Fold Faster on GPUs With FahCore_17

Discussion in 'Folding@Home - Join Team Guru3D !' started by PantherX, May 6, 2013.

  1. iancook221188

    iancook221188 Ancient Guru

    Messages:
    1,729
    Likes Received:
    20
    GPU:
    GTX 670 SLI / GTX 460 SLI
    those results sound grate Babel-17

    ive now OC my 670s a bit and there kicking out 70k ppd
     
  2. must_kill_ati

    must_kill_ati Maha Guru

    Messages:
    1,268
    Likes Received:
    9
    GPU:
    6800 XT
    @yasamoko (from rig pic thread)
    I added the flag you mentioned (I think) - only effect I've noticed so far is a lower estimated ppd by around 4k/day. Not too comfy with the interface and wondering if I've borked something. Does the config window look correct? Btw, thanks for even asking - I appreciate any/all help!
    [​IMG]
     
  3. yasamoka

    yasamoka Ancient Guru

    Messages:
    4,875
    Likes Received:
    259
    GPU:
    Zotac RTX 3090
    No, not there. You need to go to "Slots", click on gpu, edit, then add them at the bottom.

    It should not take effect instantly. It should take effect when you upload the finished Core15 WUs and the client then downloads new beta Core17 WUs.
     
  4. must_kill_ati

    must_kill_ati Maha Guru

    Messages:
    1,268
    Likes Received:
    9
    GPU:
    6800 XT
    But I want my ppd naaaaoww. :)
    tnx, i had tried the slots route and added a : behind client-type for some odd reason. when it threw errors at me i fig'd it MUST be the client cuz i'm not toopid.

    Tnx again
     

  5. yasamoka

    yasamoka Ancient Guru

    Messages:
    4,875
    Likes Received:
    259
    GPU:
    Zotac RTX 3090
    hahaha, alright. Tell me what PPD you get when your WUs become Core17.
     
  6. must_kill_ati

    must_kill_ati Maha Guru

    Messages:
    1,268
    Likes Received:
    9
    GPU:
    6800 XT
    I'll tell ya' right now, the ppd increase is big. One of my clients is on the beta, while one is still on the older project. Estimated ppd atm is 58,682 - my old high for estimated ppd was around 42k, so extrapolate that and we're looking at a possible 75k ppd? That would be a phenomenal increase. I'll post actual ppd increases later but for now, I have a question for you.
    FahCore_17.exe*32 is using 1 of my cores at 25%(the full core) whereas FahCore_15.exe32 used a max of 4% of a core.
    Is that normal/what yours is using?
     
  7. yasamoka

    yasamoka Ancient Guru

    Messages:
    4,875
    Likes Received:
    259
    GPU:
    Zotac RTX 3090
    Yes, the PPD increases come from the new OpenMM core that has greatly sped up GPU folding.

    The core usage is normal. That happens with Nvidia cards currently on Core17. It doesn't happen with me because I run an AMD card (7970).

    What's your TPF for each card, clocks and voltages?

    EDIT: If you want an accurate TPF, take a look at the log, and get the TPF by time difference.
     
  8. must_kill_ati

    must_kill_ati Maha Guru

    Messages:
    1,268
    Likes Received:
    9
    GPU:
    6800 XT
    TPF for the core15 atm is 3mins 10secs.
    TPF for the core17 atm is 3mins 27secs.

    Clocks and voltages are stock. 980MHz @ 1.0620V

    I appreciate folding and it's goals, but asking me to decipher the log is going too far.
    lol, i just realized how lazy i really can be.


