I recently handled one of those pads for an SBC. The thin film is completely invisible to the naked eye, the only reason I didn't fall into this trap was that I remembered to check right before mounting the cooler.
Former NT driver dev here: the kernel bugcheck (0x116) VIDEO_TDR_FAILURE seen at 6:13 is not actually that uncommon. It means that the WDDM timeout detection & recovery (TDR) watchdog failed to reset the driver and GPU state. This is usually indicative of either failing hardware or a seriously bugged driver that got the GPU into an unrecoverable state (the latter being pretty rare nowadays).
That film left on the pads is why some of them, e.g. Gelid, have something printed on the film. Gelids only have them printed on one of films so peel the blank side off first and stick them on, that way you can see which ones still have the other film on before you assemble the card. The resin bonding for the VRAM chips nearest the PCIe slot will be to stop them peeling off as the PCB flexes. Loads of prebuilt PCs have had that problem due to the G forces the cards endure as they're being couriyeeted.
That black residue around the memory chips is called underfill. Yes, they use it to strengthen the structure of BGA components near PCI-E socket, where bends can occur. Some reworkable resins can be removed with heat, whereas others, like the one shown in this video, are impossible to remove without brute force.
I'm a regular viewer of NorthRidgeFix and somehow I stumbled upon your channel. Glad i can find another tech who are so talented like you! I absolutely will looking forward to your other repair videos!
Northridge was supposed to send me a bunch of cards for repair. Once we agreed, I have not heard back. Not sure what his deal is but from what I understand about his operation, 90% of all GPUs on his shelf he gonna call a no fix while in my case, 90% is a fix. Just takes few days and I charge a flat fee.
@@northwestrepair yeah, Alex is a very busy person and with that much of work that has to be done, he rather choose to deem "no fix" than to actually fix it (only for devices that takes too long time or truly unrepairable) because it's not worth his time to spend. I bet he work on tens or more devices on daily basis.
Yo northwestrepair, I was actually the one who bought the RTX 2080 ti off you on ebay, (I was the guy who ordered on Nov 22nd) as I saw at the end of ur video looked like the video in the description of the RTX 2080 ti I bought, and also had no backplate as well, glad to see the RTX 2080 ti I have won't fail on me 😅, still many thanks tho, and glad I stumbled upon this video as well
Another key note, that is a Micron variant. The Micron 2080 Ti's were bombs in general. I bought a XC Black, the first two I received were Micron, and artifacting out of the box. The third came with Samsung memory, and have had zero issues since (2020).
This was a fantastic video, mesmerising to watch you work. It was also very cathartic to see the Uningine Valley pass at the end, my go-to when celebrating the birth of a new, hopefully-stable rig.
This level of repair is hardly seen these days as people either don't have the knowhow or are too impatient to get tech repaired. It takes years of skill, knowledge, failures, screwups and experience to point the exact problems get things working again after doing all the necessary changes and repairs/replacement. Hats off and 🙏 to you buddy all the way from 🇮🇳.
Expecting the average consumer to be able to solder at this level of skill just isn't going to happen. We should repair more and I always open broken tech up to give it a shot, but such small surface mount components require incredible skill and precision - not to mention all of the equipment needed, from a quality heat gun, multimeter with needle probes, a good soldering iron (usually with temperature control) and a thermal camera - plus buying the memory chips, which are often so expensive in anything other than bulk quantities that would leave consumers with a mountain of useless extra parts that consumers have no incentive to do it themselves. Repair needs to be made easier and cheaper through right to repair initiatives, not complaining that the average person somehow hasn't invested hundreds of hours practicing SMD soldering, or into nice repair equipment and doesn't have several donor boards lying around.
FYI, this residue is called "underfill" and it's purpose is to avoid/lower risk of cracks in solder joint's if it's getting under physical "stress". It's usually applied in devices which can exhibit that (like phones and other handhelded devices). Apparently wohever designed that, came to idea that memory chips near PCI-e slot are also at risk of that so that's why it's there.
its already know that early micron memory on 2080 ti failed a lot. tons of news stories about it when the 2080 ti launched. gave the card space invaders artifacting.
There are times when I'm envious of someone's knowledge and expertise and this is one of those times. In saying that - well done for the fix and i hope you see some reward for your efforts. I have a 1080ti which needs repairing, i genuinely believe t it can be fixed - I just don't have the knowledge to even start diagnosing where the problem stems from. The card is still operational but under stress it crashes.
I have almost the same issue but with an R9 280X that for some reason, I was playing call of duty warzone before the cold war forced integration and suddenly image frozen but the system didn't recover and while booting, the gpu never worked again and I had to remove it because the system was getting frozen before the POST so I think this gpu died or something horribly wrong happened and it needs dedicated repair, that would cost me way more than what I spent to buy it and some people either don't want to repair a gpu or they don't know or can't be bothered to order the parts necessary for the repair to be done.
try to underclock often times it just cant hold certain clocks anymore or try adding voltage at lower clock frequencies in frequency graphtuner from msi afterburner.
That 2080TI ,they had a waterblock on there and put the stock cooler back on later. That's why the screws are mismatched and why the thermal pads are all freaky looking. The stock thermal pads look very different.
It's interesting to watch your repairs. Wish you good luck. Some gpus repairs maybe won't pay back, but maybe some new experience you get push you bit further avoid any more same symptoms on bad/dead stuff. It reminds me my old GF 8800GTX died after 2y heavy gaming long 10 years ago. I put it to service, was something like heat treated and UV too(Not rly sure), it worked again for another 6mo until it died again. That previous 20-25USD(500CZK) repair cost I didn't wish to repeat. That was time to get new gpu.
This tells me so much,first how good you are at diagnostics and secondly,the reason for resin around the memory, is because they run this memory so hard it gets hot enough to soften the BGA solder,the resin stops the memory moving around, that explains why so many fail just after warranty, i used to buy high end cards,now i dont,just money down the drain.I remember the early days,when graphics cards usually outlasted the PC,the only reason to replace a graphics card back then,was to upgrade.Seems like everything now is made by, make fast buck companies.I liken it to buying a ferrari,it's great at waht it does,but you pay the price for this performance & its engine isnt going to see many 10000 of miles,before a rebuild.Great vid,thank you.
