how long does your system take to do 1 frame with folding?

Discussion in 'Distributed Computing' started by blue&whiteman, May 26, 2004.

  1. blue&whiteman macrumors 65816

    blue&whiteman

    Joined:
    Nov 30, 2003
    #1
    I have noticed a pattern in that my system always seems to take about 50 min to do 1 frame. its a G4/500 cpu.

    could others please post their times and cpu's so I can get a better idea by comparison.

    thanks :)
     
  2. bousozoku Moderator emeritus

    Joined:
    Jun 25, 2002
    Location:
    Gone but not forgotten.
    #2
    It all depends on which protein variation your machine is calculating.

    More information is required to give an accurate response. :)
     
  3. blue&whiteman thread starter macrumors 65816

    blue&whiteman

    Joined:
    Nov 30, 2003
    #3
    Protein: p638_L939_K12M_ext
    - Run: 1979711488 (Clone 1358954496, Gen 587202560)
    - Frames Completed: 76, Remaining: 324
    - Dynamic steps required: 4050000
     
  4. bousozoku Moderator emeritus

    Joined:
    Jun 25, 2002
    Location:
    Gone but not forgotten.
    #4
    My G3/800 takes about 29 minutes and my dual G4/800 (each processor) takes about 38 minutes per frame for the p638.
     
  5. Mustang Mac macrumors regular

    Joined:
    Apr 1, 2003
    #5
    My iMac G4 LCD/800 runs this unit at about 36 minutes per frame. Of course, even among proteins of the same unit variations are seen in run/clone/gen numbers. Also, is this a dedicated folding machine or do you use it for other processor intensive tasks? This will affect frame times as well. The unit you are running is a Tinker unit. Tinkers are not so hot when running on G3 and G4 processors. They take an eternity because no Cobalt compiler is available to enhance the Tinker core code for Macs. Yes, Tinkers are bad on PCs, but Macs are much worse. My iMac estimates completion at 10 days, so your G4/500 will take even longer to complete the unit. Keep on folding on!
     
  6. MrCommunistGen macrumors regular

    MrCommunistGen

    Joined:
    Mar 14, 2004
    Location:
    "Wherever you go, there you are..."
    #6
    About 1hr 10mins

    I folded a couple proteins before I gave up folding (i couldn't ever use my iMac when I wanted to) but it took about an hour and 10 minutes per frame on my G3 iMac 400Mhz with 512MB Ram.
     
  7. blue&whiteman thread starter macrumors 65816

    blue&whiteman

    Joined:
    Nov 30, 2003
    #7
    for now this is my only system so I do use it for cpu intensive things throughout the day. for at least 2/3 of the day and all evening folding gets the whole cpu.

    I know a G4 isn't the best cpu for folding but since folding uses altivec i'm sure a G4 must get a decent amount more done over a G3.
     
  8. bousozoku Moderator emeritus

    Joined:
    Jun 25, 2002
    Location:
    Gone but not forgotten.
    #8
    Cobalt compiler? Where did you get your mis-information?

    The Tinker core was written in FORTRAN, a mathematically-oriented language developed in the 1950s. There are FORTRAN compilers for Macintosh, but they're typically rubbish, as well as expensive. I haven't seen any news that Stanford has used IBM's XLF compiler to get any more speed out of Tinker, unfortunately. They experimented with the beta release but no news since then.
     
  9. Mustang Mac macrumors regular

    Joined:
    Apr 1, 2003
    #9
    Sorry. My bad. Too many late nights and too little coffee. ;) You can read more about it at the Folding Community forum . It is quite interesting. How do you spell relief? F-O-L-D
     
  10. bousozoku Moderator emeritus

    Joined:
    Jun 25, 2002
    Location:
    Gone but not forgotten.
    #10
    I understand. Sorry to be so mean. As long as I've been writing applications, I've been hearing about a Cobalt compiler. Unless someone has developed a language called Cobalt, there is no such compiler. :D

    It's just unfortunate that IBM wouldn't be interested in donating some compile time with their FORTRAN compiler to increase the PPC version of the Tinker core.
     
  11. aus_dave macrumors regular

    Joined:
    Jan 6, 2004
    #11
    It varies a lot with the protein etc. but here's some data from a log file on my dual 1.8 G5, 1 GB RAM:
    Code:
    [00:27:47] Protein: p520_BBA5_pf
    [00:27:50] Writing local files
    [00:27:52] Completed 0 out of 500000 steps  (0)
    [00:35:58] Writing local files
    [00:36:00] Completed 5000 out of 500000 steps  (1)
    
    That's around 9 minutes a step, some others take 4-5 and the Tinkers are slow even on a G5.

    I gave up folding as it was making the office too hot, I might start up again now it's winter :). I'm not convinced that the extra power consumption is negligible with Folding though.
     
  12. Mustang Mac macrumors regular

    Joined:
    Apr 1, 2003
    #12
    Maybe at some point. :D And I thought BASIC was fun. :p These fellows seem to be having a great time! Keep up the fold!
     
  13. bousozoku Moderator emeritus

    Joined:
    Jun 25, 2002
    Location:
    Gone but not forgotten.
    #13
    Yes, the fake language for demonstration. :D That brings back memories. I haven't been part of ACM for nearly 10 years. :eek: I'm just so used to people saying Cobalt for COBOL.

