Thanks BZ for working on my Titan X! A bit of background: This card was randomly shorting out my psu for a few months before it was consistently shorting it out to the point that it would never boot with the PCIe power plugged in. At first I thought the backplate or the frontplate were shorting with the components as I noticed they seemed to be warping against the PCB but removing those did nothing to resolve the issues. Then I suspected a cap had blown after I smelled burning electronics on one of the occasions. As you mentioned it was a pretty fruitless task to try and pull of each cap to see if by chance it was the one that had blown. I think I've included the caps I pulled in a baggy with the cards I sent but if not, the card should still turn on. Pardon my soldering skills but working with SMT components without the special tweezer iron is really messy. I am hopeful that you'll get the card to boot up though I hope it won't be too much work. Pulling the inductors was definitely beyond what my soldering tools could do, without also pulling the pads out. I also forgot to throw in the waterblock I have for this. If you would like it I would be happy to send those too as I no longer have a use for them (the other cards being a water-damaged Vega FE and an over-volted Vega 56 reference that blew the core).

"GPU manufacturers try not to use a bunch of different screws" flashbacks to gamers nexus Nvidia founders edition teardowns with over 50 screws in total, with at least 5 different types

Ooh I hope to see a follow-up for this one. Good luck!

A buildzoid repair video. Yes yes yes 🙌😁.

Very useful troubleshooting tactics. I have an X99 board with a dead VRM. The PS clicks it’s over current cutoff relay when the power button is pressed, but not if I unplug the 8 pin. Thanks for the tip on probing the shunt resistors. I hope I can find them on my board. I suspect one of the IR3550 chips has a ground mosfet shorted. Vcore is not shorted to 12v. Please make more repair (attempt) videos, they would help us all understand power distribution, which affect OC’ing as well as repairs. Cheers!

Sometimes it can be helpful to measure the gate resistance to identify the shorted stage.

You have quickly become one of my favorite youtubers. I Don't understand most of what your talking about but your videos are super fun to watch and learn from! please keep it up!

Thanks! I was hoping for GPU troubleshooting and repair videos.

The MLC ceramic caps can create a short if they sustain physical damage. They're pretty tough, but they're also quite brittle, so if the card gets dropped or something, they can get cracked and create a short. Edit: I'm working on a 980 Ti right now with this same basic board design, and the short was that bottom power stage, as you mentioned. The card works after removing that power stage; no funky safety mode or anything. I'd even go so far as to say that, if you have a hot air station, removing the power stage is probably easier than removing the chokes.

Turn on captions for intro. Builds weed here

I cracked a 980Ti die on an Matrix Platinum from this exact fault, 12v burst right into the silicon, burning, smells, sadness.

4:40 You can apply voltage to see which caps blow up so you could verify the broken one and replace it

thank you

The cooler on this is the exact same size as the one on the EVGA GTX 970 I had, and that thing got up to ~80 degrees whilst sounding like a hairdryer (even tried repasting it, which didn't help at all). It took their initial Pascal cards literally exploding for them to fix their crappy coolers.

I didn’t even know they had a dual fan design titan wtf that’s cool

G'day Buildzoid, This was really cool showing how to check for a shorts on a GPU, Thanks so much 😁 Ps I'd watch "Can Buildzoid Fix it?" videos

The RAM overclocking t-shirt is hilarious. Black magic indeed. I wish the sticker came as a vinyl decal so the background took on the color of what it is placed on. Would be perfect for a case with a window.

If this core is still alive I would be very tempted to scratch a break in the twelve volt power rail and put a fuse in both rails right near the plug-ins easy to spot. The bit of information needed to do that is what is the maximum current draw with an undamaged rail.

