Northwoods fold well....

Straight_ManStraight_Man Geeky, in my own wayNaples, FL Icrontian
edited March 2004 in Folding@Home
This is not about hardware, per se, rather about OCing stably to maximize FOLDING. Here's and example, from today, times are Zulu\GMT:

[18:22:09] - Starting from initial work packet
[18:22:09]
[18:22:09] Project: 814 (Run 7, Clone 54, Gen 3)
[18:22:09]
[18:22:09] Assembly optimizations on if available.
[18:22:09] Entering M.D.
[18:22:15] Protein: p814_p53dimer814
.
folding at ~= 3 1\3 min per percent AVERAGE...
.
[23:39:32] Past main M.D. loop
[23:40:32]
[23:40:32] Finished Work Unit:
.
.
.
[23:40:36] Unit 5 finished with 99 percent of time to deadline remaining.
[23:40:36] Updated performance fraction: 0.986738
[23:40:36] Sending work to server
.
.
.
[23:41:13] - Uploaded at ~28 kB/s
[23:41:13] - Averaged speed for that direction ~28 kB/s
[23:41:13] + Results successfully sent
[23:41:13] Thank you for your contribution to Folding@home.

Same Box, while posting on it....

[23:41:21] Entering M.D.
[23:41:27] Protein: p914_vpf909
[23:41:27]
[23:41:27] Writing local files
[23:41:28] Extra SSE boost OK.
[23:41:28] Writing local files
[23:41:30] Completed 0 out of 250000 steps (0)
.
.
.
[02:33:29] Writing local files
[02:33:31] Completed 92500 out of 250000 steps (37)

& Avg Time 4 min plus 41 to 42 seconds, floating back and forth....

Stanford shows 19545 right now for me, before the p814 credit....

Processor: Northwood, native at 2.8 GHz, running at actual 3200 MHz.
RAM: DDR333 CMX Corsair, 1 GIG.
Graphics Card: GF2 MX 400
Storage: WD BB IDE HDs
Motherboard: IC7-Max3
Temps: system 26 C, CPU 52 C, Otes 38.5 C (note Otes is defaulted to not alarm in Abit EQ until it reaches 70 C but actually does not like being over about 48 C very well).

Methinks Northwoods are NEAT for FOLDING.... :D

Comments

  • muddocktormuddocktor
    edited March 2004
    Yeah John, my 2 Northies have made a believer out of me too. They are some number crunching fools on Gro work, especially running 2 instances on the HT Northies. I just bought a refurb IS7 right before I had gone back offshore just so I could get my 2.6 folding along with my m0 stepping 2.4.
  • FoldingAddictFoldingAddict Montgomery, AL
    edited March 2004
    Finally somebody that isn't bitching about Intel's "poor performance"!

    I know what you guys mean though, I just got my piano teacher's new P4 2.53 Northwood folding some gromacs, and that has really been chewing up the WUs.:O

    ~FA
  • primesuspectprimesuspect Beepin n' Boopin Detroit, MI Icrontian
    edited March 2004
    Hey FA! Wow! Great to see you back! :rockon:
  • mmonninmmonnin Centreville, VA
    edited March 2004
    They do a lot better ar FAH than other apps because it doesnt take much for the prefetch to keep the pipe of the P4 busy. Before HT the prefetch wasnt good enough to fill it completely and there were gaps, wasted cycles.

    And after hearing about how running a 2nd client doenst slow down the first client very much only shows that almost half the pipe was empty at times. A complete waste.
  • ThraxThrax
    slowly looks into this thread and takes a slow scan before clearing his throat

    Northwoods suck. :bigggrin:

    That is all.
     🐌 Austin, TX Icrontian
    edited March 2004
    slowly looks into this thread and takes a slow scan before clearing his throat

    Northwoods suck. :bigggrin:

    That is all.
  • muddocktormuddocktor
    edited March 2004
    Thrax wrote:
    * Thrax slowly looks into this thread and takes a slow scan before clearing his throat


    Northwoods suck. :bigggrin:

    That is all.

    But they don't suck near as badly as either the P4 willie or presshott.... err, prescott. :D
  • ThraxThrax 🐌 Austin, TX Icrontian
    edited March 2004
    Yep. ;D
  • primesuspectprimesuspect Beepin n' Boopin Detroit, MI Icrontian
    edited March 2004
    you can just let it go, huh guys? ;D
  • edcentricedcentric near Milwaukee, Wisconsin Icrontian
    edited March 2004
    Two questions, that I don't have the equipment to answer.

