Думаю, что вполне возможно, дело в кабеле. С некоторых пор я сам делаю кабели. Чаще всего проблема в малом сечении. Это касается даже банальной зарядки телефона... Что уж говорить о прожорливой 4090?
Man I felt your pain here! I had a laptop that took me a few days because everything kept acting up. I slipped my probe during my final check thinking i was done and BAM slip pop short…. Laptop shut off so I took a deep breath and said yep that’s how you kill it 😂 all good now I was able to fix what I did and finish the job. Thanks for sharing we appreciate it!
Reminds me of a PS4 i fixed once, it was assembled and controller plugged in to test in front of the TV, then the doorbell rang and I stepped on the USB controller cable… when i came back, the console was off. Turns out a 5V trace in the power supply acted as a fuse and got really crispy… but all was fine after repairing the trace.
It's not a scam, people just don't bother reading the listings when they buy them. I've seen plenty of 4090s for sale all clearly marked as having the Core and Vram harvested from them.
There's plenty of scam listings too. A bunch of them have identical text pretending they don't know if the core and memory have been harvested, when it's visible from the side of the card in the video and the listings specifically avoid that view of the cards.
@@northwestrepair It's a partnership between Casio, Rolex, and Patek Philippe-a rare and extremely expensive watch. The model is Casio Digital (F91WG-9QDF)
Man, i can say i sort of understand what you've been through here. I work repairing vehicles ECUs and a lot of times we don't trust the "simulators" we use and end up having to troubleshoot them before testing the actual module. Wish i had kept track of the number of times i looked for problems on the module when the cause was on the tester, complete waste of time! Needless to say i'm impressed with how you dedicated to this board and glad it you fixed it. Great video as always!
So what are the variables on this repair. Does the core work? 🤷♂️ Does the ram work? 🤷♂️ Does the PCB have all the components? 🤷♂️ Is the replacement PCB flat? No. Does the riser work? 🤷♂️ This is so very impressive - as is the 9 seconds core lift. Nine!
Comment as requested and subscribed. Also want to say that your work is awesome. I cannot understand how you manage to keep motivation for so long. I would get so frustrated...that probe slip would make me scrap it on the spot. Then again, pretty expensive card...who cares :)
Hi Tony, what a mammoth work.. but did get to the bottom of the root cause. Great work and elimination theory worked well..! PCI-x16 raisers comes in voltage protection circuits and without. The one with voltage protection/ current limiter R-C circuits are nasty as they give you voltages but not the current as needed. Some of them has zeners which I hate. Getting few PCI risers/ extenders tested up with combination of various cards on hand, can be then labelled for few popular cards will surely fix this headache. Unfortunately, this will take some time but slowly and steadily this can be sorted out as more and more cards come in for repairs.. and your database grows on which combination of raiser / extender model needs to be used for a particular display card.. Hard job.. but well done!! 👍👍
After you complete the repairs of caps and resistors try using that riser again after switching that board power switch back to it's original position.
I like the Palit brand. My palit gamerock premium GTX1080 is still working and I'm very impressed with the longevity of this card. I used to stick with Asus GPUs, but I had problems with all of them (GTX680 blower style, 1080ti strix, 2080ti dual). As for the card featured in this repair video, it's obvious it wasn't used properly because only without a GPU stand can the PCB bend to such an extent.
However if you have any GPU that has GDDR5X or GDDR6X , you only get Micron. Nvidia and Micron co-develope these type of memory, so Micron has a monopoly.
Hello NWRepair. What I had noticed during the gpuz tests, that the card read outs were @ 2 x , and toward the end was only @ 3 x. When they should have been @ 4 x on all gpuz tests.
It could have been the capacitors on the GPU. Note that one cap was already gone completely, and as you said, the chinese maybe put a lot of heat on this board (because a lot of stuff shifted). High heat is bad for capacitors, it dries out the electrolytes inside if you do it over longer time. My guess is that the whole rows of capacitors was reduced in capacity by this. Result is that the GPU reacts much more on limited amps. If you have a riser between you have a longer line of power which means more resistance. The caps on the GPU are there to compensate low power spikes and if the caps are getting smaller capacity these spikes can freeze the GPU. You could test this theory by unsoldering one cap and measuring its capacity. If it is low replace all caps.