If you have a DMM with 4-wire ohms mode (almost all bench dmms, but almost no handheld ones) you can sometimes find the short by probing different places on the PCB and moving towards the location of lowest resistance. This is a lot harder with low ground impedance cards like this though. I'm not sure if I'd try it, but you might be able to put a "safe" voltage (1V?) to the PCI connector and see what heats up. All the "working" powerstages should not boot up due to undervoltage, but the shorted one will deliver voltage to the core (if the high side mosfet in the powerstage is shorted). This would not expose the core to unsafe voltage since the 1V is applied through the shorted one to the core. You could quickly determine which powerstage is shorted, assuming that's the issue. This really relies on having an actually smart VRM controller and/or powerstages though.

I've ran into something as simple as the backplate grounding out the card after so many owners and PC swaps. Brian from Tech Yes City found this exact issue with a Sapphire R9 390X.

I had blown powerstage on my Gigabyte GTX 1080 Ti Xtreme Waterforce, it was fused to ground plane (no hot air station or soldering iron could remove it) so i had to remove it with pliers and destroy the remaining powerstage along with all supporting circuitry from that powerstage. After that the graphics card worked perfectly again, in place of that destroyed power stage there is small 0.5mm tall copper chunk that is fused to groundplane that cannot be removed. I think it was short circuit protection build into the controller or doublers that saved the GPU from painful 5-12volt death.

t he thermal pads that came with my EK gpu block fooled me the same way. i spent an hour before i realized it was probably supposed to stay on there..

Grab a bench power supply and apply 1v and 2 or 3A to the shorted rail and check for components heating up. If the core starts to heat up, just lift up 1 side of the inductors untill the short is out and you'll quickly find the culprit. Good luck!

I have 3 ref 980ti's that are dead. all evga sc, picked up a bunch of evga epower v's. quad sli + quad epower

Hi Buildzoid, guys here!

I get 1370 Mhz on my stock GTX Titan X. Never really gets hot because of the Raijintek Morpheus heatsink with two Noctua 120mm fans. I've increased the power limit, but I can't remember getting past 1.2v for some reason. It's been a while since I edited the BIOS.

dude. Teach us :)

When I had my first pc I had a evga 960. Had quite a few lockups ended up not knowing to send it in for rma and one day it smoked up on me.

Those look like the EKWB thermal pads. I also tried peeling off the the glossy layer :D

Man, to think of what AIB partners could have made if NVIDIA had just have given them some reference TITAN PCBs past maxwell. It would have at the very least, given a different options to aesetics and perhaps warrenty as well. EDIT: @Buildzoid if you could, could you make a video about what hardware (Like power meters type hardware) and software you use to make AHOC repair and testing content?

hey man i was just curious are you a electrical engineer or something? watched your server breakdown and was curious

Hi, I'm trying to make an epower using a pcb of reference 980 Ti like is this video, what should I look into to make all power stage working (except vmem which are desolder) ? like you talked about power load balancing, do you think there is a way to force all mosfet to work ? except for that I think I only need 3.3v, 5v and to set high an enable pin to start up mosfet driver, do you think it should be enough to power lower end card such as GK104/GM204 and lower at pretty high voltage ? Thanks you

Are you interested in a PCB breakdown of a Vega 56 Nitro+ non-limited?

What could cause L14 to be cracked and the close shunt bulged? What does that circuit?

i just got a Palit GTX 970 with that exact defect. i removed the inductor but it turned out the gate of the dead powerstage (only a single powerstages was connected to the 12V power plane that was shorted so it was easy to tell which one was blown) was shorted to ground. i removed the powerstage with hot air but the card still doesn't turn on. either it's detecting the missing stage and it goes into safety mode or me removing phase#1 on a card that only runs phase#1 in idle is the problem if the card can't detect if a phase is gone EDIT: after changing nothing about the card it started to work again. maybe it was still too hot from the hot air station or something else. it is booting now.

Still waiting for the 980ti kingpin to be fixed. Maybe the titan x transplant on the kingpin pcb?!?

