Nice video, thank you very much. May be as an idea, I saw some similar videos from Russian guys and they all told, that if on of the MOSFETs blew up, it is usually not sufficient just to replace it. In the best case it will work for another week or so and, in the worst, like in yours, it will blow up immediately again. They say, that if you have to replace a MOSFET, you always also have to replace the controller, because it usually is the reason, why the MOSFETs blew up in the first place anyway.
I think you're right. This video ended up being super popular, so I might take another swing at this, just to see if we can finish the job. It'll be a while though!
@@Adamant_IT Yes, i also believe the controller has turned on both mosfets (high and low ) at the same time, creating a path to ground through the low side mosfet probably, blowing the high side one. Replacing the controller too, might solve the issue
@@yosuhara No - its the mosfet driver chip Adam mentions: NCP 81062 from ON electronics. They are available from Aliexpress.com Links to the ON datasheets site: www.onsemi.com/ www.onsemi.com/pub/Collateral/NCP81062-D.PDF www.onsemi.com/pub/Collateral/NCP81174-D.PDF Apparently the mosfet driver chip fails by turning on the lo-side mosfet simultaneously with the hiside mosfet, thus taking out the weaker of the 2.
I love the cheap compact testers/o-scopes. However in this case I'd love to see a decent analog scope, at least in an educational video. Just constructive criticism. Nicely done though.
True that, he's getting confused a lot just because the scope is limited and doesn't show properly what's going on. I think he should see full 12V on the mosfet output, mosfet is either off - 0V or on - 12V, it should not show anything in between.
@@Arek_R. That's not actually true, with such high switching frequency you can't constantly hold 12V or 0V at the output, I will always be something between. And for such purposes there will be no difference between cheap and professional o-scope, only slighly change in readings accuracy. Personally I use ad1102ca from Atten which isn't either cheap or expensive, but didn't notice difference when compared to cheap ones or some high end for thousands of $ ;)
"the short magically disappeard" :D There's no magic, the problem lays in the method you use to solder the mosfets (or actually any QFN-type components with thermal pads), placing to much solder on pads and then trying to push it out is a bad idea, 'cause you push it in all 4 directions, 3 visible and also the one that goes to G and S pads of the mosfets which you can't see but for sure will make a short circuit ;) So in my opinion it's better to clean the pads from old solder and use solder paste (even some cheap chineese one), SnPb is preffered (by me) 'cause of lower melting temperature :) @1:09:40 you start thinking why it shorted again and burn upper mosfet while it has no connection to GND, it's all caused by failed PWM controller and/or driver which drives low side mosfets constantly so at output you have GND all the time and when swiching upper one it simply burns out ( double lower side mosfets can handle that short, single upper one will blow up)
i like your videos, but... seriously a $20 "scope" with 1Msps and 200KHz badwidth for probing a buck converter most probably running at 500KHz or more ? it's more than clear that what you see is an aliased image - that's why it's so "runny" with jumping amplitude. the only thing you can see is that "there is something" but not the actual wave form at those frequencies ;) that scope thingy is good for audio or a SMPS running up to 100KHz. you literally would be not able to see any issue with the mosfet drivers based on what that scope displays at those frequencies... apart from "there is something" vs. "there is nothing" ;)
Yea the scope experiment was a definite fail. I usually use this thing for looking at simple wave forms like PWM signals, but it was way out of it's depth for this. I do plan to get a proper scope for the future though, so next time I'm better equipped.
Please don't attempt to explain things with a scope that simply cannot display anything remotely resembling the truth. The switcher runs at a few hundred kHz (datasheet says so) which is beyond that thing's ability and we see aliased nonsense. My crappy Owon from Rapid can do better. Other than that thank you for showing a repair disaster
In my experience (not specifically related to graphics cards), if your MOSFETs short, you should assume the drivers have been damaged and replace them as well.
Thank You. And yes, learned alot (and I was sure I knew everything🙄). I enjoyed the magnum length video version. I ordered a pizza and told my GF to get lost. OK, I didn't do that.. I just let the air out of her 🤤 Anyhoot, you fought the good fight. Think I will pickup one of these classics. It will feel like an old trusted friend. Cheers 🇺🇸😁🇬🇧
Interesting look into buck regulators. The only thing you didn't try was swapping out the driver i.c. I think that was faulty (thermally unstable) and provided the path to ground for the high side f.e.t. The driver i.c. probably switched the f.e.t. on for too long, killing it.