That's a BGA underfill and is not there because the solder is melting. The chips would stop working long before it ran hot enough to do that and holding it in place would only do so much anyway. Underfill is meant to relieve stress on the BGA solder balls. Usually to minimize the effects of thermal expansion over time but Nvidia doesn't really care about that. It's only on the chips nearest the PCIE slot because they're the most prone to damage from board flex. I suspect it's also there as added protection for OEMs and prebuilts shipping these pre-installed in systems. It's also likely part of the reason Nvidia has moved as many chips away from the slot as possible on the heavy 30 & 40 series cards.
@@JJFX- ok thats fair enough,i was thinking low temp solder 188 deg c,as i would imagine it would allow for a very small amount of thermal movement,under overclocking for instance,ive also had laptop system boards with resin around memory BGA's ,i wonder if it's the same reason. Do you think high end graphics cards are always going to have a high failure rate because of thermal stress,or is it something else? As for the small SMT tant caps,ive repalced a fair few of those, bit like the larger ones,they go short circuit as well. As the old saying goes,the cooler you keep components ,the longer they live.
@@enoz.j3506 Well this is almost certainly all lead-free solder which really melts well above 200C. Issues with the GPU itself tend to be on the large dies as thermal cycling has a more significant impact on the substrate joints over time. Different parts of the die heat up at different rates so the effect of thermal expansion become more significant. How the die is bonded to the substrate, the type of underfill combined with acceptable production tolerances have been difficult to get right while staying economical. This a big reason for the infamous Xbox 360 RLOD early on. Hard to say how modern chips will hold up. Package engineering is very complicated and Nvidia has made plenty of mistakes in the past. However advancements in underfill and BGA assembly have come a long way since their more notorious issues. That said, they are dancing with the devil by using bigger and bigger dies pulling more and more power. I definitely agree that those looking for longevity should keep everything as cool as possible. IMO, putting a full size water block on $1000+ GPUs capable of well over 350W is a no-brainer.
yes the thermal expansion in all dies is very difficult to control,if you think of the 28 billion transistors+ ,it is amazing it lasts as long as it does, yes i remember the xbox problems.
@@enoz.j3506 Yes it's pretty mind blowing. As processes get smaller and smaller it'll be interesting to see how the transistors themselves hold up. Interestingly, Nvidia's vcore target has largely stayed the same since Pascal despite going from TSMC 16nm to 5nm now.
That silicon around the bottom memory chips is to protect them from coming loose/solder joints cracking from the sheer weight of the cooler that the GPU manufacturers started implementing quietly a couple of months into production.
Do you know if other GTX / RTX or even RX cards tend to have that same type of Micron vram chips? Asking so I know what to look out for when purchasing used GPUs.
Browse through his older videos. He already did some repairs and mentioned it there. IIRC it´s related to 2018 manufactured Microns. You will not find it out, tho. At least not without full GPU disassembly and thorough cleaning of the chips. GPU-Z does not offer this kind of info.
can you explain more to depth about why 8 ohm on that coil means memory modules dead ? is it too little resistance ? too much ? there isn't supposed to be resistance maybe without coil ? its all a big mystery for someone like me with basic understanding.
Dumb question: Assuming that the warranty is not voided with this kind of teardown, does the unpeeled plastic cover on the thermal pad at @6:53 constitute "faulty workmanship"?
As someone who got a comp sci degree, therefore only knows programming, not hardware, I like your funny words magic man! Note: I know basic hardware not GPU and such
your CS degree DIDN'T teach you basic hardware? what sort of CS degree is that? I had multiple modules dedicated to hardware. Of course the electronic engineers have more in depth knowledge, but we were expected to be able to design (i.e traces, component positions), fabricate (milling etc...) and assemble (surface mount inductors, caps, resistors etc...) a PCB and accompanying hardware, and program the MC on said PCB to control any accompanying hardware.
@@lewisd56 I am currently doing mathematics and computer science (UK joint degree) and I never got any hardware course, even compilers were exclusively taught to computer science people. I've mostly just had theoretical stuff as compulsory courses and the rest to choose myself. And among the options we get, there's basically nothing hardware focused. Still, I have been studying more on the side because I'm afraid of coming out useless from university even if I have pretty good grades.
@@leonro UK computer science degrees too, (BSc and MSc), I had dedicated modules for embedded systems (C++ programming arduinos and other MCUs as well as designing accompanying circuitry), even had a module that went into specifics of logic gates and operations in a CPU. The MSc took a step back from the hardware, but I specifically picked the hardware oriented modules, so ended up working with more complex embedded systems and designing a whole PCB from scratch.
Crazy it got resurrected from the grave. Alot of knowledge and skill here. Remember there was a big thing with this card about the memory having alot of problem. RTX 2080 Ti owner here and I haven't had a single issue thus far and I push the GPU to the absolute limit. Still, can't wait to eventually grab an RTX 4090 this upgrade
Sure. Struggle reading? Different type of memory were outfitted on the 2080 Ti depending on the purchase people could end up using a defunct version destined to be RMA or have severe longevity issue. Luckily I didn't get this particular chipset Micron or Hynix I got Samsung, but this could be an issue going forward. Not that would I expect you to comprehend this, drug addict.
They lasted the warranty Period. Wouldn't be surprised if Nvidia knew that and used them anyway so they can sell more Cards, given their usual scumbag behavior. EVGA pulled out for a reason.
I added a shit ton of additional pads to my 2080 when i got it after replacing the old ones and it still gets crazy hot on the back plane. Probably going to have to add heat sinks to the back shield with thermal adhesive.
@@nicekeyboardalan6972 i think 3000 series will face different set of issues, mostly due to all the shortages, that were happening during their manufacturing. I expect something similar with AMD cards too. Many companies had issues during Covid shortages and were using literally anything to make their products.