    Anyway, has anyone noticed anyone noticed that folding is slightly faster with 10.3.4?
     
  14. Dreadnought macrumors 68020

    Dreadnought

    Joined:
    Jul 22, 2002
    Location:
    Almere, The Netherlands
    #14
    Also have a dual G5 1.8 and a P638 takes about 9 minutes per frame per proc. So 9*400 = 3600 minutes = 60 hours = 2,5 days non stop folding!
     
  15. krimson macrumors 65816

    krimson

    Joined:
    Oct 29, 2003
    Location:
    Democratic People's Republic of Kalifornia
    #15
    7 minutes on P257... i think..

    edit: yay, i have 1 WU completed. :)
     
  16. bousozoku Moderator emeritus

    Joined:
    Jun 25, 2002
    Location:
    Gone but not forgotten.
    #16
    I'm going to send you all my Tinkers. :D
     
  17. blue&whiteman thread starter macrumors 65816

    blue&whiteman

    Joined:
    Nov 30, 2003
    #17
    on frame 113 now. these tinkers sure are huge!
     
  18. Dreadnought macrumors 68020

    Dreadnought

    Joined:
    Jul 22, 2002
    Location:
    Almere, The Netherlands
    #18
    No thanks, I get enough of them as it is! And I can get a lot more points with normal gromacs, if they are 44 pointers, if you get the 20 pointer range gromacs, those WU's kill your daily production.
     
  19. blue&whiteman thread starter macrumors 65816

    blue&whiteman

    Joined:
    Nov 30, 2003
    #19
    this tinker i'm working on must have big frames. even when folding has the cpu to itself its taking about 1 hr 10min to get a frame done. other tinkers frames have been done as fast as 40 min on my cpu.

    btw its p638_L939_K12M_ext that i' still working on.
     
  20. aus_dave macrumors regular

    Joined:
    Jan 6, 2004
    #20
    blue&whiteman, I've got a similar protein (p639_L939_K12M_nat) and it must be pretty big. One processor of my 1.8 G5 has been folding it for just on 36 hours and is only up to frame 218 - only 182 to go :D.

    Out of interest it takes around 11 minutes to do a frame during the day, 8-9 minutes at night when there is no work being done. On slower processors these Tinkers must be excrutiatingly slow :eek:.
     
  21. Dreadnought macrumors 68020

    Dreadnought

    Joined:
    Jul 22, 2002
    Location:
    Almere, The Netherlands
    #21
    I know aus_dave, it takes about 2.5 days for this tinker to finish on a 1.8 g5...
     
  22. blue&whiteman thread starter macrumors 65816

    blue&whiteman

    Joined:
    Nov 30, 2003
    #22
    I have been on this tinker i'm on now since the day before I started this thread on 296 of 400 at the moment. this G4/500 does take a long time on tinkers. when the new G5's are out I am getting the cheapest dual. hope they are all dual so I can save more and get the entry level. once this happens the b&w will fold full time and the G5 will be the system I use and fold on.

    what kinda numbers you think I could put up with a single G4 500 and a dual G5?
     
  23. Sparky's macrumors 6502a

    Sparky's

    Joined:
    Feb 11, 2004
    #23
    I just logged on after 13 days of receiving my last Tinker and I have 298 frames completed out of 400. I run a G3 640mb RAM(upgraded to a G4-500 via Sonnet ZIFF CPU) and let it go 24/7. "Increase" says I started......

    "USER PID %CPU %MEM STAT STARTED CPUTIME COMMAND
    marclage 540 0.0 0.0 Ss 21May04 5:46.61 /Users/marclager/Library/Application Support/InCrease/bin/fah4 -local -advmethods -forceasm -verbosity 9

    Running Cores: 1

    USER PID %CPU %MEM STAT STARTED CPUTIME COMMAND
    marclage 23206 0.6 0.3 RN 24May04 13427:01.12 ./FahCore_65.exe -dir work/ -suffix 06 -checkpoint 15 -forceasm -verbose -lifeline 540 -version 400"

    So after 13 days I'm 3/4 the way done... :(
     
  24. blue&whiteman thread starter macrumors 65816

    blue&whiteman

    Joined:
    Nov 30, 2003
    #24
    just had a thought! there anything I can do to my system other than overclocking the cpu (not that i'm against overclocking but this is a sonnet cpu and I don't think their zifs can be overclocked) to make folding faster? I doubt more ram would help but if it would let me know!
     
  25. bousozoku Moderator emeritus

    Joined:
    Jun 25, 2002
    Location:
    Gone but not forgotten.
    #25
    I'm completely Tinker free for the moment, although the G3/800 got a bigger Gromacs WU yesterday that will finish next Wednesday.

    blue&whiteman:

    If you're really trying to get work done with only 512 MB while also trying to run folding@home, I would say that more RAM would definitely help your situation. Your machine is likely swapping much more than it should, which slows it down. I even noticed some slowdown going from 1 GB down to 704 MB.

    Good luck to everyone using a slower machine! :)
     

Share This Page