Question is.. have you tested the coils which the mosfets are working together with? If they all have the same characteristics to different frequencies, or if other components have measured values slightly off specs. It's a longshot, but it might be the cause of why it's the first mosfet that shorts, if the harmonics might be off tune causing standing waves, or if some other component interferes. I was thinking, maybe you could trick it into running 2 phases instead of one at lower and moderate load to smoothen the transition to 3 (or 4) phases? I really enjoy your videos and your waffling :)
id love to see you fix some rx 570s an 580s i have a bunch i could send u man, i need fixed, but to see u do it would help me fix mine i love how u go over everything an explain it, SO AWESOME MAN i love you!!
your pwm is faulty, or its pvin filtering capacitor is out of nominal. Or you have faulty low side fet. Or your fet doesnt match parameters. Or you have bad soldering and thus floating gate of some fets. FET gets shorted if you get direct conductivity from high side +V to GND. (when both fets appear to be open)
YOU ARE REALLY AMAZING. MY REGRET IS WHY I DIDNT FIND YOUR CHANNEL EARLIER! A GOOD CRAFTSMAN ALWAYS SHOW HIS MISTAKES SO WE ALL DONT MAKE THEM! YOU ARE EXCELLENT EXAMPLE! REALLY LIKE YOUR TROUBLESHOOTING CYCLES!
These High and Low sides are both N-channel. The difference between them is that the Low sides are more expensive and rated for higher current flow, so I opted to use those for the replacements. I've seen many VRMs using the same fets for Hi and Low, so I assume that it's okay to change them all out here. I don't believe that was the cause of Phase 1 melting again, as Phase 2 also got the replacement High side, and is still going strong.
@@Adamant_IT i understand wht u said and ref schematics also showed N-Fets so yes it may be right, but i fail to still understand how that works if you are switching on the positive side you need P-Fets no? which is what reference schematic also shows its switching directly 12v...maybe a complete design will help
@@Adamant_IT Again i'm confused as conventionally you have P channel switching Hi side and N channel switching low side, can you point me to a schematic that uses same FETS for hi and low cos i'm missing something here...cheers
That was really entertaining. Shame it didn't turn out as you hoped but hey, after all the replacement surgeries and explosions you still got it to work. That's a big win in my books. Keep it up! :D
if you ever need to figure out what a power rail is, i bet Buildzoid/AHOC can tell you! (i assume you don't have the time to watch a ton of his videos in order to stumble across one where he explains it) re: 36:21
I love when people make mistakes on something, notice it, then point it out in the edit... Leaves you sat there wishing you could yell *THE FET'S BACKWAAAAAARDS!!! IT'S BACKWARDS!!! DUDE!! STOP! IT'S BACKWARDS!!* even though you know they can't hear ya lmao
Hi i got a GTX 960 Asus XTRIX 2gb giving this problem , Any type or any kind of SHADOW looks GREEN OLIVE Now, i mean deep olive :s , have you seen this problem before ? , any advice ? i already tried everything changing (cables , drivers , psu , monitor, thermal paste and pads ,installed on another pc, even reballing at 200 degrees for 10 minutes ) all check , and STILL Super Green ! olive shadows, Help please
Sir, it's 3:48am in the morning... WTHECK just happened.. all I heard was turbo dead.. Mosfett just dropped dead!! 😂R.I.Pieces Mr. Mosfet. Sorry sir, I was quite happy for your accomplishment. Use the donor card.. what else can work as an inductor replacement? Dr. Frankenstein we need an Abby Normal Brain.. A better Mosfet that can handle a fuller load. This way the resistors can still take care of some of the flow.