i just want to say a big THANKS to Northwastrepair for giving me the courage to invest in a heat gun solder iron and all the toold necessary for repairing some things i have my RTX 2090 Decided to defect purple one day and dropped to lower res Dreaded Error 43 i watched these videos and ran MATS and i managed to figure out which chip had gone bad, once i got my heat gun i Re-flowed the chip with 450c air and some flux all around the chip i gave it a slight nudge and could see movment and took the air away and since then the GFX card has been running amazing, Detects fine and plays games fine - I have 2 GTX 960s that will be my Test subjects for re-balling memory and running voltage tests got to start some where . THanks so much :)
Wanted a no power MB to get into soldering /fixing and verified with the sellers everything else was good including pins. They said yes so i ordered it and they mangled the board/ pins. I told the seller this isn't fixable anymore and was told he could have broken it in half because it's for parts. I feel the frustration with the f it attitude
Райзер дело тонкое . Не каждый стабильно 3,3В выдает особенно если на импульсной схеме сделано, может в карту 5В под нагрузкой выдать. А если по другому подумать Производитель наверное предусмотрел физическое определение места куда его втыкают и напрочь отказывается работать в майнинге.
On this episode of repair man vs 2 4090 PCBs with problems galore, we see a pop, some very off-placed things on a Chinese-repaired board, a PCB so warped you can see the curve with the naked eye, AND... the deadly riser cable problem! Will our hero prevail? Watch to find out. lol
Я собі брав Asus Prime AP201 корпус білий, він рази в три менший, компактне розміщення всередині, фактично використовується весь об'єм, старі чорні гроби продав би комусь. А ще краще miniITX малюсінькі але туди відеокарта жодна не передбачається, тільки на встройці чи якогось 5600G для роботи.
Do you ever see cores with render artifacts? In this video it seems like oh it basically works so let's keep looking around the board for issues. You know you have something like a 15% riser complication ratio on this channel.
I guess it's hard to find a riser that can supply the full 75 watts the PCI express standards call for ? That's pretty bad if that riser was only giving under 2 watts. A long time ago, I had a 1650 Maxwell core GPU that got all of it's power from the pex slot, it didn't require any PSU connections. I'm guessing a card like that plugged into that riser would not work very well either, because it just about maxed out the max power draw allowed thru the motherboard pex slot.
In photoshop you can do a differential picture of two pictures. All whats the same ist black in the differential picture and you only see the differences. Maybe its easier than doing a manual comparison.
Hello from Australia... Love your videos ! Very informative. Could you please provide link and/or information to the GPU test equipment you are using ?
If you look closely to the power connector 3:39 u can see that there is a power connector from the previous Gen rtx 30 series cards is printed next to the 12vhpwr . Now what I'm asking is that couldn't we just solder the connector on its printed place and solve the problem of melting connectors on this card and similar ones??
once i faced a problem similar to this and i was getting error code 43 in task manager, or system was freezing ... and at the end it all got solved after i changed the bios and lowered the pcie frequency, after that it didnt give me any headaches
7:42 May I know if it is possible to use the double 8 pin PCI-E power connector pads to replace the 12VHPWR connector since the pads there ? That solves all the melting connector problem of the 4090 . Even better if it's EPS12V connectors that are compatible with the pads on the PCB , those things can do >1kW on Intel HEDT CPU on LN2. Both of the PCB are definitely from Palit since they are the only AIB that has 8 pin power pads on any 4090.
Hello, I have a similar problem and the same PCI riser, but it's not the problem. My RX580 XFX posts in the PCI-E x16 slot, but there is no image with a riser in the PCI-E x1 slot. It gives an image, the drivers are loaded in the manager, there are no errors, but in GPU-Z there is no information about memory bandwidth, memory clock 0MHz, and the sensors do not show the fan speeds, and the GPU load is 100%. I have performed memory tests, changed motherboards, processors and switched bios with no result. Any ideas?
Думаю, что вполне возможно, дело в кабеле. С некоторых пор я сам делаю кабели. Чаще всего проблема в малом сечении. Это касается даже банальной зарядки телефона... Что уж говорить о прожорливой 4090?
Может быть. Попробую другой.
Так и окозалось. Кабель заменил, всё заработало.
Hurray to Google translate 😂
@@guiorgy its really crazy how good it translates, even in german.
This was definitely a roller coaster of a repair. Great job!
Man I felt your pain here! I had a laptop that took me a few days because everything kept acting up. I slipped my probe during my final check thinking i was done and BAM slip pop short…. Laptop shut off so I took a deep breath and said yep that’s how you kill it 😂 all good now I was able to fix what I did and finish the job.
