RTX3080 GamiingX Inno 3D GDDR6X 10 Gb - нет изображения не видна в Виндовс второй картой Напряжения NVVDD MSVDD FBVDD 1V8 5V PEX_VDD присутствуют Сопротивления питальников - в норме (сравнивал с фото одного известного блоггера)
Прозвонка всех диф пар в норме. Отклонение - контакт а11 PERST - на нем сопротивление - 120 Ом. Также на микросхеме U811 - логический элемент 2И SN74LVC1G08 - сопротивление на выходе - пин 4 - 18 Ом - это сигнал PEX_RST_BUF - снимает ресет с ГПУ
Микросхему заменил. Без нее сопротивление на пин 4 было высокое - забыл значение - померил с уже впаянной другой рабочей - снятой с донора . Т е логику я поменял. Сопротивление на входах U811 теперь - бесконечность также и на прозвонке. На выходе - пин 4 - сопротивление 3.38 ком прозвонка 2.19 В и 0.67 В - разной полярностью если.На входах U811 - 3.3 В,на выходе - 1.8 В. Т.е. как я понимаю логика отрабатывает правильно. Но картина не поменялась .Карточка по прежнему не видна в системе. Вычитки биоса на осциллографе нет. Генерация на кварце есть. Чип достаточно быстро сильно разогревается.
Хозяин карточки утверждает что не путал райзер при включении - неисправность произошла как то сама собой в процессе работы - ну тут я ничего добавить не могу -не знаю .
Подскажите, что еще можно проверить на данной карте,так как опыта ремонта 30ХХ линейки у меня нет . И так же если у кого есть на руках RTX 3080 - подскажите пожалуйста сопротивление на пин 4 U811 - сигнал PEX_RST_BUF.

I have seen this many times, no bios activity with all voltages present and all resistances correct. This must be a common failure. I had the same symptoms on 10xx, 20xx and AMD Vega and Radeon 7. Very strange no bios activity on any pin.
Some cards are recognized in the system as Standard VGA display adapter which means the PCIe communication is alive. Of course no image but as a secondary card.

Привет на днях были две карты rx470 после падения фермы на пол имели идеальные замеры и никакой активности по гпу, отклонение было только на контакте b11 поиски куда чего идет никчему не привели и металлической метелкой не получилось найти. От безвыходности подал с лбп 0,5в на данный контакт и тепловизор помог выявить мелкую пиз-у на 6 контактов размером 1х2мм я хз что за деталь нашел такуюже в донорах и о чудо все ожило и заколосилось

выпаял микросхему логики 2 И теперь сопротивление на пин 4 - PEX_RST_BUF 3.38 кОм падение напряжения 2.19 В что одной что другой полярности тестера

B11 is a WAKE# signal. The first thing I do is to check all signals up to A18 and B18. I do it in diode mode and in resistance mode. If there is an obvious short, low resistance it is worth of immediate investigation. This is not the case on such cards as I described above.

For example I have white Gigabyte with turbine 1080ti with dead B14 HSOp(0). This card does not work through the riser but inserted into the older motherboard is normally detected and mining is possible. MATS reports no errors.
The voltage drop on B14 is 2.1V, on all others around 0.5V. It is not completely open circuit but looks like half broken.
Posted after 6 minutes 13 seconds:


This is good reading of the pure unbuffered PEXRST. You can check it with a scope. There should appear the negative pulse. PEXRST buffer, this small AND gate can fail, definitely.

---

Posted after 2 minutes 27 seconds:

Who is this guy and where are his photos?

You can't measure the resistance of the single data gpio pin. It will not work. Maybe when the pin is totally shorted you will have low resistance but you have to test GPIO pins in diode mode in both polarities. The body diodes should look intact. When the PEXRST is buffered by xxxx08 gate you have to remove this gate and measure without it from pin4 to ground.

yes of course -
without U811 resistance 3.38 Kom voltage 2.19 V - in both polarities

One more thing. I try to understand translation form Russian to English. So please write in Russian in more formal language with correct grammar because "translate" produces garbage

You say that without the AND gate the resistance was high. What was the voltage drop in both polarities? Maybe faulty AND gate passed the 3.3V from one input to the output and killed this buffered PEXRST? PEXRST is not 3.3V tolerant on these cards and this logic gate works as a level shifter. There should be additional series resistor like 100 ohm in series for additional protection.
Try to measure pin 4 when you remove the AND gate. Measure it again and them measure also the AND gate itself.
Unbuffered cards give some voltage drop in both polarities, higher is with the black probe on the ground.
Posted after 8 minutes 32 seconds:

And this is very suspicious. Get in your hand any unbuffered card and measure the voltage drop. It should be 0.5-0.8 in one polarity with red to ground and around 2k in opposite polarity.
It means that PEX RESET ball on the GPU is shoot over probably by 3.3V which passed this buffering AND gate.

You're right. I have 2 working cards P106-100. a11 - unbuffered input. Resistance 1.4 Mohm voltage 0.45 volts. Reverse polarity - resistance - infinity, voltage - infinity
I think the video chip is damaged. Thanks for the help!

On unbuffered cards there is a pull up resistor to 3.3V something like 4.7k or 10k. It is near the missing buffer gate. Remove this pull up and measure again.
Or remove 0 ohm bypassing the buffer and measure behind it.

We are like medics. Our practice is the most important factor. So please finish these measurements.

Sorry, I don't see it in the diagram - pull up resistor to 3.3V . Unbuffered card.Please clarify - I will take measurements.

Maybe it is pull down, my mistake This 10k with cross on it. It keeps the card in reset before PEXRST appears. Take picture, closeup of this missing gate please...

this is my GTX 1060 card - memory issues. Code 43. Start - yes.
The buffer is not wired from the factory

Ha! On my cards, which are already dead, 1070 and 1080 there is this resistor 10k to GND. I made mistake yesterday saying it is pull up. Definitely I have pull down 10k to GND and zero ohm bypassing the logic gate.
I will investigate it further today afternoon. I will find more cards and I will make measurements with and without resistor in both polarities.
We have to solve this puzzle what voltage drop should be and what is bad. I have many dead boards here.

Interesting. Waiting. Especially 20XX, 30XX.

I checked 7 unbuffered Nvidia cards.

Minus to GND / Plus to GND

Infinity / 0.45V
Infinity / 0.41V
Infinity / 0.41V
Infinity / 0.40V
Infinity / 0.45V
Infinity / 0.52V
Infinity / 0.45V

This Infinity is almost infinity, the last digit at the multimeter changes +/- 0.003V so multimeter sees it as almost open like 2.997V
It means that good PEXRST on Nvidia unbuffered should be 0.4V to 0.5V with red probe on ground and Infinity in the opposite.

With AMD it is more complicated but not open circuit in any direction.
It can be 1V/1V, 0.7V/0.7V or asymmetric with lower value with red on GND but not open.

Thank you !

And remember about other signals like REFCLK differential pair, WAKE#, CLKREQ and SMBUS. I had a card with one signal of REFCLK dead. The card was not detected.

When the card has PEXRST buffered I test this signal on pin 4 with oscilloscope. This is the best test and of course the voltage on pin 5. This buffer is not for protection, it is for level shifting the incoming 3.3V into 1.8V. As you can see it can kill PEX_RST_BUF directly connected to the ball of the GPU.