ALERT I WNAT TO HELP YOU FIRST when you inject power into your mobo dont say tracks can handle 19V they are rated for mby 8Amps and if you add ist R its mby 0.2 to 10.8V max SO dont say they can handle 19V say they can handle ~6A (good luck puting more than 5V on it whitout current protection) SECOND your soldering iron tip is fucked you should use temperature under 400C or under 750F (mby you turn it up when you solder on ground or bigger pads) and your tip should be tinned (shiney looking) your tip get more gray if you overheat it so its noth worth it
Found the answer for why high side MOSFET continuously failed. It's a broken MOSFET driver and not the MOSFET itself. This specific MOSFET is designed to handle upto 30V but Vgs is giving more than 30V for passing current to rest of the circuit. With a standard text book example, if Rds is 10 Ohm and Vt 3V Vgs should be more than 14.9V for it to turn on. Infact it is turning on and that's the reason why C4 did not fail. Low Side could have easily handled the voltage and required a higher of 3V to switch on and dump to ground. There is R2 and C2 which would have handled the load but there seems to be a short internally for the MOSFET driver and SW pin might have shorted and there is high demand which blew the filters. Source for my claims: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-g8I_upSEGz0.html
Forget your Kyles, Jayzs and Linusus.... This real IT problem solving! I wouldnt attempt it myself though... My clumsy fat fingers wouldnt be able to be this precise
Where did you get a Zotac 960 replacement fan? I have a Zotac GTX 960 SC 2GB. It looks ever so similar to yours (black PCB and not as many grooves in the fascia plate). The fan does not work and I need to source a replacement, but as you mentioned, this is difficult to do.
That blower card probably got cooked. You can't just blow out the dust on those, you have to disassemble them to clean them. A friend of mine blew up two blower cards before we figured that out.
Man, i am not an expert, i do repair things from time to time, but your voice and especially your English accent, really make me watch your clips. I have an issue with a keyboard: a column of keys get lights but don't send a signal, the keyboard is Logitech G pro. I try to bridge but it seems all keys ( F7, 5, T, G, V and Space) are one link on the board. Do you have any idea on how to start this repair, whats spares should I look for. This issue appeared after I try to resolder a switch ( I ve done this before, many times but never saw this type troubleshooting ). Thank you so much and Rock on!
I have a evga gtx 960 4gb...doing the same artifats but inside the game...actually red dead redemption 2..and a few segunds later crashed and go to desktop...this video card pass all the test you know,furmark,superposition,etc....normally...even i can play some old games....the problems arrive when i put modern games....previoslly i check the vram with mats...i find one bad...i do reflow...perfect....and i do a reflow to gpu too....she is working perfect until i put a modern game....i need se help about what else its happens there...please😳
Anyone know what capacitor is this rated at m.imgur.com/Zy343hY Cause some mice just climbed up to my gpu and shit on it and has piss stains so I cleaned it with alcohol knowingly it's corroded with green stuff., suspect that this cap is connected to the 12v rail from the pcie but I don't have currently a tool that checks capacitance of it, also it does boot and has image then after several minutes it would crash/pull off a no signal input. MSI RX570 4GB OC. replies are highly appreciated.
Did you put a fuse on the input supply, instead of the inductors (is it a potential bomb for your buddy)? Something like 10 - 12A perhaps. Enjoyed the video, cheers.
Hello there mate, I've got a gtx 960, my pc posts and apparently boots into windows 10 but there's no display during the whole operation , my cpu doesn't have integrated graphics and to my surprise it did actually work with another gpu. Do you have any advice?
I just works on same reference board with exactly a same issue (except mine go nuclear around inductors so it need some PCB etching and I end up replacing them with fuse as only three spots for inductors survive blast) and yes, I end up replacing control IC because first mosfet blow up few minutes after replacing. Now it work ok. Probably original problem was IC and not mosfet same like in your case.
I also have a GTX960 graphics card. But it only works when importing images. When I want to play games it overheats and shuts down. Thermal paste and pads are clean. I am not an Engineer or a Technician. But I'm still working on the problem. 😁
I like your videos very much. Your speach is 98% understandable for a non native, non supreme speaker :) You can teach me electronics - in English! - That is way more rare than you may think...
I find it funny that right after I left the video and commented you need to read schematics, you start looking at schematics. Albeit not the card manufacturer schematics. I wonder why? Also critical because the phase ground didn't end there, couldn't you have traced it farther? I do understand though you showed how you can figure out without schematics what goes where, which is handy if it is obscure or old, but that card wasn't either. Can't you get small business loans to cover ultrasonic cleaners and Schematics websites? I can't find them for free but idk. Thanks, this was awesome! I learned a ton. Entertaining as well.