Thanks for sharing we appreciate it!
Reminds me of a PS4 i fixed once, it was assembled and controller plugged in to test in front of the TV, then the doorbell rang and I stepped on the USB controller cable… when i came back, the console was off. Turns out a 5V trace in the power supply acted as a fuse and got really crispy… but all was fine after repairing the trace.
It's not a scam, people just don't bother reading the listings when they buy them. I've seen plenty of 4090s for sale all clearly marked as having the Core and Vram harvested from them.
There's plenty of scam listings too.
A bunch of them have identical text pretending they don't know if the core and memory have been harvested, when it's visible from the side of the card in the video and the listings specifically avoid that view of the cards.
I have nothing better to do...ohh but I do...I'm gonna watch new northwestrepair video
Exactly.
North Ridge has videos too! (LOL!)
I always smile when I hear, "Hello Internet" you know its going to be good.
Love the videos Man. Keep doing God's work by saving these consumers from evil manufacturers Shotty work.
Hi my name is Travis I watch your videos all the time I used your one video to fix a 3080 graphics card which I bought thanks for all the videos
god that probe slip hurt so much XD
Você por aqui, conserta meu relógio cassio do modelo Bolsonaro de 10zao por 5 reais? Se sim, eu ja mando agora
@@JoaoLui5 KKKKKKKK
No one fixes watches any more unless it's Rolex @@JoaoLui5
@@northwestrepair It's a partnership between Casio, Rolex, and Patek Philippe-a rare and extremely expensive watch. The model is Casio Digital (F91WG-9QDF)
O M G it's Gdtech, meu ídolo vendo meu outro ídolo
I just love how causally you work over your recording. Your awesome man
Man, i can say i sort of understand what you've been through here. I work repairing vehicles ECUs and a lot of times we don't trust the "simulators" we use and end up having to troubleshoot them before testing the actual module. Wish i had kept track of the number of times i looked for problems on the module when the cause was on the tester, complete waste of time!
Needless to say i'm impressed with how you dedicated to this board and glad it you fixed it.
Great video as always!
... "simulator"?
Watched the entire video! Loved every minute!
Thanks for the effort!
Incredible work!
Probe slips happen...the sparks are a bit bigger with vacuum tube equipment though :)
It's nice to see that you've used golden bars as a weight. Take that NRR😊! Now the ad is done, expect rich clienteles flowing in from Dubai! 😊😊
You sir are an artist
Perfect timing my friend, you're the best!
Nice PFP!
aAbsolutely amazing! Love the dedication! Really enjoyed watching this at 2am!
Great work man , love your perseverance, endurance and patience!
Your patience is beyond god-tier, holy shit. Great video, brother. Thank you for sharing, as always.
that was a whole lot of work. amazing!
What a cool card, i had a pair of palit 9600gt's back in the day they were bright orange
So what are the variables on this repair. Does the core work? 🤷♂️ Does the ram work? 🤷♂️ Does the PCB have all the components? 🤷♂️ Is the replacement PCB flat? No. Does the riser work? 🤷♂️
This is so very impressive - as is the 9 seconds core lift. Nine!
Comment as requested and subscribed. Also want to say that your work is awesome. I cannot understand how you manage to keep motivation for so long. I would get so frustrated...that probe slip would make me scrap it on the spot. Then again, pretty expensive card...who cares :)
That certainly was a struggle, experience gained but hard.
What a rabbit hole..... all respect to you as you dont give up...
I love your repairs they are so soothing.
I'll say it again "tenacious"; very very well solved. Thanks for the vid.
Dude, you've got lucky with the ✨"ignition". Other than that, you've got SOME patience fixing these.
Respect for your patience and good work
Wonderful recovery.
Hi Tony, what a mammoth work.. but did get to the bottom of the root cause. Great work and elimination theory worked well..! PCI-x16 raisers comes in voltage protection circuits and without. The one with voltage protection/ current limiter R-C circuits are nasty as they give you voltages but not the current as needed. Some of them has zeners which I hate. Getting few PCI risers/ extenders tested up with combination of various cards on hand, can be then labelled for few popular cards will surely fix this headache. Unfortunately, this will take some time but slowly and steadily this can be sorted out as more and more cards come in for repairs.. and your database grows on which combination of raiser / extender model needs to be used for a particular display card.. Hard job.. but well done!! 👍👍
Imagine someone watches the whole video and finds out it was just the riser hehe
The riser was one of many parts needing replacement. I wouldn't give you 2c for 99% of the risers on the market.
