GTX280 not GPU3 folding

csimoncsimon Acadiana
edited February 2011 in Folding@Home
My GTX280 hasn't accomplished much in the lines of GPU3 folding. I've set it up a few times but after a few days of folding I get a screen full of artifacts. When I reboot it goes into a solid colored screen and is non-functional. The only remedy I've found is to boot to Safe Mode and remove the drivers. After a clean install everything is fine unless I try to fold again even after an extended cooling period of a few hours. I've tried both the 6.31 console and the 6.41 console and I have had the same issue occur each time. On a side note, I've had to add a file called cudart.dll just to get it to work in the first place. I don't know if that is a contributing factor or not.

No graphics overclocking whatsoever.
extra_parms=-gpu 0 -forcegpu nvidia_g80
Whindows XP Home SP3

Right now I've just reinstalled drivers only (266.58) and I'm going to give it a few days before I try and game on it. I'm CPU folding just fine at the moment and everything seems ok.

Can't think of what it could possibly be. I was wondering if this was a common issue with GTX 280 in some particular case. I'll keep researching this for now.

Thanks,
:csimon:
«1

Comments

  • _k_k P-Town, Texas
    edited February 2011
    what about gpu2?
  • csimoncsimon Acadiana
    edited February 2011
    _k_ wrote:
    what about gpu2?

    Idunno haven't tried it. Why, is GPU3 an issue with this card?
  • _k_k P-Town, Texas
    edited February 2011
    It could actually be the board, if it is then GPU2 should provide you with the same issues. I ran into this when I was trying to build another dedicated folding rig for a 9800gx2 with an old intel media board with a 915 chipset. It would not even bother to start booting up with a 9800GX2 installed. Granted that is a wild shot.

    I would first try folding again with an underclock and watch the temps, or you could do something crazy and try loading the GPU with another program to see if you replicate what is going on with GPU3. I hoping it just ends up being the fact that you are running an old chipset on your Abit and it doesn't like the heavy traffic that is caused to the PCI-e bus. I had folding kill my asus crossfire board because of that, one PCI-e started acting weird and then not booting with both cards inserted and some other things pointing towards the south bridge and then no mas.
  • csimoncsimon Acadiana
    edited February 2011
    Why underclock instead of config it for let's say 75% rather than 100% load? Highest temps have been 83% while folding. I did see the artifacts while flying IL2 one night but after I rebooted everything was fine. The first time this happened before I yanked the card for a few months it wouldn't even go away after I removed the drivers. It was pretty bad. I was thinking about baking it but when I tried again recently everything was okay for a few days then wham. But now I can rectify it by removing the drivers and reinstallying. I'm using only drivers and nothing else ...no physics or nview.

    Also, what about the cudart.dll? Is this a one size fits all or could it be possible that I have the wrong one?
  • _k_k P-Town, Texas
    edited February 2011
    because
  • csimoncsimon Acadiana
    edited February 2011
    _k_ wrote:
    because
    k. Then what's a good software to use to underclock a video card ...ntune?
  • _k_k P-Town, Texas
    edited February 2011
    msi afterburner. The goal would be to underclock the memory by about 50-75 mhz, if you get down to 100 mhz with the issue continuing then return it to defaults and try GPU3 to only use 75% of the card.

    That cudart.dll is universal.

    Have you surfed the folding forums yet on this issue?
  • csimoncsimon Acadiana
    edited February 2011
    _k_ wrote:
    msi afterburner. The goal would be to underclock the memory by about 50-75 mhz, if you get down to 100 mhz with the issue continuing then return it to defaults and try GPU3 to only use 75% of the card.

    That cudart.dll is universal.

    Have you surfed the folding forums yet on this issue?

    Not yet. I've surfed the card itself and apparantly it happens quite a bit from what I saw. Almost sporatically though. Like artifacts when starting a game, then reboot and all is well.
    Some think it's a solder issue. I've thought about baking it too. Maybe after all else fails. Or maybe I'll wait for that after I've tried it in a new rig. That should happen in the near future.
  • TushonTushon I'm scared, Coach Alexandria, VA
    edited February 2011
    Use Furmark to test the card at current clocks and see if it artifacts.
  • csimoncsimon Acadiana
    edited February 2011
    Tushon wrote:
    Use Furmark to test the card at current clocks and see if it artifacts.