I believe you need to connect to the board via an I2C connection but for nvidia its called something else. It creates a link via usb to a pc and a GUI to view the Power Monitoring and or programming. In there you can turn off or on phases. I was thinking you could have used thermal cam to find the culprit on this GPU.
Hello there I had 4 shorted MOSFETs in my GTX 970 so I replaced them but they shorted again. I noticed something very strange is that when I spray a contact cleaner on the corresponding Gate Driver the short is gone, but when I tried to power the card again the short is back. I did that about 3 times and the short goes away every time with the contact cleaner on the Gate Driver. How can that happen?
I haven`t watched the complete video yet. But to me it looks as you have placed the mosfet on (013 or Q13) the opposite way on, to the way it was when taking it off. Can you have a look and verify this for me please. I know you posted this video last year. But i`m still interested to know. Cheers.
Cool Video. Why dont you try your luck and work on the donor card and populat the missing VRM and stuff? would be great to see. At least you know the donor card doesnt have any mysterious fault causing the short in the VRM.
1) At 1:09:55 when you say "the high side mosfet doesn't have a path to ground" that is what is confusing you. When it shorts, it has very low resistance, is not able to turn off and therefore allows the full 12v and massive amounts of current to flow through itself, the inductors and capacitors after it, and this current heads for vcore. This means there is a direct path to vcore and as you already know vcore has very low resistance which appears as a short. You're reading the vcore resistance.
Is anyone able to help me? I've got an issue where my games look a little blurry for some reason almost like the resolution scale isn't the same, but the scaling is completely fine. I'm really not sure what to do anymore, I've got an 1660 ti nvidia graphics card. Help lol
I know it is over 2 years since you posted this video but this is common called capsize from the controller, did you ever succeed to repair it and get it working? I had this fault a few times both in laptops but also in power amps LOL
I know this video is getting on a bit now but I just stumbled across it, and I couldn't help but notice that the electrolytic cap by the auxiliary power connector seems to have the polarity the other way round, just a thought...
love your videos adam but need to say tht the elastic tow rope example is verry wrong.an elastic rope would infact ad super speed to the car being towed and infact cause it to smash into the truck towing it.the toe rope needs to be solid so there is no force in towing.ive pulled hundereds of cars and just sayin.wrong example.that being said i have truely learned alot from you and your videos.bravo and keep it up your a great teacher..Robert Whalen.Ontario Canada
Why didn't you just swap the driver chip for phase 1 from the donor board? Is it possible that chip could have gone defective and was causing the high side mosfet to short?
I watch your videos like i watched The Shawshank Redemption or other great movies :))Imagine in high school i was the electronic class and none of those teacher talked like that and was so boring.Now...i watch with big curiosity, you explanation are so good and i understand no matter if my english is bad. :)) GOOD JOB!
I can't understand why you don't do something about that gpu mess. There is thermal paste coming of three of its edges onto the metal points. And it just looks damn sloppy.
i got a gtx970 that set on fire, i had it water-cooled but the pump died. there are a couple of scorch marks on the water block it looks like a couple of tiny diodes. i might have to give this a go at some point.
I always watching your videos, and thank you that u mentioning the degree Celsius of hot air. Make more videos where here and waiting and starting learning
what a stupidity to solder with hotair near big capacitors. u shortens its life to hours or less. everybody who is not dumb will cover capacitors with kapton
at the start of page 13 of the datasheet at 21:57 sec, it's clearly written "if a phase is unused, the differential inputs to that phase's current sense amplifier must be shorted together and connected to the output"
this is cool, admittedly, but you say "I've ordered some parts". My problem with this is price to fix Vs. price to buy. Is this a better deal to fix or buy?
Almost certainly to buy. You need to develop good experience with graphics card repairs before Buy-to-fix becomes viable. It's why I'm doing practise runs like this, time and parts cost isn't a factor here, just learning the layout and what parts tend to fail.
I would, have tested the Mosfets, maybe they are fried and not, but a few working Mosfets are go to have, incase needed in the future. You just stated you had the wrong ones. I would Check all your little critters out.. you never know. Forgive my answer, for I am a beginner with no experience whatsoever sir.
