Possible heads up for Linux SMPers

SnarkasmSnarkasm Madison, WI Icrontian
edited March 2008 in Folding@Home
My client recently finished a unit and decided to download itself a new core, FahCore_a2 as opposed to the a1 we've all been using forever. Looks like it then got itself a 2619 (3, 400, 0) and started churning. Right now my output is looking really queer, so I'll have to investigate what's going on a little further.

Reading file work/wudata_06.tpr, VERSION 3.3 (single precision)
Note: tpx file_version 40, software version 48
starting mdrun 'TETHERED VESICLES'
125000 steps, 5000.0 ps.
[22:01:35] Completed 10 out of 125000 steps (0%)
[22:14:59] Compl1%)
[22:14:59] 630 out of 125000 steps (1%)
[22:42:45] eps (2%)
[22:42:45] 1880 out of 125000 steps (2%)
[23:10:37] eompleted 3130 out of 125000 steps (3%)

Comments

  • QeldromaQeldroma Arid ZoneAh Member
    edited March 2008
    You're right- the first clue is the mispelled output. Those that have done that to me have almost always gone EUE ... I think one time I had partial credit. I haven't seen that in many months (last September?). It doesn't look like I've seen that project number either. You might want to take that number and search it out at the FCF.

    Also- is it an overclocked unit or something that might cause a bad computation?
  • SnarkasmSnarkasm Madison, WI Icrontian
    edited March 2008
    Nope, it's my stock laptop that's been crunching regularly for months now. It's evened itself out a little bit now, but what used to be 33-35 minute frames have now become 27 minute frames or so, and it only has 125,000 steps, as you can see, when they usually have 250k, right? So I'm guessing it's an adjustment to the WUs and I'm gonna have to wait and see what point value it gives me.

    More data:

    [00:19:52] Timer requesting checkpoint
    [00:31:29] 6%)
    [00:31:29] leted 6880 out of 125000 steps (6%)
    [00:46:44] Timer requesting checkpoint
    [00:58:43] Completed 8130 out of 125000 steps (7%)
    [01:13:56] Timer requesting checkpoint
    [01:26:54] Completed 9380 out of 125000 steps (8%)
    [01:42:09] Timer requesting checkpoint
    [01:53:55] Completed 10630 out of 125000 steps (9%)
    [02:09:09] Timer requesting checkpoint
    [02:21:26] Completed 11880 out of 125000 steps (10%)
    [02:36:38] Timer requesting checkpoint
  • QeldromaQeldroma Arid ZoneAh Member
    edited March 2008
    Yeah- I have an LT that is also a marvelous folder ... I live in a very dusty enviroment so I have to clean mine out about once every 3-4 months. Otherwise I'm out of ideas right now.

    Let us know how it turns out.
  • mmonninmmonnin Centreville, VA
    edited March 2008
    Steps mean NOTHING. I've seen WUs with like 500,000 steps be worth like 50 points and 600 pointers have much fewer steps. Time per % is the only thing that matters for PPD.

    Its a standard project:
    http://vspx27.stanford.edu/psummary.html
  • SnarkasmSnarkasm Madison, WI Icrontian
    edited March 2008
    Yeah, I ran 25-minute frames last night, so I guess it's just a new type of unit for me. I'll try and update when it finishes.
  • QeldromaQeldroma Arid ZoneAh Member
    edited March 2008
    This might be related- check it out. It is also mentioned that the 2619 is extremely memory hungry- 800mb(= 4 * 200...!)
  • mmonninmmonnin Centreville, VA
    edited March 2008
    Hmm they mentioned the a2 core in the FAQ I saw the other day but its just coming out with WUs...usually it happens the other way around.
  • Ultra-NexusUltra-Nexus Buenos Aires, ARG
    edited March 2008
    I lost 2 of these lately to EUE because of low mem and out of disk space stuff... also by just rebooting the VM where they run.

    Found that setting the BigWU to "small" limits this P2619 WU and only downloads P2605s and P2653s.

    If you find your folder machines are crapping out with this new WU, give this a try. Or experiment... whatever suits you. Its still in beta testing so either way is fine.

    Personally, I like my electricity used, rewarded. :)
  • SnarkasmSnarkasm Madison, WI Icrontian
    edited March 2008
    Mine is doing all right so far, lappy has 2 gigs and it's running Ubuntu, so it seems to shuffle memory well. Right now, the cores are running between 160 and 171.5MB, so that 800 number is about right. Still watching it, though.
  • mmonninmmonnin Centreville, VA
    edited March 2008
    I didn't think any of the options during the cfg actually worked for the SMP client such as the big WUs, forceasm and advmethods. I thought it was just set since you choose the SMP client and you got what you got.
  • Ultra-NexusUltra-Nexus Buenos Aires, ARG
    edited March 2008
    only advmethods is inoperative currently in the SMP client. The BigWU option and forceasm flag do work.
Sign In or Register to comment.