I did all that on a PIII 500 single core X86 with less than 1GB of ram on XP Pro (ok I made that part up) But then ...... https://lhcathome.cern.ch/lhcathome/show_host_detail.php?hostid=27872 |
Send message Joined: 2 May 07 Posts: 2101 Credit: 159,817,517 RAC: 132,770 |
|
Send message Joined: 27 Sep 08 Posts: 807 Credit: 652,436,425 RAC: 280,287 |
I did 108 billion (1.08^11), with my best host is 23 billion. I think I merged my hosts so this is my oldest one: https://lhcathome.cern.ch/lhcathome/show_host_detail.php?hostid=9755718 |
Send message Joined: 24 Oct 04 Posts: 1127 Credit: 49,750,905 RAC: 9,376 |
Yes Toby you and I got that VB running back in 2011 ( on a Windows OS ) I still have that #1 pc running but didn't want to update to X4 so it can't even run Sixtracks here but it still runs Einstein GPU's all day. One of these days I will finally plug in my X64 disc in that beast and it will even get to use the rest of the Ram that is already plugged in.. https://lhcathome.cern.ch/lhcathome/show_host_detail.php?hostid=10447575 You have a lot more cores running here than over at the original version. |
Send message Joined: 19 Feb 08 Posts: 708 Credit: 4,336,250 RAC: 0 |
I downloaded a bunch of them on a Windows 10 PC today, with two Atlas. I do this when Science United, in which the PC is enlisted, does not download a lot of Milkyway@home tasks, both CPU and GPU. I was one of the Test4Theory@home beta testers, and received a warm hardcopy letter from Ben Segal and two others. Thanks! Tullio |
Send message Joined: 21 Jan 20 Posts: 1 Credit: 4,335 RAC: 0 |
You so funny!! |
Send message Joined: 9 Jan 15 Posts: 151 Credit: 431,596,822 RAC: 0 |
Congrats team! Looks like have done 130 billions or 2.6% out of these events. |
Send message Joined: 12 Aug 06 Posts: 418 Credit: 5,667,249 RAC: 3 |
I did 108 billion (1.08^11), with my best host is 23 billion. You're making me jealous with that 56 core Xeon. I'm about to build a 24 core one.... but for £60 it's not bad :-) I wonder if Atlas will run on all 24, or will I get two 12s? Never tried multithreaded across physical CPUs before.... |
Send message Joined: 27 Sep 08 Posts: 807 Credit: 652,436,425 RAC: 280,287 |
Works, fine my older Xeon is dual proc machine. |
Send message Joined: 12 Aug 06 Posts: 418 Credit: 5,667,249 RAC: 3 |
Works, fine my older Xeon is dual proc machine. So you get an Atlas over all the cores of both CPUs? I thought each CPU having it's own die and its own memory would severely slow things down. Or maybe the multithreaded programming is cleverer than I'm thinking. Still waiting for some of the parts :-( Got any spare backplates? |
Send message Joined: 7 Apr 20 Posts: 1 Credit: 53,683 RAC: 0 |
Unfortunately, it appears that VirtualBox and Mac OSX 10.15 Catalina do not get along. Had to uninstall VirtualBox. Guess I can only work on Sixtracks now. |
Send message Joined: 12 Aug 06 Posts: 418 Credit: 5,667,249 RAC: 3 |
Unfortunately, it appears that VirtualBox and Mac OSX 10.15 Catalina do not get along. Had to uninstall VirtualBox. Guess I can only work on Sixtracks now. If it's only for Boinc, can you install Linux or Windows on a Mac? I know people who have installed MacOS on a PC. |
Send message Joined: 27 Sep 08 Posts: 807 Credit: 652,436,425 RAC: 280,287 |
Yes, I run say 50 tasks. In assume in theory yes but seems like for LHC, more cores is better, clock speed doesn't see so important. I guess you can compare my systems "Carbon" & "Nitrogen" they are both 56 cores but the 8173 is single proc and the 2648L is dual, I'm not running any LHC at the moment though. http://stateson.net/HostProjectStats/ You can compare them with this webapp. |
Send message Joined: 4 Apr 19 Posts: 31 Credit: 3,887,310 RAC: 9,649 |