I've had two cards die since I bought mine in 2020. First two cards were PALIT 2080ti Gaming Pros, both died due to issues with the GDDR6 memory - pretty serious problem from what I've heard from other 2080ti owners. Sucks even more that the RMA process takes around a month due to PALIT being in china. Thankfully, they ran out of PALIT 2080ti Gaming Pros to send me last year due to the 3000 series coming out. 😂 They sent me a PALIT 2080ti Gaming Pro OC which is a card that costs £600 more than the £999 card I bought due to much higher quality components, thermal cooling additions and a much bulkier design for heat dicipitation. Haven't had the same problems with this card as the cheaper ones (touch wood), so I'm assuming it's due to thermal performance or the issue was fixed post-june 2021 (first two cards died within 4-6 months).
with the resin on the bottom memory ic's... I think evga might have determined that the pcb flex near the pcie connector plus thermal cycles leads to those sm components breaking contact.
Bought a used Strix 2080Ti for $274 (incl. an installed Strix waterblock + the original OEM shroud/fan/heatsink assembly w/backplate). As I do not do water cooling, I removed the watercooler for resale and installed the air cooling assembly/backplate (after a deep clean, thermal-pad check, and re-paste). Unfortunately, the white shroud for the male FAN power connector to the GPU board was damaged (as-is) and although the pins make good contact, the shroud does not securely maintain the connection. Tie-wrap to the rescue. Done. Now the card runs perfectly well. Benchmarked SAT. I then did an under-volt using MSI Afterburner...for safe measure. GPU runs cool and stable...very happy with a fine 2080Ti that cost me, NET, about $220.
Remember Micron from 2019? It has big troubles for 2080's as Hynix GDDR6 1gb modules for 3060ti. Yet appearing after years of service. Noone is safe from such things to happen, sadly. It also depends on cooling too :p
Could you please make a video explaining mods you make to your own cards to make them more reliable and less likely to experience part failure from overheating? What assortment kits you recommend, assuming it's shims or pads, etc.? Thanks!
@@northwestrepair my HD7850 worked for 12 years non stop for me. Gaming 9 hours or using 3d max 9 hourse a day, it didnt matter. This gpu still can work after all this time when it was maxed out everyday for 12 years. I got a 3080 but that amd card can still work I had a golden rule. Which were never broken when i had this gpu. NEVER EVER overclock NEVER. Over clock is most stupid thing people are doing with their gpu. It will sap away life span heavily. At very least this was my experience. My last pc was 20 years ago as well. It worked for 10 years. I never overclock that either. Normally i build new pc for every 10 years, hope new 3080 work fine for next 10 years. Its undervolted and kept cool while i am working with pc
Overclocking isn't dangerous as long as you don't overvolt too much and keep the temperature in check. My old ATi 4870 1GB is still working fine, even tho it has been heavily overclocked since the day that I bought it in 2008.
@@waifuhunter9709 "Over clock is most stupid thing people are doing with their gpu. It will sap away life span heavily." yeah you're actually retarded if you think people use garbage GPU's for more than 2-3 years bro these days, even back when overvolting was a thing it wasn't relevant Either way, just overclock anyways, why? Well... Overclock doesn't mean OVERVOLT, you can't even do that anymore on 99% of gpu's stop being stupid just because you fucked up
@northwestrepair could you explain to me how to find the gpu schematics showing the layout and values of each component like the one shown in this video? Also how do you know exactly where to touch your voltage tester probes when looking for a short in memory, resistors, vrms ect..? Would you be willing to do a video explaining how others like myself can get a beginners guide to checking for damage/ short circuits using a voltage tester to check for ourselves? Any tips or advice would be greatly appreciated. I genuinely enjoy watching all your videos and each one inspires me to do some troubleshooting myself (I have a decent but still beginners level of experience with electronics) but i am to nervous to attempt anything more than a thermal pad/ repaste on any of my GPUs.
When watching any of his video's you can see the results of a true technician not somebody that thinks they know what they are doing! He has repaired a card for me with 100% excellent results, so if you have a card and were worried about his honesty or work ethics fear not he is the best !
Makes me wonder how my 1070 is still chugging along with almost no issues. A motherboard swap when I upgraded from my r7 1700 to an r5 5600x got rid of intermittent snowflakes... Why? No idea.
Repairmire:Diggie digi good work partner if i were near you would buy this art sadly we are far from each others always watching your videos for learning something new keep it up my friend
This is why ebga has stopped producing GPU. I have a 1 year old 3060 ti sc gpu that I suspect memory has died. They sent me a replacement and not even a week after I have artifacts and cannot be reinstalled anymore. Pretty sad.
Seems like using some kapton tape on those small components would make life a bit easier in that they wouldn't blow away when you apply heat to the memory chips.
I suspected it was failing memory on my Sapphire 280x it would have a square on squeen randomly like swap places with another square and parts of the textures would get mixed up/missing, and would commonly get horrible tearing issues with meshes. Also motherboard handshaking issues. It cleared up a bit after a oven bake (not proud of lol but it was that or it was dead) interesting to see and not as scary as I thought to replace memory now that I've gotten alot more experience soldering just haven't worked much with BGA
Hmmm, why are you so certain that 2019 micron chips will not die as well? Wasn't it confirmed that even Samsung memory version has a similar failure rate? Were there some recent discoveries about this that I've missed? I have a failing 2080 Ti Founders Edition, it's still 'working', but launching any game will quickly cause artifacts and a crash to desktop. Underclocking memory by -300 MHz fixed the problem temporarily, but that didn't last. Surprisingly, it passes MATS test. Now I want to open it up and check which year micron chips were produced, just to make sure if this is GPU is worth trying to save.
@@northwestrepair But that's why I'm asking... I can't find any info about this factory defect on 2018 micron chips. What's the source of this? Only info I'm seeing is that memory chips aren't at fault, and all of them fail equally.
It's wrong method to use the results of first test in this linux build - after test starts you have some artifacts in bottom of screen because linux dhcpd service is trying to resolve network addresses and its logging data to system console which is used to mats test and mats reading it as memory failure which is not true. On rtx 2000 series mats linux build (400, 450, 455) tou need to run mats again after first automated one.