    *ninja edit
     
    Last edited: May 20, 2013
  9. yasamoka

    yasamoka Ancient Guru

    Messages:
    4,875
    Likes Received:
    259
    GPU:
    Zotac RTX 3090
    hahaha well just choose the slot for core17, and take a look at the last adjacent percentages, then subtract the time. That's all there is to it ;)
     
  10. must_kill_ati

    must_kill_ati Maha Guru

    Messages:
    1,268
    Likes Received:
    9
    GPU:
    6800 XT
    Last edited: May 20, 2013

  11. PantherX

    PantherX Folding@Home Team Leader

    Messages:
    1,380
    Likes Received:
    0
    GPU:
    Nvidia GTX 1080 Ti
    Regarding the CPU Usage, it is an Nvidia "feature" which will "use" the entire CPU thread for a single GPU. Each GPU needs a CPU. This is the OpenCL design that Nvidia has taken. AMD on the other hand, took a different approach thus doesn't require a full CPU Thread for each GPU. I do hope that Nvidia will "fix" this in the future.

    Regarding the TPF, it is normal to expect variations of couple of seconds between each percentage. That's the way the FahCore_17 works for now. Also, the GPU Usage will dip for few seconds as it doesn't some verification. Both of these may or may not change in the future. Do note that FAHControl will take a couple of frames to get the correct average.
     
    Last edited: May 20, 2013
  12. PlayN@Home

    PlayN@Home Active Member

    Messages:
    50
    Likes Received:
    0
    GPU:
    3 @ GTX 570 WC
    Apparently I'm missing something. I've been trying the get the beta option working and it never saves it in the config. I'm trying to use 7.3.6, trying. I click on Configure, then the slots tab, select one of the GPU slots, then click edit at the bottom. scroll to the bottom to extra options, click add and put beta in the box that pops up and save the config. When I go back and edit the GPU slot, the extra options is empty. Am I doing this in the correct location?:bang:

    V6 client was so much easier to configure.
     
  13. iancook221188

    iancook221188 Ancient Guru

    Messages:
    1,729
    Likes Received:
    20
    GPU:
    GTX 670 SLI / GTX 460 SLI
    that strange, im on 7.2.9 have not heard of any problems in 7.3.6 .

    it taking me a little time to get used to V7 as well
     
  14. mooney4321

    mooney4321 Member

    Messages:
    34
    Likes Received:
    2
    GPU:
    2080S Seahawk EK

    use client-type in Name field and beta in Value field..
     
  15. PlayN@Home

    PlayN@Home Active Member

    Messages:
    50
    Likes Received:
    0
    GPU:
    3 @ GTX 570 WC
    Thanks. I'll give it try after I get home from work.
     

  16. must_kill_ati

    must_kill_ati Maha Guru

    Messages:
    1,268
    Likes Received:
    9
    GPU:
    6800 XT
    Sometimes seeing something done makes it click inside my head. If you're like me, this video may help with adding beta flags to gpu clients.

    In the video, he first removes his beta flag, in order to add it again.

    http://www.youtube.com/watch?v=d1AQfs6Hk24

    You'll start getting beta projects after your current ones are done.
     
  17. yasamoka

    yasamoka Ancient Guru

    Messages:
    4,875
    Likes Received:
    259
    GPU:
    Zotac RTX 3090
    So that explains this then?:
    [​IMG]
     
  18. PantherX

    PantherX Folding@Home Team Leader

    Messages:
    1,380
    Likes Received:
    0
    GPU:
    Nvidia GTX 1080 Ti
  19. PantherX

    PantherX Folding@Home Team Leader

    Messages:
    1,380
    Likes Received:
    0
    GPU:
    Nvidia GTX 1080 Ti
    Not exactly. There are two kinds of dip:
    1) From 99% (or the max on your GPU) to 0% for few seconds. This usually happens for verification purpose
    2) From 99% to 93% (basically it dips a few percentage and oscillates between them). This is how FahCore_17 functions and is considered normal.

    Do you see this kind of dip frequently? Are there any other dips that you see on your GPU?

    Do note that a new version of FahCore_17 has been published so you may want to upgrade after the current WU finishes -> http://foldingforum.org/viewtopic.php?f=66&t=24209&start=120
     
  20. must_kill_ati

    must_kill_ati Maha Guru

    Messages:
    1,268
    Likes Received:
    9
    GPU:
    6800 XT
    Last edited: May 21, 2013

Share This Page