Всем привет.

I have a strange problem. One DrMos NCP302155 which is under control of the NCP81022 doesn't get HIGH state on DISB# pin number 31.
The problem is that I can't find the component it is connected to. I tried to tough the whole board with a steel wool but it sims that either I can't find it or the trace is broken somewhere. There are no faulty components, I think, and no missing. I was the first inside and I removed the seal. PWM signal is present on the input (1) of the NCP302155 but the DrMos is disabled by the pin 31. I can't locate how is it formed since we don't have schematic nor a board view.

This DISB# pin 31 has a voltage drop of about 570mV in reverse polarization and a resistance around 500 kohm so it is alive, not "KZ" Definitely something does not assert this signal into the high state or it is unconnected.

The second, exactly the same circuit, gets this signal directly from NCP81022 (pin 32 DRON) as I have drawn on the image and 0.9V works. All other voltages work. Resistances are as on many 5700 cards and on this faulty choke it is around 60 ohms the same as on the memory choke on the right. But the right choke has a different DrMos and is controlled by a separate PWM along with core phases.

This card stays at b2, typical behavior.

This card uses split memory power supply 1.35V and for an unknown reason the left doesn't work. Both are apparently controlled by NCP81022 on the left of the PCB.
I can't find a trace or components where this DISB# of the NCP302155 signal is connected to.

OK, I have progress but I need help if someone has PowrColor Rx 5700 Red Devil, (NOT XT) Please look at board what is in place of the left Integrated Driver MOSFET for memory voltage.

I measured PCB and the enable signal goes to all three chips but memory mosfets have enable on the 30th pin !!!!!!!! It will not work !!!!!

Maybe somebody tried to repair this board and probably replaced NCP302045 -> NCP302155 by mistake. There are mostly NCP302155 and he made mistake. The difference between pins is such that pin 31 is on pin 30 and pin 30 is on 31. This pins are swapped between these integrated circuits !!!!

The warranty sticker was on the screw, maybe fake, board was very clean but definitely enable signal goes to the wrong pin of the left NCP302155. It will never work this way. This is magic.

Мне нужна картинка печатной платы, особенно память mosfets. Какие символы на них?

Looks like NCP302155 is in place

Damn it, so what is responsible for pin 31 DISB# ?
Definitely multimeter beeps at pin 30 of this left Mosfet and pin 31 looks unconnected in LOW state. LOW state = Pin 31 DISB# LOW = Mosfet OFF. This card gives a picture in bios but hangs immediately after and this Mosfet never gets DISB#=HIGH. PWM is delivered to the Mosfet but no movement on the choke. Like disabled transistor.
My schematic of the DISB# trace is good.
Posted after 9 minutes 1 second:
Moment, my friend Images are from Red Dragon version. I have Red Devil. Maybe there is a difference? Maybe someone will have it in hands one day and could tell when the left portion of VMEM starts and where is DISB#...
Under Linux I can flash the bios and I flashed the bios from the techpowerup to one bank. No change of course.

Under linux there is communication with the GPU so I can flash a bios but I can't run any test, it hangs immediately and the left VMEM never starts. There are four bios files on the techpowerup. I checked them all.

Please look at attachment. How can it work when the DISB# signal is connected to the open drain output of the overtemperature sensor ??????? Output (enable signal) is connected to overtemperature output.

DISB# is internally (inside the chip) pulled low so when unconnected will never allow it to run.

What the damn mystery?

302155:
"30 THWN Thermal warning indicator. This is an open−drain output. When the temperature at the driver die
reaches TTHWN, this pin is pulled low."

swap u702 u704

I will have 302155 and 302045 next week. I could replace and test but I still don't know whether this left is to work from the beginning or not. And of course why this enable signal goes to the wrong pin.
This is sensitive operation and overvoltage there will kill the GPU so before I do something crazy I would like to know

There are just some questions regarding Red Devil version.

1. Is the left MVDD Mosfet 302155 or 302045? (They may have a different PCB revision)
2. Is the DISB# signal on pin 31 high or low?
3. Does this VRM start from the beginning, are there pulses on the choke?

Replacing these fine pitch Mosfets is very dangerous in general and I don't tough them when not needed. You can always break something, make KZ between pins. I have many other cards to work on so some days will not make a difference.