Hi there, great videos, keep on going, and i whant to ask you how do you test memories when they are off board? Ohms or diode mode but wich pins will tell that is bad memory, thanks...
Very informative video. I have a faulty 2080Ti (error 43) with a KO memory module detected by the "mats" program. I did a 6 min reflow with a heat gun on the memory but it's worse, the card no longer sends a video signal, the fans not working but it seems Windows starts up and I hear the drivers updating. Did the reflow cause a short circuit or break the RAM or something else ? I had put aluminum foil around the memory. My measurements : PLL connection (0,5Ω), PCI buffer(8Ω ), motherboard connection 12V (76Ω) & 3.3V (240Ω) ? Did the reflow cause a short circuit or break the RAM or something else ? Thks a lot
Hi i love the videos i have one question i have a 3060 ti from gigabyte with faulty vram chips. I removed all and i only happen to order one vram chip with is like 30 dollars. So my question is can install just one still just to see if the vram is good and them order the rest and put all. Because i reballed the first set and the card will boot but with terrible image. So can i just use one still and test??
Were all the vram chips faulty? If you know which ones show as faulty after reballing, replace that chip. Even if there is still a faulty chip, you should see an improvement. That is, unless the fault is a shorted chip.
@@AnnaVannieuwenhuyse thing is the gigabyte 3060ti's come with faulty sk hynix chips. It was a big problem at the time. The gpu it self is healthy and ok. But the vram chips are the problem. I've seen so many people replacing there vram chips due to gigbyte using bad batches of chips.
@@trr4gfreddrtgf you are correct but it was not shorted because of reball the chip itself died. I believe is because it was already giving issues and the heat when soldering it back killed it. I've this for my rx 480 and is fine so i believe the chips are just not good
Noob question why does the memory being Micron from 2018 mean we have to replace all the chips? Is it soem imcompability betweene 2019 and 2018 modules?
How do you get these diagrams? I'm new to this channel, I will spend some time looking through what you have available this weekend. I watched Rossman for years but I will never get into a situation to mess with apple products. This is way more along the lines of somethin I would love doing.
In your honest view, does amd apu's have a better lifespan potential if they mostly use system Ram u can replace VS gpu's with ram soldered on the board like theses ? and probably much better repairability too ?
If you need a repair, please find contact info in the description.
When I saw the custom thermal pads I immediately thought it was user error. Lo and behold a plastic film left on the thermal pad. Wow!
I recently handled one of those pads for an SBC. The thin film is completely invisible to the naked eye, the only reason I didn't fall into this trap was that I remembered to check right before mounting the cooler.
You're the kind of guy that sits in the corner and your sisters honeymoon and when she opens her legs you quickly point out that she's not a virgin.
@@1415gatewayable What a crappy comment.
@@conyo985 what did he say
@@johnjonson6377 Well Chris wrote a sarcastic comment about me being a person stating the obvious, but he said it in a childish way.
I find these GPU repairs pretty fascinating! I learn something new from you with each new video. Keep up the GREAT work!!
yea, but also fascinating that people forget to remove that plastics on the heat spreader and destroys all the memory chips.
Former NT driver dev here: the kernel bugcheck (0x116) VIDEO_TDR_FAILURE seen at 6:13 is not actually that uncommon. It means that the WDDM timeout detection & recovery (TDR) watchdog failed to reset the driver and GPU state. This is usually indicative of either failing hardware or a seriously bugged driver that got the GPU into an unrecoverable state (the latter being pretty rare nowadays).
so its partially your fault lol
As interesting and informative this information may be, it is absolutely useless for my field of expertise.
Very sorry.
I sort of had 2 of them in the last week usually when coming out of a game. Updated the drivers let's hope it's gone.
I experienced this using a pcie 3.0 riser with a pcie 4.0 gpu. Setting pcie to 3.0 on the bios and obtaining a 4.0 riser are both solutions.
why is it not uncommon?
I love seeing fully comprehensive repairs like this with somebody that knows what they're doing
I didn't even know that this kind of repair was possible. Truly amazing. Congrats on the excellent work!
Thank you very much!
same
That film left on the pads is why some of them, e.g. Gelid, have something printed on the film. Gelids only have them printed on one of films so peel the blank side off first and stick them on, that way you can see which ones still have the other film on before you assemble the card.
The resin bonding for the VRAM chips nearest the PCIe slot will be to stop them peeling off as the PCB flexes. Loads of prebuilt PCs have had that problem due to the G forces the cards endure as they're being couriyeeted.
That black residue around the memory chips is called underfill. Yes, they use it to strengthen the structure of BGA components near PCI-E socket, where bends can occur. Some reworkable resins can be removed with heat, whereas others, like the one shown in this video, are impossible to remove without brute force.
@@avetruetocaesar3463 Something we learned from the Xbox 360 ring of death era.
Couriyeeted! Nice one!
I'm a regular viewer of NorthRidgeFix and somehow I stumbled upon your channel. Glad i can find another tech who are so talented like you! I absolutely will looking forward to your other repair videos!
Northridge was supposed to send me a bunch of cards for repair.
Once we agreed, I have not heard back.
Not sure what his deal is but from what I understand about his operation, 90% of all GPUs on his shelf he gonna call a no fix while in my case, 90% is a fix.
Just takes few days and I charge a flat fee.
@@northwestrepair yeah, Alex is a very busy person and with that much of work that has to be done, he rather choose to deem "no fix" than to actually fix it (only for devices that takes too long time or truly unrepairable) because it's not worth his time to spend. I bet he work on tens or more devices on daily basis.
Awesome job man! Looks like you know what you are doing....respect!
Yo northwestrepair, I was actually the one who bought the RTX 2080 ti off you on ebay, (I was the guy who ordered on Nov 22nd) as I saw at the end of ur video looked like the video in the description of the RTX 2080 ti I bought, and also had no backplate as well, glad to see the RTX 2080 ti I have won't fail on me 😅, still many thanks tho, and glad I stumbled upon this video as well
this is going to be a nice day because my day is start by northwestrepair awesome video
Morning!
