Dual SMP in Linux

_k_k P-Town, Texas
edited January 2009 in Folding@Home
Help me ^title. I assume you can, it would be silly if you could not. Just started using Linux this week and I really don't get the command structure. All I want to do is get a second SMP client up because I have a quad that is running 60% load with a single client. Yes I have 4GB of ram to handle the dual clients. I am running Ubuntu v8.1. If you could explain the command syntax for launching in terminal and show me exactly what I need to input that would be super sweet. I am doing this as a learning experience, please teach me.

...........Phenom 9600 folds so slow on the big WUs.

Comments

  • edited January 2009
    _k_ wrote:
    ... I am running Ubuntu v8.1 ...

    Here is your first problem. Run 8.04.1 if you are wanting to fold under Ubuntu. The performance increase is worth loading it up again.

    Also, you are not the only one seeing < full load on Phenom quads. Check foldingforum.org

    Oh, and: http://icrontic.com/articles/ultimate_newbie_guide_gnulinux_shell

    and one more, my post on setting up SMP under linux: http://icrontic.com/forum/showpost.php?p=663318&postcount=3

    Ignore the parts referencing a virtual machine, the rest is exactly what you need to get going. To run dual SMP (although you probably shouldn't) just install to a separate directory and change the machine ID of the second client.
  • _k_k P-Town, Texas
    edited January 2009
    How much performance increase are we talking about cause remember how lazy I am.
  • edited January 2009
    On C2Q it's about 20%
  • _k_k P-Town, Texas
    edited January 2009
    Why shouldn't I run dual SMP
  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska
    edited January 2009
    Why shouldn't I run dual SMP
    Well, you tell me! Windows with a quad core, Linux with a quad or dual.
  • edited January 2009
    This is a debate that has been had for years, since the first Pentium 4 with Hyper-Threading. Running multiple clients will often result in higher PPD but it is more useful to Stanford for you to get results back to them ASAP. Running dual clients will often return more work, but not as quickly.

    Also, the F@H client for Linux is so efficient, I wouldn't be surprised if running dual SMP actually produced fewer PPD. My only reservation with that theory is in the case of the Phenom II, which do not seem to be fully utilized by the Linux client at this time.
  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska
    edited January 2009
    This is a debate that has been had for years, since the first Pentium 4 with Hyper-Threading.
    The debate isn't very strong anymore. The power users will continue to run double SMP clients as long as they can, or until Pande Group comes out with a client that utilize mult-core CPUs more efficiently. I suppose Pande Group would also be happy if everyone ran the fastest process on the fastest motherboard with a max number of the very fastest video cards. That would also get work units back faster. But obviously, they happily accept less than the ideal.

    Mas0n, I know exactly what you meant, but that debate has pretty well died, as most everyone ignored Pande Group in that regard.
Sign In or Register to comment.