Announcement

Collapse
No announcement yet.

Nvidia Tesla P40 (Q0V80A) Enterprise GPU, I put the power cable on backwards.

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

    Nvidia Tesla P40 (Q0V80A) Enterprise GPU, I put the power cable on backwards.

    Ok I screwed up big time and hoping I can fix. I will break down what happened and the images.
    • Have Dell PowerEdge R730
    • Bought Tesla P40
    • Researched power cables and order correct one, pic had two black connectors in photos
    • Power cable arrived, same connectors (EPS-12v) on both ends
      Both sides are the same plug but different colors. The listing stated that colors would be random. OK I plug it in, red side to riser board, black side to GPU.
    • Power on.
    • System immediately turns off.
    • I only had 1 PSU plugged in and it was blinking orange, I thought I drew too much so I plugged in the other PSU and turned it back on again.
    • Immediately turned off again but that oh-so-familiar smell.

    I went to the computer right away looked at a blog post and saw that only one end (with the four yellows on clip side) goes to GPU and other to board. The listing I bought from did not state that so I didnt know. So I screwed it up, bad.
    • I put the power cable on the right way and then turned it on and the computer would see the card but nvidia-smi wouldnt detect any cards. ChatGPT said to try and older driver, we did, no go. Then it said that the card may be damaged and is powering it enough through the pci power to ID itself but something else isnt right.
    • I disassembled it
    • Visual inspection didn't show anything
    • Sniff test brought me to this one 1R0 inductor, its the only source of burnt I can find
    • I tested the four grounds and all of them were ok except the one noted in the photo, it shows OL, not sure if it should or not. Circuit is only closed to a couple pads opposite the inductor which smelled.
    • I watched the entire Learn Electronics Repair GPU Step by Step fault finding to no avail.
    • Have requested the schematic here (unable to find via search)

    The computer sees its a P40 but the driver won't initialize it. I believe that is because the board is getting power through the slot and that is enough to tell it what it is through PEX/bios but the rest of the card isn't getting the power it needs from one of the VRMs so the driver can't initialize it.

    I can soldier on a new inductor. Im hoping thats all it is. Could there be something under the inductor? Can any one help me diagnose the VRMs further to make sure nothing else is damaged? Please.

    TL;DR I put power in backwards polarity to this card twice. Im an idiot.


    #2
    when u see that videos u can provide measurements

    Comment


      #3
      Originally posted by ktmmotocross View Post
      when u see that videos u can provide measurements
      Thanks for the reply. So I have been busy and researching. Strangely enough all the the reading that I can be sure that I do seem to be OK. The power cable I was sold doesn't have the "sense pin" for the server's riser power port. So when I hooked it up properly the card got powered by the slot but not with the pcie power I purchased another card P40 and I had another identical server present. In that testing I found I got the same result with a known working card and server. So now I'm trying to get a proper cable going and then I will be able to sure that the card is working correctly or not and will update with findings and if it is not working I will dive deeper on the readings. So in short I did reverse the polarity not once but twice but the card and server may have survived.

      Comment


        #4
        Check locations Q503 and Q504 as well

        cmac

        Comment


          #5
          Thanks for the tip. Hope to get back to this soon!

          Comment

          Working...
          X