Another key note, that is a Micron variant. The Micron 2080 Ti's were bombs in general. I bought a XC Black, the first two I received were Micron, and artifacting out of the box. The third came with Samsung memory, and have had zero issues since (2020).
This is the reason I will not touch the 4090 but the new 5000 series will hopefully use Hynix /Samsung
How do you know if it's a micron variant?
@@KMRscoo GPU-Z
Amazing, those are not easy repairs. Very few people have the technical knowledge and soldering skills to complete something like this.
This was a fantastic video, mesmerising to watch you work. It was also very cathartic to see the Uningine Valley pass at the end, my go-to when celebrating the birth of a new, hopefully-stable rig.
I am so glad I got this channel recommended.
Great stuff !
This level of repair is hardly seen these days as people either don't have the knowhow or are too impatient to get tech repaired. It takes years of skill, knowledge, failures, screwups and experience to point the exact problems get things working again after doing all the necessary changes and repairs/replacement.
Hats off and 🙏 to you buddy all the way from 🇮🇳.
Expecting the average consumer to be able to solder at this level of skill just isn't going to happen. We should repair more and I always open broken tech up to give it a shot, but such small surface mount components require incredible skill and precision - not to mention all of the equipment needed, from a quality heat gun, multimeter with needle probes, a good soldering iron (usually with temperature control) and a thermal camera - plus buying the memory chips, which are often so expensive in anything other than bulk quantities that would leave consumers with a mountain of useless extra parts that consumers have no incentive to do it themselves.
Repair needs to be made easier and cheaper through right to repair initiatives, not complaining that the average person somehow hasn't invested hundreds of hours practicing SMD soldering, or into nice repair equipment and doesn't have several donor boards lying around.
I've never pressed on a youtube notification so fast. Really nice videos ! Keep it up man !
Same, I always watch his videos, It's like Therapy or ASMR to me.
i keep waiting for more videos from you. like the way you explain. learning a lot from you. Thank you.
FYI, this residue is called "underfill" and it's purpose is to avoid/lower risk of cracks in solder joint's if it's getting under physical "stress". It's usually applied in devices which can exhibit that (like phones and other handhelded devices). Apparently wohever designed that, came to idea that memory chips near PCI-e slot are also at risk of that so that's why it's there.
its already know that early micron memory on 2080 ti failed a lot. tons of news stories about it when the 2080 ti launched. gave the card space invaders artifacting.
There are times when I'm envious of someone's knowledge and expertise and this is one of those times. In saying that - well done for the fix and i hope you see some reward for your efforts. I have a 1080ti which needs repairing, i genuinely believe t it can be fixed - I just don't have the knowledge to even start diagnosing where the problem stems from. The card is still operational but under stress it crashes.
I have almost the same issue but with an R9 280X that for some reason, I was playing call of duty warzone before the cold war forced integration and suddenly image frozen but the system didn't recover and while booting, the gpu never worked again and I had to remove it because the system was getting frozen before the POST so I think this gpu died or something horribly wrong happened and it needs dedicated repair, that would cost me way more than what I spent to buy it and some people either don't want to repair a gpu or they don't know or can't be bothered to order the parts necessary for the repair to be done.
Same issue with my gtx 770, so I bought a second hand rx 5500xt. For 96 pounds or about 115 dollars.
try to underclock often times it just cant hold certain clocks anymore or try adding voltage at lower clock frequencies in frequency graphtuner from msi afterburner.
Very entertaining video, keep up the great work and thank you for sharing the knowledge!
woah! I didn't know these kinds of repairs were possible 🤯 Next level stuff
wow...I am speechless at your genius level skills
INCREDIBLE! loved it, i don't know any person who do that in my country, you have my respect 🙏
That 2080TI ,they had a waterblock on there and put the stock cooler back on later. That's why the screws are mismatched and why the thermal pads are all freaky looking. The stock thermal pads look very different.
6:20 how does one go about creating a linux usb with vram tests? I can't find anything on google about it
i love to watch your videos as its informative on about the issues that cards can have over time. great job on the fixing of the rtx 2080 ti!
7:00 i'm curious as to how plastic film can lead to a capacitor blowing?
Thats Work of A LEGEND! wish theres like you on my place :D
This is like the Tech equivalent of those AMSR videos,🤫its just so enjoyable and relaxing to watch 😀
It's interesting to watch your repairs. Wish you good luck. Some gpus repairs maybe won't pay back, but maybe some new experience you get push you bit further avoid any more same symptoms on bad/dead stuff.
It reminds me my old GF 8800GTX died after 2y heavy gaming long 10 years ago. I put it to service, was something like heat treated and UV too(Not rly sure), it worked again for another 6mo until it died again. That previous 20-25USD(500CZK) repair cost I didn't wish to repeat. That was time to get new gpu.
You're amazing, hope you get your money back for that hard work :)
This tells me so much,first how good you are at diagnostics and secondly,the reason for resin around the memory, is because they run this memory so hard it gets hot enough to soften the BGA solder,the resin stops the memory moving around, that explains why so many fail just after warranty, i used to buy high end cards,now i dont,just money down the drain.I remember the early days,when graphics cards usually outlasted the PC,the only reason to replace a graphics card back then,was to upgrade.Seems like everything now is made by, make fast buck companies.I liken it to buying a ferrari,it's great at waht it does,but you pay the price for this performance & its engine isnt going to see many 10000 of miles,before a rebuild.Great vid,thank you.
That's a BGA underfill and is not there because the solder is melting. The chips would stop working long before it ran hot enough to do that and holding it in place would only do so much anyway.
Underfill is meant to relieve stress on the BGA solder balls. Usually to minimize the effects of thermal expansion over time but Nvidia doesn't really care about that. It's only on the chips nearest the PCIE slot because they're the most prone to damage from board flex. I suspect it's also there as added protection for OEMs and prebuilts shipping these pre-installed in systems.
It's also likely part of the reason Nvidia has moved as many chips away from the slot as possible on the heavy 30 & 40 series cards.