Even if this Overtemperature pin is used on the left 302155 to shut down enable to others so the question is why this way and what happened to the enable to pin 31? Swapping chips when you see that definitely there is missing one signal is not a good idea. I don't want to force pull up this 31 pin yet before I understand what I am doing.

Why is the measurement on the other side, because the first pin and 30th is in the opposite corner?

Yes you are right, my mistake when drawing this schematic of DISB#.
I measured the right 302045 properly. It was the bottom right corner. The left is drawn properly.
I came from work and started drawing at home from memory. There are empty pads (for 0402 components) also connected to this pin 30 so it is easier to probe by multimeter. I will draw again in the moment...
Posted after 21 minute 38 seconds:
Yes, the left was drawn properly. The problem is that on the left 302155 the pin 31 goes to nowhere, there are empty spaces for 0402 components. There is a via hole and there is maybe a trace in interlayers but I can't find where it goes No doubt it is in LOW state. The third chip on the bottom left, connected to the same net i 302155 and has this DISB# on pin 31 and it goes HIGH, the same as pin 30 on the right 302045.
Posted after 2 minutes 58 seconds:
I have seen many fake warranty stickers but this board looks a little bit suspicious in this place we are talking about. Maybe it was ultrasonically cleaned, maybe I am wrong but the question is why pin 30 of this left MVDD gets DISB# on the wrong pin...
Posted after 20 minutes 39 seconds:
I measured it again 1 minute ago. No doubts, right 302045 pin 30, left 302155 pin 30, bottom left 302155 pin 31.

---

Posted after 42 minutes 6 seconds:
1. I can write, read, test, verify bios from Linux, no problem.
2. I can run dmgg.py which gives errors on all banks.
3. Tserver hangs, so I can't run anything else

It's definitely 155 there, proper board this time)

Try to use Gigabyte 5700 boardview from this topic, device looks like mostly similar to Powercolor

Than you for help but your image is from 5700XT version probably. There are 2 x 8 pin Molex, but I have 5700 without XT 1x8pin Molex. + 1x6pin Molex.
Posted after 21 minute 49 seconds:
OK, I am analyzing this board view. It looks like it is made on SIC620A(302155) and SIC632A(302045) on the right, they look similar but all SIC chips have DISB# on pin 31. So the PCB revision of the Powercolor may be another revision correcting location of the DISB# pins or somebody tried to repair my board and made mistake.
Nothing wrong happend yet because LEFT MVDD simply do not run.
I still wait for somebody with this exact PCB

What exactly version do you have? What is on the image from your post above? Could you please ring DISB# traces?
Posted after 10 minutes 21 second:
I will tell you more...

According to the boardview:

NCP81022 PWM pin 32 DRON is a chip select for these three power stages. They go through 0R resistors

1. R829 to U704 (Bottom left 0.9V) -> Pin 31
2. R828 to U703 (Right Memory power supply) -> Pin31
3. R827 to U702 (Left Memory power supply) -> Pin31

All is OK because both SIC620A and SIC632A have DISB# pins on 31 !!!!!
Posted after 11 minutes 43 seconds:
People on the forum ask "where to dig" all the time. This is the best example how to dig and where to dig when you face such impossible case. Warranty sticker, chip looks almost intact, not too many traces of flux but there are some small stains of flux. Almost invisible and the chip which after deeper analysis is not going to work in this circuit. A damn mystery.

Please friends, who has Powercolor Rx 5700 Red Devil (NOT XT) with Molex 8+6 just please help

It is 302155 that need to be there. Here it doesn't show full chip, but you can compare part number that is visible with the one on picture of core drmoses https://www.guru3d.com/articles-pages/p ... iew,4.html

On another site they write "The PCB corresponds to that of the Red Devil XT – except for the power connection"
Thus, the same as 5700 XT

Take off the drmos then check again

I have been waiting for 302045 but actual 302155 will not work without enable signal anyway. PWM comes to the PWM pin from 81022. These are facts :-) I would understand missing the enable signal, but this is enable signal to the wrong pin, to the output of the overtemperature :-)
I will replace it, sure, when I get them. But what the damn mystery :-)

Bottom left 0.9V 302155 is properly connected and works. On the boardview from Gigabyte this is also properly connected but DISB# is on 31 so the reference designs had to be changed by manufacturers. These integrated power mosfets don't fit in place of others, always one leg is different.

