Общее потребление с/без pcie соединения - 0.914 A, и растет, соразмерно нагреву чипа.
1.8v - - 420 ohm - 22.9 mA
1v - - - 93 ohm - 13.5 mA
5v - - - 769 ohm - 154.5 mA
NVDD 0.83v - 0.22 ohm - 1 A
FBVDDQ 1.35v - 56 ohm - 2 A
На диодной:
smclk/dat - ol
pex_reset - 1.31v
refclk - 0.793v
По стороне A - 0.453v
По стороне B - 0.308v
8915FN-56 не греется, видимых прогаров нет. На всех логических "И" есть 3.3.
Генерация на кварце есть, все напряжения есть, шимка получает VID 0.901, дает pwrgood, шевеление на биосе есть.
Биос прошит оригиналом с techpowerup, обоими версиями поочередно. Судя по всему, что-то мешает плате выйти из состояния "ожидания" pcie соединения в полноценный запуск, хотя соединение она видит, раз начинаются шевеления на биосе. Напряжение на чипе и памяти не меняется. Я вот и не знаю, что отвечает за последующую инициализацию платы, но думаю, что скорее всего чип. Возможности перекинуть на другую плату чип и проверить там - нету, как и обратного.

Is pex_reset buffered or not? Take a picture please. 1.31V looks a little too high. Please measure it with red and then black on the ground.
0.9A consumption is not enough, should be closer to 1.5A. Something is not running.
1.8V ~400 ohm .. should be closer to 800-900
By the way welcome to the club

My bad, I had measured pex_reset with a red probe. With black vdrop are 0.637. And what do you mean by "buffered"? Logic comparator?
1.8v seems wrong to me too. My other 1080(inno3d) has 750 ohm and consumption around 14.5 mA.
Thanks. I'm checking this site from the beginning of 2015, but this is my first post)

This is the best website I have ever seen
Oh this 1V8 ~400 ohm does not look good. Maybe another chip on this rail is faulty, maybe the bios chip. Replace the bios and write the proper from techpowerup. When you remove bios measure resistance...

I have such cards every day. Even with everything in order but not detected.

Without bios ic - resistance are same. Removed 1.8v from "AND" gate - same 400 ohm.
Removed 0ohm resistors from 1.8 power to gpu, got 857 ohm on gpu and 767 ohm on 1.8v choke.

Interesting. 1/800 + 1/800 = 1/R so it is ~400 Take a measurement of a working card, what is the resistance of GPU alone and you will know and we will know for your particular card and GPU model.

да сопротивление 1.8 В 400 Ом маловато ,мне кажется .pex_reset - 1.31v - что Вы подразумеваете под данным сигналом - сигнал с пина а11 PCI-e слота ? если да - то имхо там должно быть 3.3 в сигнал вроде называется PEX_RST под буферизацией вероятно понимается отсутствие или наличие ИМС логики 2И - см кусок схемы - т.е. 3.3 вольта с гребенки идут через резистор непосредственно в ГПУ - этим сигналом снимается ресет с видеочипа и он начинает работу - ну в частности вычитку биос - гляньте любую схему 10ХХ - все сразу станет понятно. Т.е микросхема логики бывает просто не запаяна на плате. Если запаяна - то она выдаст насколько я понимаю логическую 1 - 1.8 в а не 3 В . Если ресет не снят с ГПУ (снят - высокий уровень не снят - низкий) - то карта естественно нигде не будет видна . Может Вам стоит обратить внимание на сигнал PEX_RST если он у Вас прям на слоте PCI-e равен 1.31 в (хотя мож я ошибаюсь и это не его Вы имели в виду) И мне не совсем понятно в таком случае как у Вас происходит вычитка биос - если не снят ресет с ГПУ . А по остальным сопротивлениям питальников,дифф линий - вроде норма . И да - попробуйте для очистки совести подключить карту с другой материнкой блоком питания - ну если еще не делали этого - всякое бывает
---------- Добавлено спустя 10 минут 56 секунд: ----------
фотки раньше не обратил внимание - 1-я фотка - да микросхема 2 И не распаяна - сигнал через резистор должен идти прямиком в чип

Dear crocodile70. We are talking about voltage drop measured between PEX_RST and GND in diode mode when the card is not connected to anything.
Every pin of the integrated circuit has some kind of static electricity protection formed mostly by diodes, one to ground and one to the working, expected power supply. Some diodes are parasitic diodes formed from semiconductors, for example between drain and source in mosfets.