@@JJFX- ok thats fair enough,i was thinking low temp solder 188 deg c,as i would imagine it would allow for a very small amount of thermal movement,under overclocking for instance,ive also had laptop system boards with resin around memory BGA's ,i wonder if it's the same reason. Do you think high end graphics cards are always going to have a high failure rate because of thermal stress,or is it something else? As for the small SMT tant caps,ive repalced a fair few of those, bit like the larger ones,they go short circuit as well. As the old saying goes,the cooler you keep components ,the longer they live.
@@enoz.j3506 Well this is almost certainly all lead-free solder which really melts well above 200C. Issues with the GPU itself tend to be on the large dies as thermal cycling has a more significant impact on the substrate joints over time. Different parts of the die heat up at different rates so the effect of thermal expansion become more significant. How the die is bonded to the substrate, the type of underfill combined with acceptable production tolerances have been difficult to get right while staying economical. This a big reason for the infamous Xbox 360 RLOD early on.
Hard to say how modern chips will hold up. Package engineering is very complicated and Nvidia has made plenty of mistakes in the past. However advancements in underfill and BGA assembly have come a long way since their more notorious issues. That said, they are dancing with the devil by using bigger and bigger dies pulling more and more power.
I definitely agree that those looking for longevity should keep everything as cool as possible. IMO, putting a full size water block on $1000+ GPUs capable of well over 350W is a no-brainer.
yes the thermal expansion in all dies is very difficult to control,if you think of the 28 billion transistors+ ,it is amazing it lasts as long as it does, yes i remember the xbox problems.
@@enoz.j3506 Yes it's pretty mind blowing. As processes get smaller and smaller it'll be interesting to see how the transistors themselves hold up. Interestingly, Nvidia's vcore target has largely stayed the same since Pascal despite going from TSMC 16nm to 5nm now.
That silicon around the bottom memory chips is to protect them from coming loose/solder joints cracking from the sheer weight of the cooler that the GPU manufacturers started implementing quietly a couple of months into production.
Do you know if other GTX / RTX or even RX cards tend to have that same type of Micron vram chips?
Asking so I know what to look out for when purchasing used GPUs.
Browse through his older videos. He already did some repairs and mentioned it there. IIRC it´s related to 2018 manufactured Microns.
You will not find it out, tho. At least not without full GPU disassembly and thorough cleaning of the chips. GPU-Z does not offer this kind of info.
2000 series cards do.
You cant look without opening it up.
If its micron and starts with 8, get something else of problem will come sooner or later.
can you explain more to depth about why 8 ohm on that coil means memory modules dead ? is it too little resistance ? too much ? there isn't supposed to be resistance maybe without coil ?
its all a big mystery for someone like me with basic understanding.
nice. unlike the other youtuber, you talk soft, no rude and work clean.
i dunno why i like computer repair videos so much
Would it be possible for you the mention the price of the memory chips during the videos? Always interesting to know!
I think he just salvages them from other boards so it's not like they have a specific price
I don't think you'll get those GDDR6 Micron chips soo easily in the market. Nvidia and AMD makes specific orders which are then delivered by Micron.
You done very good job. The question is how long it will last until break up again?
Dumb question:
Assuming that the warranty is not voided with this kind of teardown, does the unpeeled plastic cover on the thermal pad at @6:53 constitute "faulty workmanship"?
It's likely voided.
As someone who got a comp sci degree, therefore only knows programming, not hardware, I like your funny words magic man!
Note: I know basic hardware not GPU and such
I only know hardware, the programming is a foreign language to me. Funny seeing the opposite end of the spectrum. Lol
@@OCtheG same
your CS degree DIDN'T teach you basic hardware? what sort of CS degree is that? I had multiple modules dedicated to hardware. Of course the electronic engineers have more in depth knowledge, but we were expected to be able to design (i.e traces, component positions), fabricate (milling etc...) and assemble (surface mount inductors, caps, resistors etc...) a PCB and accompanying hardware, and program the MC on said PCB to control any accompanying hardware.
@@lewisd56 I am currently doing mathematics and computer science (UK joint degree) and I never got any hardware course, even compilers were exclusively taught to computer science people.
I've mostly just had theoretical stuff as compulsory courses and the rest to choose myself. And among the options we get, there's basically nothing hardware focused.
Still, I have been studying more on the side because I'm afraid of coming out useless from university even if I have pretty good grades.
@@leonro
UK computer science degrees too, (BSc and MSc), I had dedicated modules for embedded systems (C++ programming arduinos and other MCUs as well as designing accompanying circuitry), even had a module that went into specifics of logic gates and operations in a CPU. The MSc took a step back from the hardware, but I specifically picked the hardware oriented modules, so ended up working with more complex embedded systems and designing a whole PCB from scratch.
Very nice video, keep a good work :)
4:58 every PBS show at the begining be like
Crazy it got resurrected from the grave. Alot of knowledge and skill here. Remember there was a big thing with this card about the memory having alot of problem. RTX 2080 Ti owner here and I haven't had a single issue thus far and I push the GPU to the absolute limit. Still, can't wait to eventually grab an RTX 4090 this upgrade
RTX 4090?
Are you puffing the magic smoke dude?
Sure. Struggle reading? Different type of memory were outfitted on the 2080 Ti depending on the purchase people could end up using a defunct version destined to be RMA or have severe longevity issue. Luckily I didn't get this particular chipset Micron or Hynix I got Samsung, but this could be an issue going forward. Not that would I expect you to comprehend this, drug addict.
Just get a 3070/3080
Not likely, the 4090 is nearly 2* stronger
@@lfsky strong enough to burn down your house
This dude is an absolute Guru, why is your channel only 16k subs wtf lmao.....
Please start speaking about GPUS, PCB quality and such.....
You are some kind of wizard sir.
I'm surprised how there are no class action cases for failing of 8 leading GDDR6 micron chip.
They lasted the warranty Period.
Wouldn't be surprised if Nvidia knew that and used them anyway so they can sell more Cards, given their usual scumbag behavior.
EVGA pulled out for a reason.
Good job team.