Those look like EK thermal pads if they are that's how they are shiny on one side

Is there a follow-up with that card? Curious if the core itself survived 🤔

this video was too short wasn't able to fall asleep ;D

I hear 2 buildzoids at the beginning

those kind of thermal pads don't have plastic on them. They are just... crappy like that. I destroyed one trying to get the "plastic" off.. I don't know why they exist. I don't know where they come from but I see them sometimes on used cards where an owner changed the pads for some reason or another.

Does the titan x maxwell (same gpu as in your vid) have a fuse ? Been looking all over this card and still have not found any.

Actually Hardcore Overclocking, What would you say in this situation: My Titan x Maxwell when running load like Valorant and sometimes it does this with League of Legends but the issue is that I get a graphics card error, white screened until the game is is 'end tasked'. I noticed it was getting really hot about 83c. I changed the thermal pads and paste and it has been running cool but its still doing the same thing with the white screen and graphics card error. I inspected the card's components but I didn't see anything out of the ordinary visually.

"If you have an overpowered VRM it is extremely unlikely it will fail, ever". This is basically why I buy mid range overclocking motherboards even though I'm not an overclocker. The better components are good for longevity. I have a Rampage IV formula motherboard that's close to 8 years old and a Sabertooth X58 that's about 10 years old and they're still going strong.

I didn't know that a custom titan x existed.

BZ should make his own graphics card, I'd buy one lol & I got a GTX 460 cyclone that posts but the image is all fudged up, not sure what's wrong with it . Got it for only like $20 on eBay 😂

Louis Rossmann uses some kind of alcohol to detect shorted MOSFETs on Macbooks Pros, why not try something similar here?

I never buy reference Nvidia PCB's anymore. They fail so many times. Had a GTX 780 from EVGA with reference Nvidia PCB, it died after two years. They then gave me a refurb 780, which also died within a few months. These cards easily hit 85-87 C under load despite custom cooler, imagine how hot those VRM's must have been. Subsequently bought a custom PCB MSI 1070 and it still runs perfectly after nearly 3 years (also much lower temps). Only buying custom PCB cards from now on, I want cards that can survive at least 4 years of daily abuse.

“Let’s get the plastic tub for all the parts” No GN modmat?

How did those heatpipes get so dented up?

I have an EVGA 980 Ti with the bottom VRM blown up, im in the process of replacing it. I got the VRM removed already and im waiting for the component to arrive. Soldering and desoldering on that spot is a pain tho.. I needed 400 degrees on a SMD reflow station to warm up the groundplane enough. Im not sure if I can resolder the new component properly without overheating everything. Can the card also run with a powerstage less?

Why buildzoid uses such an expensive multimeter? Are there any other benefits apart from, I assume, reliability and precision of the measurements?

i cant remember how i clicked an unlisted video what

I think I just blew my GTX 1060 after overclocking it, its detected but error code 43 on device manager, if I connect HDMI directly to the card, nothings are shown up.

ek sends the same thermalpads

Buildzoid Mod Mat aka blue shop towel

I have a question, is a 16 phase vrm overkill for a 3700x?

Could you theorecticaly upgrade the power stages?

Last time i was this early Buildzoid hadn't grown hair on his arms yet!

rs1 not reading and rs2 and rs3 and my apw7142 vr883 chip is getting hot. can you help me pls?:D

I have never had an amd gpu ever die on me. Can`t say the same for novideo gpus tho.

it's funny because I've bought dead cards that also had the plastic on the pads, lol always a good feeling!

LoL that swiss army knife gives me the verge flashback

epower it!!!!!!

Just scored one of these for $120usd.

I had no idea there were titan cards not directly from Nvidia

You know what would be really cool? If the 3300x existed and could be purchased for $120.

HELLO

First repair then oc until broken!!!🤣

..NVidia has prob invested in a shitton of those VRM chips and is going to use em even though a different design with better controllers would be better for long term performance..the reality of economics..

Please clear that thermal paste. Thanks for video.

Kind of shitty to get a (even if broken) Titan X, forgetting about it for ages, and not even bothering to contact the sender with a "sorry, what was the problem?" or looking back at it, tbh.

Builds weed hahahaha

first