RX580 all voltages good but won't boot and fan won't spin

[debidoobs]

Hi,

Im actually trying to find what is going wrong with my sapphire Pulse RX580 8gb... I purchased it used and supposely dead a year ago and found out (thanks internet) that U1000 (And gate sot23-5) was not giving me the high state to the power management IC NCP8122...

At this time, with no spare parts, i decided to put a 1kohm resistor to create the voltage as commented by a guy named Lorkur on this forum (https://*******/viewtopic.php?f=48&t=122873), you ll need to translate it in english.

and voila ! the gpu worked again and i didn't care much since. But a week ago, my computer didn't start up.. so here we go again i dismantle the GPU and plugged it to my riser and my laboratory psu.... and checked all the voltages and everything is right as it should but the fan won't spin which means to me the gpu is not booting even though i do have my 0.9vcore... by the time i received the AND Gate ic i should have solder since the beginning.. but no improvement. In my case, Perstb_buf is missing at the PIN 2 of the AND GATE U1000.

And really strange is if i turn on my psu then touch the gpu chip, sometimes, fan will spin as the gpu is booting normally

So my question is do you think the NCP8122 can be the culprit even though my vcore rails are correct ? Perstb_buf on some schematic are coming from a U100 ic but this ic is nowhere to be found on the Sapphire Pulse

It's really strange because every test point mentionned on youtube are good..

Thanks for you help and your tips !

Debi

[прямо]

The 2 input AND gate tiny logic receives 1V PGOOD signal from the PEX PWM controller and PCIE RST signal from the mainboard via the PCIE slot lane, and the logic output signal is called PCIE RST GATED.

When this tiny logic IC is missing, you can bridge PCIE RST signal solder pad directly to the PCIE RST GATED pad with a resistor.

It has no direct correlation to the symptom you described.

Touching the core while the system is powering up sometimes will make the fan spins, align more on an issue with solder balls connection.

You can try to apply pressure to the core while powering up and see if you will get a picture on the monitor. If you do, then the solder balls are definitely cracked. Solution is to reflow or reball the core.

[debidoobs]

Regarding to the schematic, the logic and gate is as attached. In my case, 3.3v is at pin1 but nothing at pin 2 even though it happens i do have 3.3v at pin2 randomly.

May be you have the schematic where your informations come from to compare ?

About touching the gpu is only by touching slighty the metal case of the gpu chip not the center core. I do not push at all so it doesnt look like a reballing issue.

Regards

[прямо]

Your PERST_BUF signal on pin 2 should come from somewhere. Look it up in the schematic or a boardview.

[debidoobs]

Your PERST_BUF signal on pin 2 should come from somewhere. Look it up in the schematic or a boardview.

Hello,
unfortunatly i didn't findthe right schematic for a Sapphire pulse RX580 8gb, the only schematic i found is called "rx580V341" and it shows as the screenshot attached. On my board there is no U100 and also the PERST signal is coming from te PCIE from the motherboard ? That could explain many things but it does not explain why my gpu stopped working since i put the 1K resistor it booted up very find for few months... unless doing that damaged the NCP81022, but having my Vcore 900mv right is strange if the NCP was bad it should not i guess.

[прямо]

It's U4513 on the back. From this logic IC it outputs a buffered PERST signal to pin 2 of U1000.

Because you have VCORE, then 3.3V BUS and 1V8_AON are ok, so that leaves the PERST signal from the PCIE slot. This is what you need to check if it is present at U4513 input.

[debidoobs]

Hello, thank you for your help. After looking around with my multimeter, Perst signal is going from PCIE pin to the U4513 logic... Now it makes sens, the fact im using a riser make some voltages as the Perst signal not present when it would be present if i was using a true motherboard ?

In this case, i dont understand why in the first place i couldn t get the board botting up when connected to my pc motherboard... Unless U1000 or U4513 was faulty since the beginning.

Just tested u4513, he s fine. Now i really dont understand why by forcing high state from u1000 to the NCP81022 power management ic made the card working for months.. with the same set up (mobo + psu).. but now it s not booting up anymore when the card is plugged in. No short anywhere on the card and right voltages also...

[прямо]

If you are testing your card with a riser that is not connected to the PCIE slot of your mainboard, then yes, you don't have a PERST signal.

[debidoobs]

Hello ! just an update, so yes i had to jump solder the pcie riser to allow the PERST high state. So yes i fixed the card.... kind of cauz it boots normally with a lab psu and also inside the pc. Unfortunatly, the graphic card won't show any signal in the monitor even though fans are spinning et pc starts normally as test leds on the mobo process correctly. VGA led does not stay lit. So bad, im wondering if the bios is still in good condition after the few restart i made with the card alone during the troubleshooting. Monitor and hdmi are the same i used before when the rx580 was working.. i don't find forum talking about graphic card with the same symptoms..

[DynaxSC]

Maybe the hdmi cable, happens sometimes.

[прямо]

Hello ! just an update, so yes i had to jump solder the pcie riser to allow the PERST high state. So yes i fixed the card.... kind of cauz it boots normally with a lab psu and also inside the pc. Unfortunatly, the graphic card won't show any signal in the monitor even though fans are spinning et pc starts normally as test leds on the mobo process correctly. VGA led does not stay lit. So bad, im wondering if the bios is still in good condition after the few restart i made with the card alone during the troubleshooting. Monitor and hdmi are the same i used before when the rx580 was working.. i don't find forum talking about graphic card with the same symptoms..

Ok, so boot into Windows with the integrated graphics so you have a display and see if Windows detects the card.

If the card is detected, then you will see an error code pertaining to the no display symptom (code 43 most likely) which is usually caused by broken connections from VRAM to GPU or from GPU to VRAM.

Run nVidia MATS tool to check which VRAM channel has that problem, and let me know.

[прямо]

Why did I say nVidia MATS?

I meant to say, use AMD Tserver/AGT to test the memory.

[debidoobs]

Hello unfortunatly i dont have integrated graphics i know it's bad ! i try to figure out how get one quickly just to test. I let you know. Cheers

[debidoobs]

Hello,

i reopened the card today to check again some voltages cause since i repair a bad diode, the card boots (fan are spinning) inside the pc, but it will not show any pictures. And surprisingly, my vcore is 2.8v ! instead of 0.9v.... May be the gpu is dead ....