Thank you
I added a shit ton of additional pads to my 2080 when i got it after replacing the old ones and it still gets crazy hot on the back plane. Probably going to have to add heat sinks to the back shield with thermal adhesive.
wrong thickness=temps
hopefully 3000 series will not suffer a similar fate 💀
it does not
@@northwestrepair I knew the 20 series was one I was gonna pass on
@@nicekeyboardalan6972 i think 3000 series will face different set of issues, mostly due to all the shortages, that were happening during their manufacturing. I expect something similar with AMD cards too.
Many companies had issues during Covid shortages and were using literally anything to make their products.
@@Morpheus-pt3wq well if you got a gigabyte card or a zotac I would worry
@@nicekeyboardalan6972 i have 2070 super from zotac and it does worry me if my gpu memory will fail. memory manufacturer is micron.
bro love your vids by the way
Thats why I still used the good ol 1080ti thing is a tank.
I've had two cards die since I bought mine in 2020.
First two cards were PALIT 2080ti Gaming Pros, both died due to issues with the GDDR6 memory - pretty serious problem from what I've heard from other 2080ti owners. Sucks even more that the RMA process takes around a month due to PALIT being in china.
Thankfully, they ran out of PALIT 2080ti Gaming Pros to send me last year due to the 3000 series coming out. 😂
They sent me a PALIT 2080ti Gaming Pro OC which is a card that costs £600 more than the £999 card I bought due to much higher quality components, thermal cooling additions and a much bulkier design for heat dicipitation. Haven't had the same problems with this card as the cheaper ones (touch wood), so I'm assuming it's due to thermal performance or the issue was fixed post-june 2021 (first two cards died within 4-6 months).
with the resin on the bottom memory ic's... I think evga might have determined that the pcb flex near the pcie connector plus thermal cycles leads to those sm components breaking contact.
Dude Ur a genius
Bought a used Strix 2080Ti for $274 (incl. an installed Strix waterblock + the original OEM shroud/fan/heatsink assembly w/backplate). As I do not do water cooling, I removed the watercooler for resale and installed the air cooling assembly/backplate (after a deep clean, thermal-pad check, and re-paste). Unfortunately, the white shroud for the male FAN power connector to the GPU board was damaged (as-is) and although the pins make good contact, the shroud does not securely maintain the connection. Tie-wrap to the rescue. Done.
Now the card runs perfectly well. Benchmarked SAT. I then did an under-volt using MSI Afterburner...for safe measure. GPU runs cool and stable...very happy with a fine 2080Ti that cost me, NET, about $220.
Remember Micron from 2019? It has big troubles for 2080's as Hynix GDDR6 1gb modules for 3060ti. Yet appearing after years of service. Noone is safe from such things to happen, sadly.
It also depends on cooling too :p
It’s why I soak the plastic and let it rehydrate, then ram work better then new activating all the cells whenever and not just that teturns
So my 2080 ti its getting blue screen when i play some games or driver try to install, idk what it is, i need to investigate more.
What an amazing skill and experience!! You make it sound that it is, piece of cake!!
How often do you see instances of the plastic covering on thermal pads not being removed?
4:47 whats the name of this software an where i can download the shematics ? greetings
Could you please make a video explaining mods you make to your own cards to make them more reliable and less likely to experience part failure from overheating? What assortment kits you recommend, assuming it's shims or pads, etc.? Thanks!
I don't do mods.
Don't overclock.
Keep it cool.
Fresh paste and pads. Nothing else.
@@northwestrepair You may disagree but i also undervolted my GPU as much as i can before i start to sacrifice performance and stability,
@@northwestrepair my HD7850 worked for 12 years non stop for me. Gaming 9 hours or using 3d max 9 hourse a day, it didnt matter.
This gpu still can work after all this time when it was maxed out everyday for 12 years.
I got a 3080 but that amd card can still work
I had a golden rule. Which were never broken when i had this gpu.
NEVER EVER overclock
NEVER.
Over clock is most stupid thing people are doing with their gpu. It will sap away life span heavily.
At very least this was my experience.
My last pc was 20 years ago as well.
It worked for 10 years.
I never overclock that either.
Normally i build new pc for every 10 years, hope new 3080 work fine for next 10 years.
Its undervolted and kept cool while i am working with pc
Overclocking isn't dangerous as long as you don't overvolt too much and keep the temperature in check. My old ATi 4870 1GB is still working fine, even tho it has been heavily overclocked since the day that I bought it in 2008.
@@waifuhunter9709 "Over clock is most stupid thing people are doing with their gpu. It will sap away life span heavily."
yeah you're actually retarded if you think people use garbage GPU's for more than 2-3 years bro these days, even back when overvolting was a thing it wasn't relevant
Either way, just overclock anyways, why? Well... Overclock doesn't mean OVERVOLT, you can't even do that anymore on 99% of gpu's
stop being stupid just because you fucked up
@northwestrepair could you explain to me how to find the gpu schematics showing the layout and values of each component like the one shown in this video?
Also how do you know exactly where to touch your voltage tester probes when looking for a short in memory, resistors, vrms ect..?
Would you be willing to do a video explaining how others like myself can get a beginners guide to checking for damage/ short circuits using a voltage tester to check for ourselves?
Any tips or advice would be greatly appreciated. I genuinely enjoy watching all your videos and each one inspires me to do some troubleshooting myself (I have a decent but still beginners level of experience with electronics) but i am to nervous to attempt anything more than a thermal pad/ repaste on any of my GPUs.
For mw B0 had write failures and sometimes it would be okay so i bought a rework station and reflowed it and its working for now.
When watching any of his video's you can see the results of a true technician not somebody that thinks they know what they are doing!
He has repaired a card for me with 100% excellent results, so if you have a card and were worried about his honesty or work ethics fear not he is the best !
Makes me wonder how my 1070 is still chugging along with almost no issues. A motherboard swap when I upgraded from my r7 1700 to an r5 5600x got rid of intermittent snowflakes... Why? No idea.
Yeah... this video explains why there are now so many 2080ti's on the market (for cheap... like $200 cheap in case there are those not in the know.)