The history of this card is unknown, purchased cheaply for 200 euro or so. I had cards from consumer returns as it is in EU, such that they had seals but inside, the board was without components, without memories or even GPU. This card is suspicious because is clean inside.

---

Posted after 16 minutes 49 seconds:

This drmos looks like taken off :-) It is disabled :-) The power comes from the right MVDD until bigger load.

To be honest, i believe that NCP302045's datasheet contains a mistake at DISB part. It explains everything mysterious. For example, i saw Palit RTX 2060 with 045, 155, SICs and AOZ5332QI at same design boards.

support BrasilBack
here is a piece of boardview and bom file from which it is clear that 31 pin is an enable.

No, It can't be Because the right 302045, from my PCB and my measurements, beeps at pin 30. 0.9V 302155 beeps at pin 31 The same with the left MVDD, it beeps at 30.
0.9V has it at pin 31 as expected and it works.
Right also works with enable on pin 30.

I thought at first, half the memory chips only from right, other hallf from left. Of course not.

Left side appears to be intentionally left disconnected, because powering single rail from 2 separate PWMs can be complicated.

The problem is likely with right drmos or even elsewhere. If right drmos is fine, you can try take if off and put instead of left 302155 but dont try to run with both 302045


Is it the case or not?

No No No forget it. It was initial thought (I can't edit the post) and I found it is controlled by NCP81022. All three Power stages are controlled by it. It works in 2+1 mode. 2 for the memory and 1 for 0.9V.

Please look at the last picture.

Since the left memory power does not start and it does not look damaged in any way I started to dig. Then I discovered this discrepancy in the ENABLE signal to the mosfets. 0.9V has it in place at 31 as the NCP302155 requires, the right memory pin 30 302045 as required but the left will not work without ENABLE signal to the 31 pin. It gets it on 30.
The left is 302155 with ENABLE to the pin 30 !!! So I suspect that there should be NCP302045.

According to the PDF of both power stages they have swapped pins 30 and 31.

I understand that my theory is crazy but I have seen many fake warranty stickers and boards after repairs inside, literally devastated inside.
So we need somebody with exact same card to confirm and ring this DISB# (ENABLE) trace.

Update:

I got this image from a friend. It is Powercolor 5700 probably XT but please look at the left memory VRM. 302045!!!! I did not measure it yet, I don't have it but it is very suspicious. At the bottom side there is a second NCP81022 for the core voltage. Simpler construction just in case of failure.

This example card is not detected and I don't have idea yet what is wrong with it but I have evidence that Powercolor shuffles these power transistors.

I have forgotten to write. I repaired this card and it was indeed the wrong chip soldered by someone. It was so perfectly soldered that it was hard to notice. As I expected it was the wrong chip !!!

Ваша настойчивость и уверенность в себе помогла разоблачить преступника. Поздравляю

It was repair attempt by someone with skills and equipment and a n ultrasonic cleaner. He put warranty stickers. I had some 5700 cards from this first post and every manufacturer uses a different set of memory power chips. They may be the same or may be different, total randomness without a reason.

But this chip select, enable/disable should go to the proper pin.

Вы не один с такой проблемой. Похоже Powercolor выпустил бракованную партию видеокарт где в левом канале вместо ncp302045 была установлена ncp302155.
ну и память питалась от одного канала. Вот у меня сейчас такая же видеокарта на руках.

ncp302155 на этой видюхе отлично работает, он кстати и мощнее.
Эти дрмосы отличаются сигналами THWn и DSBL, они на разных ногах, THWn на этой видюхе не используется, он висит в воздухе
и поэтому остается только перекинуть DSBL с 30й ноги на 31. С Для этого нужно просто по одному нулевику переставить.
место для этого есть.
Не знаю как будут работать вместе ncp302045 и ncp302155 ? на моей видюхе ncp302045 сдох. Не вынес мук что он один за двоих работал )))
Я поставил вместо него ncp302155 передвинул нулевики и оба канала нормально заработали.

Did you fully repair this card after this operation? Is it functional? So you say that it worked on one MEM power phase for some time in someone's computer and failed? It is almost impossible, I thought that someone did the wrong repair and returned the card to the customer.

The history of my card was unknown because a customer brought it as a not working, second hand purchase.

Блин тоже попался, увидел тему уже после того как разобрался в проблеме. Доставил отсутствующий R4231.

Tried that, flickering screen and some artifacts appeared in windows, but driver installed good, pulses are there too.