пытаюсь отремонтировать 2070. Он пришел с коротким замыканием на 12v и на 5. Я заменил 12 на 5 (IAMXJ), контроллер памяти up1666Q, все 6-фазные контроллеры UP1965P.

после этого карта сработала. Я решил провести несколько тестов с пластиковой задней панелью и без нее, чтобы посмотреть, не будет ли без нее лучше. Они были такими же. Я поставил заднюю панель на место, а затем после сборки частота ядра не превысила 300 МГц. Память доходит до 2000 МГц как обычно.

сопротивление выглядит нормально кроме 12в. После того, как я заменил закороченные части, это было примерно 3ком, а не 385ом. Снял IAMXJ, все еще 400ом.

Я думаю, что мосфеты памяти подключены к 12v от PCI, но память работает.

Какие-нибудь советы? у вас есть схема, пожалуйста?
Posted after 35 minutes 48 seconds:
NCP45491XMNTWG это один виноват в 385ом

---

Posted after 1 hour 39 minutes 11 seconds:
после замены NCP45491 4,5 ком на 12в.
все еще не работает. 276%TDP


что очень обидно, так это то, что работало, пока я не снял и не поставил пластиковую заднюю пластину

Take a photo of a back side please.

I will tomorrow morning, I left work.

Завтра утром пойду, ушел с работы.

When you have such big TDP it means you broke a resistor somewhere, maybe a shunt resistor, maybe something smaller responsible for analog part of the measurement. Higher TDP means higher voltage drop on the shunt maybe there is a voltage divider along the traces, maybe NTC somewhere, depends on application. But definitely you broke something.

What are voltage drops over all shunt resistors? Did you try to measure them? There is an Ohm's law, you can immediate tell when it is 0.05 or 0.5 ohm.

I will search tomorrow, I figured there is something missing.
It worked ok after repair. But I wanted to see if taking down the plastic backplate can make a difference in temp. It worked again, without the plastic backplate. Then when I put it back, it stopped working. 300 mhz max on GPU.
I measured resistance and 12v PCIE had 400Ohm. Since about everything on 12v was shorted before, I figured that I must change that ON chip NCP45491. After I replaced it that higher tdp appeared. Clearly old NCP45491 since 400ohm is not ok on 12v. And I guess that chip is involved in power management?

Maybe I didn't put it back right, or I moved a resistor when soldered that QFN. or there is another part missing somewhere.

---

Posted after 5 minutes 26 seconds:
what are shunt resistors? sorry but I don't know ?

This chip is like INA on other cards. It is quad voltage/current sense. Maybe you solder it not properly? Maybe this chip is dead, maybe some components around?

https://www.arrow.com/en/research-and-e ... -resistors

thanks. So my guess is right, it has to do with power management, like INA.
I took the chip from another card, a 2060 super with a broken GPU memory controller. I was careful in soldering, but who knows. I will check.
I read about shunt resistors, will check them also.

---

Posted after 4 minutes 35 seconds:
Can you please tell me what is this chip ?

This is mosfet, power switch.

anyone has the code for that mosfet ? P8 GUE

The problem was a resistor that didn't sit well after I soldered QFN. I attached pictures with the board in case someone else needs them for other repairs.

MUX_SEL - it is the main clock for the operation of this chip. It is not strange it did not work

anyone knows if this is palit? it says pc partner LTD

it came with a burn on pcie 12 v. I cleaned it. Now all voltages work except vcore. I replaced UP9512R, same thing. enable is 2.5v, pgood is 3.3v phases controllers receive 12v and 5v enable.

maybe there is a connection on the old dremeled burn, but it doesn't look like. Anything specific to look at UP9512?

strange thing is that on 12v pcie, on first measure there are 6.27kom, and if I measure again, keeping the same polarity resistance is 9Mom.
Posted after 4 hours 15 minutes 23 seconds:
I guess these three lines must be connected.

PEX Reset is not connected
Posted after 21 minute 37 seconds:
pex reset is buffered, seems PC Partner knows their ways
Posted after 1 hour 33 minutes 1 second:
I guess this is not OK. GPU_NVDD_SENSE and Therm Overt are almost shorted. please see pictures

---

Posted after 1 hour 28 minutes 37 seconds:
all ok now. sorry for my monologue....