Thank you soo very kindly for showing the music. Asher is amazing and its my new fave song!
where did you get the gpu schematic?
is it me or are graphics card failures becoming more frequent in recent years ?
I've had the same 1060 for 6 years no issues besides when I replaced the thermal paste the original was like cement on the gpu.
Repairmire:Diggie digi
good work partner
if i were near you would buy this art sadly we are far from each others
always watching your videos for learning something new keep it up my friend
Very cool video. If I had the time, I'd like to experiment with component level repair, but I lack the necessary tools and shouldering abilities.
search a local shop in your area
If I'm not mistaken, the 2080Ti had memory failure issues at launch as well 🤔
This is why ebga has stopped producing GPU. I have a 1 year old 3060 ti sc gpu that I suspect memory has died. They sent me a replacement and not even a week after I have artifacts and cannot be reinstalled anymore. Pretty sad.
Nice channel , fixing like real pro =) , what temp you use in pcb heater , and hot air gun when solder memory? ty your channel is awesome
400
@@northwestrepair Preheater In gpu bottom 400c too? Ur channel is awedome
How to fix ripped pads`?
I have an old Radeon 7xxx that I would donate. The fans turn on but it seems no signal over hdmi. I can't test the dp or dvi outputs.
Could you show more tools/materials/techniques of the trade? It would be super useful for my personal projects, even microcontroller projects.
Seems like using some kapton tape on those small components would make life a bit easier in that they wouldn't blow away when you apply heat to the memory chips.
The glue on the memory chips near the PCI connector is there to keep them from detaching in case of excessive board flex?
That is for reflow soldering
where i find the program who test the gpu? thx
what program specifically ?
What would do that to that first card? That seems like too much damage to be a single bad component.
I suspected it was failing memory on my Sapphire 280x it would have a square on squeen randomly like swap places with another square and parts of the textures would get mixed up/missing, and would commonly get horrible tearing issues with meshes. Also motherboard handshaking issues. It cleared up a bit after a oven bake (not proud of lol but it was that or it was dead) interesting to see and not as scary as I thought to replace memory now that I've gotten alot more experience soldering just haven't worked much with BGA
Hmmm, why are you so certain that 2019 micron chips will not die as well? Wasn't it confirmed that even Samsung memory version has a similar failure rate? Were there some recent discoveries about this that I've missed?
I have a failing 2080 Ti Founders Edition, it's still 'working', but launching any game will quickly cause artifacts and a crash to desktop. Underclocking memory by -300 MHz fixed the problem temporarily, but that didn't last. Surprisingly, it passes MATS test. Now I want to open it up and check which year micron chips were produced, just to make sure if this is GPU is worth trying to save.
i am not certain. they simply dont have a factory defect.
@@northwestrepair But that's why I'm asking... I can't find any info about this factory defect on 2018 micron chips. What's the source of this? Only info I'm seeing is that memory chips aren't at fault, and all of them fail equally.
Do you think that this same issue could be on rtx 2000 series cards?
This is 2000 rtx card.
It's wrong method to use the results of first test in this linux build - after test starts you have some artifacts in bottom of screen because linux dhcpd service is trying to resolve network addresses and its logging data to system console which is used to mats test and mats reading it as memory failure which is not true. On rtx 2000 series mats linux build (400, 450, 455) tou need to run mats again after first automated one.
Why did they use resin/glue only for those chips?
Hi there, great videos, keep on going, and i whant to ask you how do you test memories when they are off board? Ohms or diode mode but wich pins will tell that is bad memory, thanks...
Very informative video. I have a faulty 2080Ti (error 43) with a KO memory module detected by the "mats" program. I did a 6 min reflow with a heat gun on the memory but it's worse, the card no longer sends a video signal, the fans not working but it seems Windows starts up and I hear the drivers updating. Did the reflow cause a short circuit or break the RAM or something else ? I had put aluminum foil around the memory. My measurements : PLL connection (0,5Ω), PCI buffer(8Ω ), motherboard connection 12V (76Ω) & 3.3V (240Ω) ? Did the reflow cause a short circuit or break the RAM or something else ? Thks a lot
How much does it cost to replace all those chips?
depends how many
Very interesting and satisfying, how on Earth do you learn to solder like that, I can’t even solder 2 wires without problems 😅
Very informative video.
Hi i love the videos i have one question i have a 3060 ti from gigabyte with faulty vram chips. I removed all and i only happen to order one vram chip with is like 30 dollars. So my question is can install just one still just to see if the vram is good and them order the rest and put all. Because i reballed the first set and the card will boot but with terrible image. So can i just use one still and test??
Were all the vram chips faulty?
If you know which ones show as faulty after reballing, replace that chip. Even if there is still a faulty chip, you should see an improvement. That is, unless the fault is a shorted chip.
@@AnnaVannieuwenhuyse thing is the gigabyte 3060ti's come with faulty sk hynix chips. It was a big problem at the time. The gpu it self is healthy and ok. But the vram chips are the problem. I've seen so many people replacing there vram chips due to gigbyte using bad batches of chips.
@@AnnaVannieuwenhuyse and yes after reballing two got shorted and had to get rid of them
@@zane383 I don't know much about this stuff, but if they got shorted when reballed, couldn't you just reball them again until there's no short?
@@trr4gfreddrtgf you are correct but it was not shorted because of reball the chip itself died. I believe is because it was already giving issues and the heat when soldering it back killed it. I've this for my rx 480 and is fine so i believe the chips are just not good
Noob question why does the memory being Micron from 2018 mean we have to replace all the chips? Is it soem imcompability betweene 2019 and 2018 modules?
Because replacing one can cause adjacent chip to artifact.
Or it will artifact very soon after.
How do you get these diagrams? I'm new to this channel, I will spend some time looking through what you have available this weekend. I watched Rossman for years but I will never get into a situation to mess with apple products. This is way more along the lines of somethin I would love doing.
In your honest view, does amd apu's have a better lifespan potential if they mostly use system Ram u can replace VS gpu's with ram soldered on the board like theses ? and probably much better repairability too ?