Folding at home guides

TushonTushon I'm scared, CoachAlexandria, VA Icrontian
edited December 2011 in Folding@Home
Trying to gauge interest in what people would want to read/use.

Currently, F@H is beta-ing v7 client, which allows you to manage CPU/SMP and GPU folding from one interface, and rather automatically. There are some bugs they are still working out, but it is mostly solid.

There is folding bigadv (on Windows and linux is fairly similar, but linux ends up with a decent gain in PPD) that I have successfully setup (on Ubuntu) and will have monitoring done shortly.

I have been tossing around ideas for guides, but if people want a specific one (v7 howto vs bigadv), I'd work on that one first. Stuff has settled down some for me so I have some time to work it. Let me know if you have ideas/preference or what would get YOU folding.
«1

Comments

  • clifford_cooleyclifford_cooley Arkansas, USA Member
    edited October 2011
    I've been using the V7 Client for sometime now. I've grown to appreciate the flexibility with the program. There was some confusion at first but I've slowly learned how to use the program.

    Instead of multiple clients there is what they call slots. These slots can be configured to use different User ID's, Team ID's, or Passkey's. Adding a custom value for a slot will override the default global values set for the client.

    I'm currently folding with the V7 Client configured so that bigadv points goto one team and the GPU points goto another. I'm using the same passkey for both Team's, so I didn't have a need in changing the global passkey by adding a passkey value for either slot configuration.

    I've even made a remote connection and controlled a V7 Client on another machine. However for some reason after the last V7 Client update was unable to make this connection. At this point though, I'm not convinced the problem wasn't myself.

    I think these guides for the V7 Client would be very useful.
  • TushonTushon I'm scared, Coach Alexandria, VA Icrontian
    edited October 2011
    Thanks for the input. I run v7 on my main rig and file server, v6 for bigadv on my dedicated box. I also had issues with remote access, but I only tried during the first beta, so I should update both clients and try again.
  • _k_k P-Town, Texas Icrontian
    edited October 2011
    Ubuntu install with Kraken wrapper.
    FahMon install
    Possibly a short linux file sharing guide for remote monitoring.
    Issues with remote viewing a windows PC and GPU3 issues.
    The trade offs from banking on SMP/Bigadv v. GPU.
  • TushonTushon I'm scared, Coach Alexandria, VA Icrontian
    edited October 2011
    I wouldn't do Kraken only because it should only be used for multi-CPU setups. I can do the rest of it
  • _k_k P-Town, Texas Icrontian
    edited October 2011
    You do it all or I pee on your car.
  • TushonTushon I'm scared, Coach Alexandria, VA Icrontian
    edited October 2011
    That's just rude. Give me a SR-2 or G34 with quad procs and I'll do it big!
  • csimoncsimon Acadiana Icrontian
    edited October 2011
    Yeah Tushon go for it.
  • csimoncsimon Acadiana Icrontian
    edited December 2011
    What are minimum requirements for BIGADV?
  • GargGarg Purveyor of Lincoln Nightmares Icrontian
    edited December 2011
    csimon wrote:
    What are minimum requirements for BIGADV?

    Soon, it will be 16 threads. I think it's 8 now?

    The move to 16 threads saved me a hundred bucks. I bought a 2500k last week instead of a 2600k when I realized I wouldn't be able to do BIGADV :D
  • TushonTushon I'm scared, Coach Alexandria, VA Icrontian
    edited December 2011
    Gargoyle wrote:
    Soon, it will be 16 threads. I think it's 8 now?

    Correct
  • csimoncsimon Acadiana Icrontian
    edited December 2011
    I just read that some guys are doing big wu's on X6's ...right now I am set with client-type advanced. So should I not bother?
  • TushonTushon I'm scared, Coach Alexandria, VA Icrontian
    edited December 2011
    It is possible, but requires essentially constant folding in order to turn it in on time (source). Not recommended
  • csimoncsimon Acadiana Icrontian
    edited December 2011
    Tushon wrote:
    It is possible, but requires essentially constant folding in order to turn it in on time (source). Not recommended
    Ok that's what I thought. Though this machine is on 24/7 it isn't stable and reliable enough yet. I'll hash that around at a later date perhaps.
    Wasn't there a reference to 12gb ram minimum that I remember from someplace?
  • TushonTushon I'm scared, Coach Alexandria, VA Icrontian
    edited December 2011
    csimon wrote:
    Ok that's what I thought. Though this machine is on 24/7 it isn't stable and reliable enough yet. I'll hash that around at a later date perhaps.
    Wasn't there a reference to 12gb ram minimum that I remember from someplace?

    No, I believe the new "recommended" spec is .5GB/thread, so minimum of 8GB for the new stuff in March.
  • csimoncsimon Acadiana Icrontian
    edited December 2011
    Who the heck has 16 threads? I remember when points were benchmarked from a Pentium 25mhz sheesh.
  • clifford_cooleyclifford_cooley Arkansas, USA Member
    edited December 2011
    I know you are being sarcastic :) because there was no Pentium 25MHz.
  • csimoncsimon Acadiana Icrontian
    edited December 2011
    clifford_cooley wrote:
    I know you are being sarcastic :) because there was no Pentium 25MHz.
    You're right ...it was an Intel 386DX.

    Tushon is that pee on your car?
  • TushonTushon I'm scared, Coach Alexandria, VA Icrontian
    edited December 2011
    Wut?
  • csimoncsimon Acadiana Icrontian
    edited December 2011
    _k_ wrote:
    You do it all or I pee on your car.

