GTX 295 VRM1 Flashing On and Off During Load (Riva 2.24)

Discussion in 'RivaTuner Advanced Discussion forum' started by LedHed, Jun 9, 2009.

  1. LedHed

    LedHed Banned

    Messages:
    6,827
    Likes Received:
    1
    GPU:
    eVGA GTX 295 /BenQ FP241W
    Hey Guys,

    I recently got my 3rd GTX 295 (2nd RMA) and I have noticed that during load (mainly COD W@W) the VRM1 only will continue to flash on and off. I noticed before that it will flash incorrect readings for temperatures or amperage for a split second but never flash off for 20-25 seconds at a time and then show back up for a split second. However when I'm not in load it never flashes or acts funny at all.

    The reason I'm asking is because it would be very important if this is a hardware flaw as my last 295 had a faulty VRM1 and it's just too perfect to be random. If anyone knows of a bug or glitch that can cause this (be it settings/drivers/clocks/voltages) please let me know, because I'm going to call eVGA Tech Support in the morning and ask them about this if no one can give me an answer on the hardware side.

    In the meantime I'm going to monitor with GPU-Z and see what it says during COD W@W.
     
  2. LedHed

    LedHed Banned

    Messages:
    6,827
    Likes Received:
    1
    GPU:
    eVGA GTX 295 /BenQ FP241W
    I just wanted to confirm this is happening in all heavy GPU load applications, not just COD W@W. I'm going to call eVGA shortly and ask them about this if no one has any idea on the software end of things.

    I'm hoping Unwinder has some idea of what is causing this, or maybe he'll know it's a hardware problem. I noticed that even my 8800 GTS 512 would flash the VRM with weird readings (like 200C) for a split second but then return to normal. It really seems like a hardware malfunction and I'm about to test with GPU-Z and compare my results to RivaTuner.

    One more thing to note is this doesn't start happening until the GPU/VRM heats up to above 70C, before that it doesn't flash on and off at all even with full load. This really makes me feel like it's the hardware since RivaTuner shouldn't be effected by the heat, however the GPU-Z log file doesn't cut out during this time.
     
    Last edited: Jun 9, 2009
  3. LedHed

    LedHed Banned

    Messages:
    6,827
    Likes Received:
    1
    GPU:
    eVGA GTX 295 /BenQ FP241W
    Ok so I found out the cause to this (at least I'm pretty sure), I talked to upper level tech support at eVGA and listed below is what I found out from them:

    I found out that RivaTuner has to be giving a false reading on the VRM1/VRM0, because when showing above 60A on each is above the rating that a standard (US) wall plug can provide. They think that RivaTuner is taking both VRMs and reading them as one but then listing it twice as two separate VRMs even though it's already combined both VRMs ratings to 50-60A (in peak stress). I verified this with GPU-Zs VDDC Current reading, which were 55-60A and tech support said my power rating was around 95-97W per PCB which is well within spec.

    So basically the reason I think VRM1 flashes on and off is because it's not an accurate reading when both VRM1 and VRM0 are showing above 60A each (which is impossible) and it's probably glitching from some type of conflict in the software due to how it's displaying/measuring the amperage.

    Unwinder is there anyway to remedy this issue and get correct VRM readings? GPU-Z is displaying it as "VDDC Current (A)" and these readings are correct, so there must be a way to get the correct reading. You could just change VRM1/VRM0 to VRM and list the current as a whole for the entire VRM instead of listing each separately.

    Also I noticed that if I removed VRM1 Current from the LCD (G15) and leave just the 3 temperatures then it would not flash on and off anymore, so there must be a conflict here.
     
    Last edited: Jun 10, 2009
  4. Unwinder

    Unwinder Moderator Staff Member

    Messages:
    14,437
    Likes Received:
    1,462
    Such a long posting and short usual reply: read documentation. There is the following note for a long long time:

    "Added support for voltage regulator output and average current monitoring on NVIDIA GeForce GTX 260 / 280 series. Please take a note that Volterra voltage regulators are rather sensitive to frequent polling and may return false data under heavy load, so it is not recommended to use VRM monitoring in daily monitoring sessions."

    Also, the tech's comment about "They think that RivaTuner is taking both VRMs and reading them as one but then listing it twice as two separate VRMs even though it's already combined both VRMs ratings to 50-60A (in peak stress). " is a complete nonsense. Doing so whould be just a plain stupid and has no logic from any software developer POV. And for real techs all monitoring plugins inside RT are open source meaning that any competent person may examing what is read, how and where.
     

  5. LedHed

    LedHed Banned

    Messages:
    6,827
    Likes Received:
    1
    GPU:
    eVGA GTX 295 /BenQ FP241W
    Thanks for taking the time to explain it Unwinder, I apologize for not fully reading the documentation. It's good to know that this isn't a single case that I'm experiencing and thanks as always for your amazing work with RivaTuner. I also forgot the 280/260 use the exact same Volterra VT1165MF as the 295, so while I did see postings on the 280/260 having some issues I wasn't sure if the same applied to the 295. Again thanks for the reply.
     

Share This Page