Graphics card not working after cleaning ? Guess what killed this 2080Ti

foreign hello internet today we have this EVGA 2080 TI info repair don't know exactly what's wrong with it so let's take it apart and see if we find anything interesting and boy oh boy look what I found [Music] a nut that was stuck inside a thermal pad my guess is this knot was shorting the RM area because it was sitting directly on top of the solder pads here I don't yet know what damage it did to the card so let's continue with the disassembly okay let's see if we have any shorts no short on 12th nor shorter 3.3 North Shore in these connectors not sure any of these coils fuse is good this fuse is good no short no short 15 on memory picks okay 1.8 no short data line look the same on the back data line also look good and clock reference and text reset looking good okay let's power it on and see what we have one and a half amp that's good we have 1.8 do we have packs yes memory oops no memory I also noticed the M draw dropped low it's very interesting we seem to be having a new symptom let's look at it again yep that must be GPU Ebola maybe loose connection nope that doesn't help yet as soon as I touch this area power cuts off so what are we losing here we have 5 volt 1.8 next is the core and there's nothing on it which means we are losing core and then picks and memory as a result what I'm noticing is that as soon as I touch something in this area here power is lost my guess is we have a bad solder joint somewhere in this area so let's Reflow this whole area and see if that helps Reflow did not help but problem became more consistent where power will come on and then disappear next let's take a look at the multi-phase controller and see if we're losing enable signal it looks like enable signal stays on even after a power loss and if we check the power good pin it goes cold when power cuts off in order for power good signal to stay high controller must have every enabled phase working properly so let's check every pwm signal at the controller maybe we'll find something weird looking at the oscilloscope all pwm signals are generated correctly so next thing I want to check is to see if the signal for each pwm makes a full circle from the controller to the driver mosfet and back into the sense pins it looks like every sense pins is a flat line which is not good out of curiosity let's take a look at the signal at the driver mosfet when it Powers the core Echo Echo no Echo but noisy clean Echo clean clean memory okay so we have a bunch of Echoes and some noise which is likely caused by the controller itself and I want to replace it however I don't have this controller available so what I'll do is I'll swap it with the one that drives the memory phase and see if it makes any difference foreign [Music] [Music] [Music] swap is complete let's check for shorts no shorts was found so let's power it on and see if anything changed looks like power consumption is now consistent but I still don't know if there are any issues with the memory face so let's verify the signal for all three memory phases as you can see they are looking good so let's continue checking the core faces no noise anywhere and right until I accidentally flipped the attenuator switch on my probe so let's put the switch back on and continue with testing time to book the card and run some tests but you may be wondering how come this bad controller works for memory since memory uses only 3 out of 8 phases my guess is that we got lucky and it uses those three phases that are working on that controller in any case let's power on and see if it works and we have no picture okay let's put into Linux and see if we can detect the card with the lspci command nope it's not detected okay next order of business I will read the BIOS copy the version from it and download that exact bias from Tech PowerUp flash it back and then back to the scope I want to see if we get a conversation between the motherboard and the bias chip first set of pulses on pin 2 is where bias chip replies to the motherboard after that there's no more talk but there should be in other words all we're getting right now is motherboard sees the GPU and says hi GPU says hi back and then they don't talk anymore this isn't normal and I almost wanted to go look at the PEX reset logic and then I saw this it's a blob of solder that shortened some of the pins that are also responsible for detecting the board I have no idea how it got there there's no one here but me so I'm the only one to blame I'll take care of that with the wick hopefully that's all there is and try again what I want to see now is more than just one request coming in there's hello my guess GPU responded with hello as well and now there is what's your name after which I'm sure GPU replied I am 2080 TI and the motherboard beep and said cool and look we even managed to pass a memory test with that said I can confidently say this card is fixed so now I will deal with the pads owner had sent two size pads but only one was useful I could only use one of his and then I ended up using some of my own so it is what it is at that point I wasn't even sure why he included these pads my guess was that he wanted all pads replaced so that turned out to be false in the end and then he just turns out that he was just being generous and wanted to donate me some pets uh but you know by that time it was already replaced all the pads on the card so eh who cares pads are free anyway card was assembled and I ran furmark just to see that I'm not very happy with the results and judging by the footprint heatsink is not exactly parallel to the Core I've seen this problem before it's caused by the pad that's on the right being too thick so let's make that pad thinner and try again this looks a lot better but only test will show the truth so let's load framework and see what we get 65 on the core 81 on the hot spot in comparison to an RTX Titan we are running 5 degrees lower three degrees lower than the founders Edition so my guess this is as good as we can get and at that point I thought that was it for this repair and then while doing more stress testing I found that it runs at X8 instead of x16 which means we need to check data lines and what do you know we have one signal missing that signal goes directly under the core so next thing to do is to lift the core and check for connections [Music] [Music] [Music] foreign I was able to confirm the connection to the pad so next is to check the core itself and unfortunately the connection is not present on the core which means there's nothing I can do except to put it all back together [Music] [Music] thank you [Music] I will apply a sticker before the assembly so that way it'll serve as a reminder to whoever is going to try to fix it next this is all I could do I'm sorry what's interesting about this is that after I ran Benchmark it gave me higher score than an identical card I tested a while back I guess this is a performance gain mod I don't know in any case this repair is now over I hope you learned something new today if you need a repair the contact information is in the description goodbye

Channel: northwestrepair

Views: 671,459

Keywords: best, how to, repair, fix, diagnose, identify, findout, lightning, boot, detect, nvidia, artifacts, micron

Id: LxlxdVVPFRE

Length: 14min 6sec (846 seconds)

Published: Sat Jul 08 2023