    Write the guide or the tire gets it.
  • TushonTushon I'm scared, Coach Alexandria, VA Icrontian
    edited December 2011
    Oh that ... yeah. I have the linux guide all written out, but prime was waiting for me to add v7 stuff, though it (what I would write for v7) doesn't really improve any on the available guides.
  • _k_k P-Town, Texas Icrontian
    edited December 2011
    To be fair I already had that taken care of. Had one of my minions destroy the inside of his car.
  • TushonTushon I'm scared, Coach Alexandria, VA Icrontian
    edited December 2011
    Bastard!
  • csimoncsimon Acadiana Icrontian
    edited December 2011
    Tushon wrote:
    Oh that ... yeah. I have the linux guide all written out, but prime was waiting for me to add v7 stuff, though it (what I would write for v7) doesn't really improve any on the available guides.
    Just update us on which flags are optimal for which systems or something like that. What hardware takes advantage of which flags and stuff. You know. :rockon:
  • TushonTushon I'm scared, Coach Alexandria, VA Icrontian
    edited December 2011
    Get your AMD cards out of here. That is all you need to know.
  • shwaipshwaip bluffin' with my muffin Icrontian
    edited December 2011
    Eh, AMD cards can be quite good. On my phenom X6, I dropped SMP from 6 cores to 4 (smp only likes even numbers of threads), and folded on dual 5870s for a huge PPD increase. I stopped because it wasn't worth the heat increase or power usage.

    It is a shame that you can't fold with CF enabled and that the CPU usage is much higher than nvidia folding, but they can still put out a lot of ppd.
  • csimoncsimon Acadiana Icrontian
    edited December 2011
    Don't you get bonus points on GPU folding? I thought that GPU's far outproduced CPU's. What the hell happened?
  • GHoosdumGHoosdum Icrontian
    edited December 2011
    shwaip wrote:
    On my phenom X6, I dropped SMP from 6 cores to 4 (smp only likes even numbers of threads)...

    Is the even/odd reference based on half the cores? Or does SMP run a thread across two cores (is that even possible)?
  • shwaipshwaip bluffin' with my muffin Icrontian
    edited December 2011
    RE bonus points:
    I don't know the details. Adding 2x5870s to folding greatly increased my PPD, vastly more than enough to overcome dropping 2 threads from the SMP client.

    RE: even/odd
    I think it has to do with how folding divides up the work. The client restricts you to an even number of threads - I read somewhere else that the problem is with prime numbers of threads...but who knows. Just use an even number :)
  • TushonTushon I'm scared, Coach Alexandria, VA Icrontian
    edited December 2011
    shwaip wrote:
    Eh, AMD cards can be quite good. On my phenom X6, I dropped SMP from 6 cores to 4 (smp only likes even numbers of threads), and folded on dual 5870s for a huge PPD increase. I stopped because it wasn't worth the heat increase or power usage.

    It is a shame that you can't fold with CF enabled and that the CPU usage is much higher than nvidia folding, but they can still put out a lot of ppd.

    They can be, but they are still nowhere near the efficiency and ease of use as Nvidia at this point in time. Almost negligible CPU usage + vastly less interference with other GPU usage if set to idle (i.e. watching videos/playing games ... only get noticeable stutters if set to low rather than idle or watching/playing 1080p high content) vs BSOD/driver crash when watching videos if folding before video started, having to set environment variables to get better performance, and using an entire core to feed itself (as of the last time I used AMD). I agree that you can get a big PPD increase by comparing two cores of your six core folding SMP vs two 5870s but the "management" cost is much higher. The dual card problem afflicts SLI as well :/
    shwaip wrote:
    RE bonus points:
    I don't know the details. Adding 2x5870s to folding greatly increased my PPD, vastly more than enough to overcome dropping 2 threads from the SMP client.

    RE: even/odd
    I think it has to do with how folding divides up the work. The client restricts you to an even number of threads - I read somewhere else that the problem is with prime numbers of threads...but who knows. Just use an even number

    The even numbers thing is applicable to v7 client, but not to v6. In v6, many people use 
    fah6.exe -smp 7 -bigadv
    on i7s to fold bigadv and the last core to run GPU/other stuff. The v7 client does not accept odd numbers at all, and will round down. In your case, it was doubly beneficial because you needed two cores for GPU usage.
    csimon wrote:
    Don't you get bonus points on GPU folding? I thought that GPU's far outproduced CPU's. What the hell happened?

    GPUs did better when quad or more was not a "default" on decent machines. For dedicated folding, bigadv is the way to go. See my response here (that response is assuming 24x7 dedicated folding for the numbers I cite).
  • shwaipshwaip bluffin' with my muffin Icrontian
    edited December 2011
    Tushon wrote:
    They can be, but they are still nowhere near the efficiency and ease of use as Nvidia at this point in time. Almost negligible CPU usage + vastly less interference with other GPU usage if set to idle (i.e. watching videos/playing games ... only get noticeable stutters if set to low rather than idle or watching/playing 1080p high content) vs BSOD/driver crash when watching videos if folding before video started, having to set environment variables to get better performance, and using an entire core to feed itself (as of the last time I used AMD). I agree that you can get a big PPD increase by comparing two cores of your six core folding SMP vs two 5870s but the "management" cost is much higher. The dual card problem afflicts SLI as well :/

    I had no problems watching 1080p streamed content while gpu folding. Only problem I had was when I forgot to stop gpu folding before toggling crossfire for gaming.

    I'm not trying to argue that folding with nvidia isn't better - just trying to make it clear that folding on amd is pretty painless (and quite productive).
Sign In or Register to comment.