wootness

GrayFoxGrayFox /dev/urandom Member
edited January 2006 in Folding@Home
I just upgraded my server from 2X 2ghz xeon's to 2x 2.8ghz Xeon Ht :).

I have FAH running 4 times :)

http://graynetwork.org/phpsysinfo

I should have a pretty big production increese.

Comments

  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited January 2006
    A virtual CPU (hyperthreading) will not double your production, but can increase it over just the physical CPU by around 25%. So yes, considering the boost from 2Ghz to 2.8GHz, and with the addition of two virtual CPUs, your production should go way up. Good work! HT and Folding are good friends. Make sure to set your clients to accept large packets and add the -advmethods flag for all four Folding instances. Assuming your computer has the necessary memory, you should be able to fold four QMDs simultaneously.
  • GrayFoxGrayFox /dev/urandom Member
    edited January 2006
    Im adding more ram tomorrow. (And yes im aware that ht wont double my production but it will still help :).)
  • profdlpprofdlp The Holy City Of Westlake, Ohio
    edited January 2006
    Looks like a nice boost for you, GrayFox - I've alerted the 50K Toaster factory to be prepared. :D
  • QeldromaQeldroma Arid ZoneAh Member
    edited January 2006
    :: Thinks that someone is going to have a contest of 250K, 500K and 1Mega cogs and signature stars if much more of this is going to go on ::

    Yes!

    wOOtness!emoticons!:woowoo:
  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited January 2006
    I just get all warm and mushy inside when the conversation steers towards dual core, hyperthreading, and dual CPU. sniff sniff


    gotta love it
  • muddocktormuddocktor
    edited January 2006
    GrayFox, the Xeon's won't output nearly as many points/hr increase as you figure they will, since your dual Xeon rig's processors share the same bus to the memory, especially with QMD work. Multiple QMD wu's flood the memory bus and that will limit their folding effectiveness. With 4 clients running, be sure to set up only 2 of the clients to do QMD work and then set the other 2 clients up to run regular production/non-BigWu work and that will net you a better ppd output than setting all 4 clients to do QMD work.
  • GrayFoxGrayFox /dev/urandom Member
    edited January 2006
    Im just goina stick them all on normal production (The server does other things other then fold :P)
  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited January 2006
    I've no experience with Xeons; I had assumed too much. Thanks Mudd for filling in where I advised badly.
  • muddocktormuddocktor
    edited January 2006
    GrayFox, that's a real good idea if that Xeon is doing other work besides folding. If you do have 1 GB or over of ram, you might be able to get away with running 1 client on a QMD and the others on regular work. You shouldn't see much performance penalty that way, if you feel like experimenting some day.

    That's no problem, Leo. There are quite a few folks on the overclockers.com team that are using Xeon workstation class boards for folding rigs and that is what they have found with trying to run 4 QMD wu's. Overclocking the fsb and ram helps a bit but too many QMD wu's flood the bus and no additional points increase is seen and can actually slow points (and wu) production down a little.
  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited January 2006
    Well that does make sense. I had forgotten about Xeon motherboards' antiquated data buses. With either two real cores Folding simultaneously, or a P4 Northwood with two cores - one real, on virtual Folding simultaneously, I have not seen production degradation running only QMDs. You had suggested to me before that I only run one QMD per computer with HT and dual core machines. I tested it. Mulitple, simultaneous QMD's far outpace QMD/non-QMD production. But four instances of Folding on one computer... oh boy, some day!
  • DonutDonut Maine New
    edited January 2006
    muddocktor wrote:
    GrayFox, that's a real good idea if that Xeon is doing other work besides folding. If you do have 1 GB or over of ram, you might be able to get away with running 1 client on a QMD and the others on regular work. You shouldn't see much performance penalty that way, if you feel like experimenting some day.

    That's no problem, Leo. There are quite a few folks on the overclockers.com team that are using Xeon workstation class boards for folding rigs and that is what they have found with trying to run 4 QMD wu's. Overclocking the fsb and ram helps a bit but too many QMD wu's flood the bus and no additional points increase is seen and can actually slow points (and wu) production down a little.

    Currently my baby is at 223 fsb with 2 gigs of ram, more than 2 QMD's and the completion times start to tank. With 3 the computer starts to feel sluggish, mind you I don't do anything too intensive (ie. browse Short-Media).

    fwiw, 1.6lv@2.9, PC-DL
  • muddocktormuddocktor
    edited January 2006
    Donut wrote:
    Currently my baby is at 223 fsb with 2 gigs of ram, more than 2 QMD's and the completion times start to tank. With 3 the computer starts to feel sluggish, mind you I don't do anything too intensive (ie. browse Short-Media).

    fwiw, 1.6lv@2.9, PC-DL

    Yep, that's exactly what some of those guys there have found too. More than 2 QMD's floods the data busses so bad that the machine gets non-responsive, no matter how much ram it has.
  • GrayFoxGrayFox /dev/urandom Member
    edited January 2006
    Im getting more ram tomorrow.
  • Liquid81Liquid81 Westbrook,CT
    edited January 2006
    pretty sweet indeeed !
Sign In or Register to comment.