When you connect two semiconductors N and P the phenomenon happens. The adjacent depleted region P attracts electrons from enhanced region N nad they create uniform region with equal amount of electrons. In order to force electric current through it, called a barrier, you have to overcome this phenomenon. This is why silicon diodes have a voltage drop around 0.7V, Schottky 0.3V, some LED diodes even volts.

When we can measure this barrier we can be sure that there is a junction present and probably this GPIO was not shoot over. We are talking about forward direction where electrons flow from cathode to anode. The reverse behavior is a different story, it looks close because this barrier repels electrons and extends the size of barrier preventing the flow of any current until it breaks.

Your schematic presents a circuit with a 0 ohm resistor instead of a logic gate. We call it a buffer, buffering logic, glue logic. Without this AND gate PEX_RESET will definitely die when you discharge a static electricity on the card. They saved 1 cent but more cards will go to trash.

This is the same situation with other signals on the PCIe bus like Reference clock, Lane outputs. Inputs are somehow protected by a series capacitors.

Yes - I understand It was about diode dialing and not about voltage measurement Sorry

Meneer Pandemonium6, do you have vcards with 1080 gpu which have separate connection to the rest of a 1.8v power rail? Like I mentioned - mine have a 0ohn "fuse" resistors, so I can measure somewhat "pure" gpu resistance on it's 1.8v power. Because my second 1080 - inno3d seems to have everything connected together on a rail, without separating with resistors. Indeed it have a two of them right behind a choke, but for obvious reason this is useless for me.

Yes I have, plenty of them. All with 0 to 10 ohms, maybe 10 of them All completely dead. I will measure if I find one with good 1.8V. I can even unsolder GPU and measure it on balls on top capacitors.

I'm very appreciate your support. I'll be glad if you share any good 1.8v resistance =)
I think that resistance on taken off board gpu still can be different. I don't certain with that, but it's my thoughts.
I'll be waiting for any further information, thanks!

I just checked on boardview the one candidate. Only 3 balls connect 1V8 to the GPU (BA10, BB14, BC14) and this rail goes to plenty of small transistors, resistors and logic gates. 1V8 goes also to the missing AND gate of the PEX_RST* haha.
So in this case PEX_RST*(3V3) would become PEX_RST_BUF*(1V8) with pull down on the output of the gate in order to keep it low when the gate is not yet powered from the 1V8. Very clever because it prevents GPU from latch up. But of course PEX_RST* is not buffered and this pin is susceptible to die from static discharges, overvoltage, short, reverse riser.

But in my case pex_rst seems good, even without this safety measure.
But anyway, I'm very unsure in 1.8v on gpu, and what role it plays in there.
I'll try to find more information over ethernet about 1.8v resistance on this gpu.

I did not remove this chip yet, no time today but 1V8 should be around 800-900 ohms. This is what people say and you can read or watch even on Youtube. The role is simple, connect some higher voltage signals, lower voltage signals through level shifters composed of logic gates with TTL inputs, open drain and open collector transistors, Schottky diodes and of course a 1V8 bios chip. This is such intermediate low voltage bus, not as high as 3V3.
Gates in mosfet transistors are very thin and if the technology process is several nanometers it is a big concern because insulated gates are thin and it is easy to kill it. Transistors are smaller, less electric charge is needed so the voltage must be lower. Gates do not conduct current in mosfet transistors but too high electric field will kill the gate. So it is irresponsible to connect 3V3 to the PEX_RST* (negative logic, active low) when it is probably 1V8 pin and maybe only higher voltage tolerant. Signal could pass through a resistor divider and it would be protected, even 100 ohms in series instead of 0 ohms would prevent damage.

Before removing GPU resistance was around 870 ohms on 1V8 rail.

Next step GPU removed, measured on 1V8 inductor: 902 ohms

There are 3 balls under GPU so I measured each one.
Measured on GPU with BLACK to the GND: 45k, 5M, 10M
So two measurements look like open circuit in megaohm ranges.

Measured on GPU with RED to the GND: 43k, 96k, 130k

Your GPU is dead with 400 ohms. I am not sure what values should be on factory new GPU. Why do I have 45k on one ball and an open circuit on two others? I don't know. This card was dead anyway.

Aww, that's slightly sad, but I was ready for some sort of this outcome from the early beginning of diagnose this vcard.
Anyway, I'm very gratitude for your participation in this little "investigation", and I'm very appreciate your contribution, mr. Pandemonium6, at least now I know slightly more about 1.8v rail. Good repairs to you!