Folding gromacs, a caution

RiddickRiddick Malaysia Icrontian
edited September 2003 in Folding@Home
After countless amounts of lost WU progress i have come to the conclusion that when folding gromacs, one MUST use a system running WinXP (or other stable OSes like linux or win2k) to avoid any loss of work. (unless of course ur win98/winme machine never crashes)

I have noticed over a period of long time that whenever my non-WinXP machines shut down improperly whenever f@h was running it would almost 100% start back from its initial work packet ( 0 frames again ) and after switching to XP, these problems ceased even when the improper shutdowns occur.

just a heads up for new folders. :fold:

Comments

  • mmonninmmonnin Centreville, VA
    edited August 2003
    What client are they all running. Some of the older clients are more problematic when shutting down the clients.
  • WuGgaRoOWuGgaRoO Not in the shower Icrontian
    edited August 2003
    ive been experiencing problems with restarts aswell...in addition..sumtimes the gromac wont even start up properly
  • RiddickRiddick Malaysia Icrontian
    edited August 2003
    all of them console 3.24

    i dont think i can find a old client download?
  • mmonninmmonnin Centreville, VA
    edited September 2003
    They still have some for download if you know the links. I still have several old versions.

    I actually had a WU restart on me today doing M$ updates.
  • SlickSlick Upstate New York
    edited September 2003
    Doesnt the -forceasm flag help keep your work protected from unexpected shutdowns?
  • danball1976danball1976 Wichita Falls, TX
    edited September 2003
    Forceasm only enables the use of 3DNow! or SSE depending on what client you have.
  • ThraxThrax 🐌 Austin, TX Icrontian
    edited September 2003
    No, Danball.

    Forceasm enables SSE on AMD chips that use 3dnow! by default.
  • muddocktormuddocktor
    edited September 2003
    Thrax said
    No, Danball.

    Forceasm enables SSE on AMD chips that use 3dnow! by default.

    But only with the beta 3.25 client on AMD procs.

    -forceasm doesn't protect you from bad_checksum_errors, which is the main cause of the client restarting from scratch. All -forceasm does on Intel rigs does is to make sure the client starts while using assembly loops(SSE) without checking prior termination but AMD's with the beta 3,25 client will also force the use of SSE instead of the default 3DNow! instructions, as well as forcing the client to use assembly loops without checking prior termination of the client.
  • mmonninmmonnin Centreville, VA
    edited September 2003
    Danball was correct.

    Welcome back Jim.:)
  • BuiesCreek847BuiesCreek847 In a van, down by the river, NC Icrontian
    edited September 2003
    I wonder if it's not the OS, but the filing system used. FAT32 or NTFS. NTFS has all the additional security features. Is it possible these include better means of recovery? :scratch:
  • mmonninmmonnin Centreville, VA
    edited September 2003
    Never thought of that. I always use NTFS and I dont have problems but very once in awhile.
  • RiddickRiddick Malaysia Icrontian
    edited September 2003
    my guess is on the filesystem, if only winme/98 could run on NTFS LOL
  • muddocktormuddocktor
    edited September 2003
    I have some Win2K rigs running FAT32 and some running NTFS and I've noticed no difference in folding reliability between them myself.:) If not using SSE optimizations, they will all stay up for weeks on end. I do get the occasional lockup while using SSE optimizations on my Tbred B and Barton rigs, but no difference when all are using 3DNow! instead of SSE. The SSE lockup problems seems to primarily be only related to Tbred B and Barton procs.
Sign In or Register to comment.