I watched the whole video 😂
an older video was the audio driver
I watched the entire video thinking he made a mistake lol
Bro... No spoilers!!!
After you complete the repairs of caps and resistors try using that riser again after switching that board power switch back to it's original position.
I like the Palit brand. My palit gamerock premium GTX1080 is still working and I'm very impressed with the longevity of this card. I used to stick with Asus GPUs, but I had problems with all of them (GTX680 blower style, 1080ti strix, 2080ti dual). As for the card featured in this repair video, it's obvious it wasn't used properly because only without a GPU stand can the PCB bend to such an extent.
yeap, asus really likes to mess up the protection circuits and overall having terrible designs in a lot of products not just their gpus.
My main card is a Palit and it works very well. It has Samsung memory chips.
However if you have any GPU that has GDDR5X or GDDR6X , you only get Micron. Nvidia and Micron co-develope these type of memory, so Micron has a monopoly.
@@fleurdewin7958 no mine is the regular model 3070 so it has regular gddr6.
Snubber networks are used to protect the switching device from ringing (voltage spikes) caused by the parasitic elements in the circuit.
That is why you always test your new setup with known to be working components first.
woww that is a lot of work
Good work man
Im pretty sure I've seen a few of these 4090's floating around on eBay before I see them end up in your videos. Lol
Hello NWRepair. What I had noticed during the gpuz tests, that the card read outs were @ 2 x , and toward the end was only @ 3 x. When they should have been @ 4 x on all gpuz tests.
It could have been the capacitors on the GPU. Note that one cap was already gone completely, and as you said, the chinese maybe put a lot of heat on this board (because a lot of stuff shifted). High heat is bad for capacitors, it dries out the electrolytes inside if you do it over longer time. My guess is that the whole rows of capacitors was reduced in capacity by this. Result is that the GPU reacts much more on limited amps. If you have a riser between you have a longer line of power which means more resistance. The caps on the GPU are there to compensate low power spikes and if the caps are getting smaller capacity these spikes can freeze the GPU. You could test this theory by unsoldering one cap and measuring its capacity. If it is low replace all caps.
i just want to say a big THANKS to Northwastrepair for giving me the courage to invest in a heat gun solder iron and all the toold necessary for repairing some things i have my RTX 2090 Decided to defect purple one day and dropped to lower res Dreaded Error 43 i watched these videos and ran MATS and i managed to figure out which chip had gone bad, once i got my heat gun i Re-flowed the chip with 450c air and some flux all around the chip i gave it a slight nudge and could see movment and took the air away and since then the GFX card has been running amazing, Detects fine and plays games fine - I have 2 GTX 960s that will be my Test subjects for re-balling memory and running voltage tests got to start some where . THanks so much :)
33:00 yeah that IC hold the line and save the core
Wanted a no power MB to get into soldering /fixing and verified with the sellers everything else was good including pins. They said yes so i ordered it and they mangled the board/ pins. I told the seller this isn't fixable anymore and was told he could have broken it in half because it's for parts. I feel the frustration with the f it attitude
Райзер дело тонкое . Не каждый стабильно 3,3В выдает особенно если на импульсной схеме сделано, может в карту 5В под нагрузкой выдать.
А если по другому подумать Производитель наверное предусмотрел физическое определение места куда его втыкают и напрочь отказывается работать в майнинге.
Надо будет посмотреть как себя ведёт 3.3в. может продает.
Excellent save!
That blue screen crash was related to drivers and IRQ. Even current days that can happen to anyone
On this episode of repair man vs 2 4090 PCBs with problems galore, we see a pop, some very off-placed things on a Chinese-repaired board, a PCB so warped you can see the curve with the naked eye, AND... the deadly riser cable problem!
Will our hero prevail? Watch to find out. lol
Awesome video
I don't understand why utube sent me here. But its amazing.
How do you learn to do all of this kind of work?
Я собі брав Asus Prime AP201 корпус білий, він рази в три менший, компактне розміщення всередині, фактично використовується весь об'єм, старі чорні гроби продав би комусь. А ще краще miniITX малюсінькі але туди відеокарта жодна не передбачається, тільки на встройці чи якогось 5600G для роботи.