Thank you for providing such a great educational resource at http://mcplots.cern.ch/ where anyone can see how our research in particle physics has progressed so far and what is the current state. Until now I was assuming scientists has modeled the LHC experiment to a precision on the order of 1e-10 or better, but the plots and predictions at mcplots is a bit of, disappointing. It is not uncommon to see theorical prediction go completely off the experimental uncertainty, sometimes by an order of magnitude! No wonder why scientists need such a strict standard of 5σ confidence in order to announce a discovery. I have been diving in CERN publications in the hope that I can interpret these plots which I contributed to one day. But apart from authors of LHC@home team, I failed to see any "mcplots" being published. Then a big question arises: are these resources just "educational"? Has any scientist not from LHC@home team published one of these plots? |
Send message Joined: 12 Aug 06 Posts: 418 Credit: 5,667,249 RAC: 3 |
Yes, I run say 50 tasks. In assume in theory yes but seems like for LHC, more cores is better, clock speed doesn't see so important. I can't believe that. If I have a CPU with double the clock speed with the same number of cores, it will do twice the work on any project as it will take half the time. Likewise, doubling the cores at the same clock speed will allow either twice as many WUs to run at once, or multithreaded ones to run twice as fast. Of course there's also changes in instructions sets with newer CPUs.... I still don't have my 24 core system running, I'm waiting on one single part, the front panel so I can switch it on! Can't be bothered finding the pinout of the board's connector to power it up manually. But I notice whenever virtual machine is running, my machine I'm typing this on is a little sluggish, much more so than with non-VM tasks. I wonder if that is significantly worse if 24 virtual machines are running, or whether the slowdown would be the same? Not that it matters for the usability, as the 24 core machine will only run Boinc. As long as it doesn't slow the OS down too much so it can't feed the tasks from the disk or internet fast enough. |
Send message Joined: 4 Apr 19 Posts: 31 Credit: 3,887,310 RAC: 9,649 |
Yes, I run say 50 tasks. In assume in theory yes but seems like for LHC, more cores is better, clock speed doesn't see so important. Believe it or not, I see identical credits/elapsed time ratio when I underclock my CPU to 800MHz from its stock frequency of 2500MHz. It seems LHC@home gives credits only based on your total time and BOINC benchmark. Your actual contribution to LHC@home can be seen at mcplots website. So if you want more science results, boost all your cores and improve your network speed, or even better install linux inside your VirtualBox and run natives there. But if you just want more credits/W, underclock your CPU, make your network slower, and run as many tasks in parallel as possible without regard to cache performance. |
Send message Joined: 12 Aug 06 Posts: 418 Credit: 5,667,249 RAC: 3 |
Yes, I run say 50 tasks. In assume in theory yes but seems like for LHC, more cores is better, clock speed doesn't see so important. It's about the science not the points [shakes head in disgust] |
Send message Joined: 4 Apr 19 Posts: 31 Credit: 3,887,310 RAC: 9,649 |
The jet charge plots at http://mcplots.cern.ch/?query=plots,ppppbar,jets,jet_charge look so bad that all of these generators are off by huge differences. However the referenced experimental article from CMS said predictions of PYTHIA8 and HERWIG++ matched almost perfectly with experiment results. Quite strange. http://cms-results.web.cern.ch/cms-results/public-results/publications/SMP-15-003/ |
Send message Joined: 4 Apr 19 Posts: 31 Credit: 3,887,310 RAC: 9,649 |
Rivet currently has 991 analyses, see https://rivet.hepforge.org/rivet-coverage . However only 110 are available on mcplots, while it is stated it is almost automatic to add a Rivet analysis to mcplots. Are there any plans to add all of them? Seems only some basic information needs to be filled for each paper. |
©2024 CERN