New Core 21 0.0.18 lower PPD

Discussion in 'Distributed Computing' started by Skillz, Feb 2, 2017.

  1. Skillz

    Skillz [H]ard DCOTY 2017

    Messages:
    21,974
    Joined:
    Aug 14, 2004
    Anyone else noticing a poor performance with the latest core 21 update? I'm away from my boxen, but my PPD has ranked by about 1M since the update. Need to know if it's because of the New Core or something wrong on my end.
     
  2. Core32

    Core32 [H]ard|Gawd

    Messages:
    1,065
    Joined:
    Mar 3, 2012
    Lower PPD, higher power required, so far. I'm down about 18% on PPD on what ever project is now running.
    Edit: Oh, and I also had a "core_update" error in the beginning on one of the 1050s that would not self-clear, so you may also have that issue if the numbers are off much more than 18%. I let the other board finish it's WU, deleted both GPU slots, saved, let the erroring WU clear out, then added both GPU slots back and no more errors last 24 hours.
     
    Last edited: Feb 2, 2017
  3. Skillz

    Skillz [H]ard DCOTY 2017

    Messages:
    21,974
    Joined:
    Aug 14, 2004
    Well I should be around 5.8M PPD but it seems to have dropped to around 4.8M PPD now. Which is right at 18%. So guess it's not an issue on my end, yet.
     
  4. ChristianVirtual

    ChristianVirtual [H]ard DCOTM x3

    Messages:
    2,522
    Joined:
    Feb 23, 2013
    For a final judgment I think we need to wait until nV removed some workaround in their driver ... it seems to be in a state of transition.
     
  5. jfb9301

    jfb9301 [H]ard DCOTM February 2016

    Messages:
    1,650
    Joined:
    Jan 11, 2005
    Huh... and I just restarted my boxen and the core_update cleared and the machine started folding. I've thought I saw a cut in production on that machine. My other machines are still running the 372.xx drivers, if there is any cut in production, I've not really seen it. My 1080 SLI is getting 2x600-800KPPD, with some units up to 900. Seemingly normal, though the average is a bit lower, but not any lower than it was 3 months ago, I just attributed it to the mix of WU.
     
  6. tmossman01

    tmossman01 n00b

    Messages:
    29
    Joined:
    Jun 5, 2010
    I am not seeing much drop in PPD for my GPU slots (1080s and 1060s), maybe a 5% drop, but this could be due to WU variance. (running 373.06 drivers) I did have the transient drop while the systems waited for the GPU slots to go idle to allow the core update.
     
  7. mmonnin

    mmonnin Limp Gawd

    Messages:
    190
    Joined:
    Jan 11, 2004
    Same PPD on my 980Ti with the new core with 346.96 in Linux. Must be the crappy new driver everyone thinks they have to upgrade to.
     
  8. ChristianVirtual

    ChristianVirtual [H]ard DCOTM x3

    Messages:
    2,522
    Joined:
    Feb 23, 2013
    Depend on the hardware you have. Newer hardware need newer driver. Enjoy your "antique" 980Ti (I do too :))
     
  9. mmonnin

    mmonnin Limp Gawd

    Messages:
    190
    Joined:
    Jan 11, 2004
    Only a 1050 or 1050Ti needs a broken driver.
     
  10. Linden

    Linden [H]ard|Gawd

    Messages:
    1,181
    Joined:
    Sep 8, 2005
    About the same performance lately as before. Maybe one or two projects are slower, maybe. Drivers are 373.06.

    Update: I keep a log. I've checked it thoroughly. On 373.06 drivers, I'm not seeing a difference between current performance and from a couple weeks ago on like projects.
     
    Last edited: Feb 4, 2017
  11. Core32

    Core32 [H]ard|Gawd

    Messages:
    1,065
    Joined:
    Mar 3, 2012
    The 1050s are still averaging lower PPD and for the first time since I bought them I am noticing some screen lag on everyday tasks.
    One card per monitor.
    Not really an issue much longer since I am retiring at 1.2B this week anyway.
    About to drop out of the top 10 for the first time in years.
    Too old, not enough time or patience to keep running optimally, and no longer much fun with Stanford remaining belligerent.
    Hopefully they eventually report some worthwhile results from all this effort.
     
  12. pututu

    pututu [H]ard DCOTM x2

    Messages:
    1,681
    Joined:
    Dec 27, 2015
    Might want to consider other DC projects (CPU and GPU) within BOINC? WCG provides regular update on their findings.
     
  13. Skillz

    Skillz [H]ard DCOTY 2017

    Messages:
    21,974
    Joined:
    Aug 14, 2004
    Damn. Hate to see you go Core32, but its understandable. Maybe if you just take a beak for a year you'll get back into it later. I've taken a few periods of extended breaks since I started folding back around 2001.
     
  14. Linden

    Linden [H]ard|Gawd

    Messages:
    1,181
    Joined:
    Sep 8, 2005
    I have too, brother. Hey, Core32, I understand. I just came back to F@H after an 18-month hiatus on BOINC. Do what you need to do. Please though, continue to contribute somewhere.