Watching your vid almost 2am 😂
😴
Do you ever see cores with render artifacts? In this video it seems like oh it basically works so let's keep looking around the board for issues. You know you have something like a 15% riser complication ratio on this channel.
48:05 yeah, different revisions, it made a hard job even harder
Good work 😅
My first thought when you said you were going to solder on the ram then bend the board back was surely it'd be better to do the bending first no?
I guess it's hard to find a riser that can supply the full 75 watts the PCI express standards call for ? That's pretty bad if that riser was only giving under 2 watts. A long time ago, I had a 1650 Maxwell core GPU that got all of it's power from the pex slot, it didn't require any PSU connections. I'm guessing a card like that plugged into that riser would not work very well either, because it just about maxed out the max power draw allowed thru the motherboard pex slot.
Top👍
Reminds me of sling blade . When Carl asks does it got gas in it MMMMHMMM .. lol
crazy man! cool@
In photoshop you can do a differential picture of two pictures. All whats the same ist black in the differential picture and you only see the differences. Maybe its easier than doing a manual comparison.
That was impressive mate, Jesus! what do you charge for all that work was it worth repairing? Bob
Trolled by your test rig. Sometimes it just feels like ...
Hello from Australia... Love your videos ! Very informative.
Could you please provide link and/or information to the GPU test equipment you are using ?
17:26 if you're talking about that tool, it's DIY.
good bideo
If you look closely to the power connector 3:39 u can see that there is a power connector from the previous Gen rtx 30 series cards is printed next to the 12vhpwr . Now what I'm asking is that couldn't we just solder the connector on its printed place and solve the problem of melting connectors on this card and similar ones??
i am so surprised at the end components stuck to the thermal pads.
That Palit looks like it got hit by a tornado.😮
once i faced a problem similar to this and i was getting error code 43 in task manager, or system was freezing ... and at the end it all got solved after i changed the bios and lowered the pcie frequency, after that it didnt give me any headaches
so bright so see the video bro
Praise comment
Is there any purpose to those 220nH inductors (R22, LR22 in this v9ideo), or are they putting them just to complicate things?
I never thought about having a rhinestone encrusted GPU... For a reason.
7:42 May I know if it is possible to use the double 8 pin PCI-E power connector pads to replace the 12VHPWR connector since the pads there ? That solves all the melting connector problem of the 4090 . Even better if it's EPS12V connectors that are compatible with the pads on the PCB , those things can do >1kW on Intel HEDT CPU on LN2. Both of the PCB are definitely from Palit since they are the only AIB that has 8 pin power pads on any 4090.
Looks like it.
Noice 👊🏽🤠🇨🇦
Que modelo se calentador usas para el reballing
47:50
Its much easiert to compare those two boarda with a 3d technic called cross eye.
If anyone has a 4090 kicking around that they would like to donate, I would be very happy to receive it! 😁
They should make PCB more rigid and thicker to prevent this.
That’s a ridiculous shroud haha
I had riser that had to be set to GEN3 PCIE in bios rog strix x570-e gaming
How do you find you video card diagrams and schematics? I need a schmatic for my zotac 1660 super twin oc so i can replace a knocked off component
"and remove the googles" lmao
At 10:30 what kind of flux did you wipe away so easy without residue?
@20:22 was that a Crack in the pcb? I haven't watched to the end yet, didn't want to forget to ask
what are riped pads?
Next episode: So about raisers..
@ 4:00 do those smd ploymer caps count as electrolytic? I guess they are polar..
how do you get that kind of bending ?
38:58 - those large box two things just before video source change - looks like those are short circuited?
No.
12:55 , 5 sec later , heheheheheh . the magic of edit
Would it not help adding additional power to the riser via the molex or 6pin ?
Don't think so. I am not sure exactly which power rail is sagging.
How much does it cost the customer for something like this I wonder
Hello, I have a similar problem and the same PCI riser, but it's not the problem. My RX580 XFX posts in the PCI-E x16 slot, but there is no image with a riser in the PCI-E x1 slot. It gives an image, the drivers are loaded in the manager, there are no errors, but in GPU-Z there is no information about memory bandwidth, memory clock 0MHz, and the sensors do not show the fan speeds, and the GPU load is 100%. I have performed memory tests, changed motherboards, processors and switched bios with no result. Any ideas?
maybe you should have a working GPU around to test the riser
☺
big boss says remove the goggles
👍👌👍