    Good idea ...I think I'll try that first. Will it work with only video drivers?
  • _k_k P-Town, Texas
    edited February 2011
    Why wouldn't it? I have no idea if Furmark will load the GPU more than OCCT or MSI's program will. Really you could be our test subject and track down a few programs to stress your GPU with for a few hours and log the temps then report back.
  • csimoncsimon Acadiana
    edited February 2011
    I attempted Furmark benchmarking. I set 1680x1050, fullscreen, temp monitor, then start. As soon as I hit start I have instant frozen artifacted screen.
    Reboot - same thing as soon as drivers have time to load.
    Safe Mode - remove drivers, and drier cleaner pro.
    Reboot - install drivers.
    Reboot - driver load, artifacts again.
    Typing from Safe Mode now.
  • TushonTushon I'm scared, Coach Alexandria, VA
    edited February 2011
    That means your card is bad, assuming it is not the slot. Sorry!
  • _k_k P-Town, Texas
    edited February 2011
    RMA or sell it on craigslist.
  • TushonTushon I'm scared, Coach Alexandria, VA
    edited February 2011
    _k_ wrote:
    RMA or be a huge doucher (unless you meant sell it to someone who intends to fix it).

    You could also try baking it at this point.
  • csimoncsimon Acadiana
    edited February 2011
    Well I'm not so quick to write off this card yet. It could be the fact that this old motherboard can't support it. It is PCI-E 1.0 afterall. Shouldn't that be an important factor?
  • _k_k P-Town, Texas
    edited February 2011
    Not for it being able to display without throwing artifacts on the screen. When you run a faster card in a slower slot what really happens is the FPS will not be as high as possible because the socket cannot provide enough data traffic some of the times the extra side effect of having a PCI-e 1.0 slot is that the older proc and ram hinders the GPU more than anything else.

    When I switched from a 1900xt to a 8800GT g92 I had to run a wicked OC on the proc to get my bus and HT up plus get my ram at 474, while still holding 2-3-2-5, to not have parts of games where huge amounts of frames would be lost and get closer to the max performance the card would do.
  • csimoncsimon Acadiana
    edited February 2011
    Yeah I figured I would save it for the new build. I'm torn between the Bloomfield and Zambezi. Just waiting to see Zambezi performance and price when it comes out. May not wait though still undecided. Maybe when taxes return it will itch in my pocket. I think Zambezi will be rush out of the door because of the P67 setback. At least that's what I've been reading.
  • _k_k P-Town, Texas
    edited February 2011
    but when are we going to get our hyper-threading....me want.
  • csimoncsimon Acadiana
    edited February 2011
    Why hyperthreading over hypertransport?
  • _k_k P-Town, Texas
    edited February 2011
    why can't I have everything, hyperRAM hyperHDD hyperPCI. Because if anything has taught us it is that adding more threads equals more performance.

    But depending on their speeds the 8-core amd procs will be really fast.

    X3440 @3.8GHz Bigadv - 25,000
    i7 965 @3.03GHz SMP2 - 13,560
    E5506 @2.13GHZ Bigadv - 10,389

    the e5506 is slower than them all but when you put two of them together you and are able to run bigadv then you have a ppd around 20k at a much lower speed than the hyperthreaded procs.
  • csimoncsimon Acadiana
    edited February 2011
    I have 8 cores of 2.0ghz xeon at work supposedly can that make bigadv? Probably dual quads I suppose.
  • _k_k P-Town, Texas
    edited February 2011
    Yup, you have to be able to take 8 threads either by real cores or HT. If you launch a bigadv assume it will take 1GB to 1.5GB of ram. The machine I used to run it on was using 1.75 out of 2GB with only running a fresh install of server 2008 R2.
  • csimoncsimon Acadiana
    edited February 2011
    How is bigadv set up differently than regular SMP?
  • TushonTushon I'm scared, Coach Alexandria, VA
    edited February 2011
    It isn't. I do believe you just add -bigadv to your switches (stop service, launch with -configonly, add -bigadv to the switches in the advanced section, start service)
  • _k_k P-Town, Texas
    edited February 2011
    you also have to have memory set to large, otherwise you keep pulling SMP2 WUs.
  • csimoncsimon Acadiana
    edited February 2011
    I can do that. I already have memory set to 100% I believe but I don't have the -bigadv switch set I know for sure. I thought that was already an option in the config? Like bigpackets=big.
  • TushonTushon I'm scared, Coach Alexandria, VA
    edited February 2011
    I think the config option you are referring to has to do with memory usage, not flagging your client for bigadv WUs
  • csimoncsimon Acadiana
    edited February 2011
    Idunno. I'll be running a -configonly config setup to add -bigadv when I get back to work and I'll see it then though.
    I'm really curious to see what zambezi will do with f@h and what the price will be. I'm holding off on a new system for that.
  • csimoncsimon Acadiana
    edited February 2011
    It looks like it will take roughly 5 days total (120 hrs) for my first wu. Is this sufficient or should I just envoke the -oneunit and hang it up?
Sign In or Register to comment.