    1. On an Intel with HT, does running two instances of F@H change the output of hte first at all? Does the total production double?, go up 50%? or what...

    2. How do the 'M's fold? We know that they are a lot stronger in general than the P4's. But what about in a straight math task.
  • muddocktormuddocktor
    edited March 2004
    1. HT P4's slow down around 25% or so I would guess with 2 instances running, which still yields quite a bit better points/hr yield from both instances combined compared to running just 1 instance on it.

    2. No actual experience with the P-M procs but I know that the P3 would run neck and neck with an XP at the same clock speed as the P3 and the P-M is supposed to be much improved over the P3. It's just too bad there aren't any desktop boards to be found with some overclocking features for these P-M procs. It would seem like a great choice to build one of those little Shuttle type rigs with due to the much lower heat and power requirements of the P-M.
  • primesuspectprimesuspect Beepin n' Boopin Detroit, MI Icrontian
    edited March 2004
    I have a 1.3ghz banias, an athlon xp 2400+, and a thunderbird 1.33ghz in my house.

    I don't feel like getting the hard numbers right now, but the 1.3ghz Pentium M is WAY faster than the 1.33ghz tbird, and just a bit slower than the 2400+

    For example, pretend there is a protein that takes 8:00 per frame on my thunderbird 1.33. The same protein would take, say, 4:00 on the 2400+ and maybe 5:00 on the banias.

    I know, i know, not very scientific.. I'll just say they fold very fast for their clock speed.
  • mmonninmmonnin Centreville, VA
    edited March 2004
    P-Ms They freakin smoke FAH. I had some on the school LTs. I think I still have some of those suckers folding for me. The P-Ms are THE best thing Intel ever designed. I had been reading a bit about them. They take the best from the P3 and the P4 and made the P-M with power saving solutions.
  • Straight_ManStraight_Man Geeky, in my own way Naples, FL Icrontian
    edited March 2004
    edcentric wrote:
    Two questions, that I don't have the equipment to answer.

    1. On an Intel with HT, does running two instances of F@H change the output of hte first at all? Does the total production double?, go up 50%? or what...

    2. How do the 'M's fold? We know that they are a lot stronger in general than the P4's. But what about in a straight math task.

    One thing about HT to keep in mind.... First, internally, the cpu does not have bridge pipe for each HT pipe, second they use a common larger caching structure for whole CPU, not two complete cache sets per HT pipe. So, improvement will not be double for these CPUs as opposed to two Xeons on a board with same base speed and same RAM clocking and rate, and same FSB.

    I would expect them to be about 1.6-1.7 (at 3.4 GHz OR up TRUE speed) times the effective speed of a single piped CPU of same rate, cache size, and on a motherboard with same basic chipset and FSB structure and same RAM. Most of the reason for the huge caching is to work as a holding space for movement to and from main RAM, as the RAM definitely cannot be piped as wide or fast as the internals of the CPU are. Anand Shimpi showed this on his site in articles about Northwood versus Prescott, in more detail. His site is at <A href="http://www.anandtech.com/">this virtual place</a>, and he proved why the speed versu relative performance thing is true.

    So, for 2 instances with an O\S running, if you use an HT able Northwood, you will want the fastest RAM you can get and migth consider piping most of the O\S processes and services to CPU0 which is pipe 0 and one folding instance to pipe 1-- alias the processes to balance load onto pipe 0, user pipe 1 for folding almost exclusively. By doing this, you get a mostly dedicated folding pipe that is always on as most other things can use the main pipe that Windows favors (CPU0 or pipe 0). If you can alias this way, things will probably almost equal two instances within 10-15%. The client will be grabbing almost all of the pipe 1's time slicing as Windows will be running on the other pipe and BOTH will be running at the speed of the CPU rating or above if you OC.

    Do not bother running the Folding client as a graphical, run it as a console client set up as a service if you want to try this.

    My Northwood is a single pipe Northwood, BTW.

    John D.
  • edcentricedcentric near Milwaukee, Wisconsin Icrontian
    edited March 2004
    prime, that is what I expected since the P-M has sse3 and the TBird is only 3dn. My tbred at 1.4 out folds a tbird at 1.6 thanks to sse.

    Yes, a micro atx mobo for P-M would be nice.
Sign In or Register to comment.