Ремонт: Ноутбуков, Компьютеров
https://vlab.su/

Asus RX 470 4 GB Mining - блок питания невероятно горячий
https://vlab.su/viewtopic.php?f=48&t=122172		 Страница 1 из 1
Pandemonium6 [ 07 янв 2022, 08:24 ]
Заголовок сообщения:  Asus RX 470 4 GB Mining - блок питания невероятно горячий
Привет дорогие товарищи :-)

This article is about this card:
https://www.techpowerup.com/vgabios/213986/213986

The 4 phases of the core voltage are incredibly hot!. I can say 100'C, maybe more. PWM controller works in a strange way, it turns outputs almost at the same time.
Digi+ ASP1106 does not look as already replaced by someone before.

This card boots to Windows, drivers are accepted but it goes to the black screen or image flickering when 3D is run.

Channels on the oscilloscope 1,2,3,4 from the top choke to the bottom choke.

I spotted now, during writing this post that BIOS is for Elpida but I have Hynix... It was another bios inside and this overheating was present anyway. I found unverified, but original bios for Hynix.

Please help friends :-)
mia [ 07 янв 2022, 19:15 ]
Заголовок сообщения:  Re: Asus RX 470 4 GB Mining - блок питания невероятно горячий
Механические повреждения, трещины дорожек ?
Pandemonium6 [ 08 янв 2022, 03:17 ]
Заголовок сообщения:  Re: Asus RX 470 4 GB Mining - блок питания невероятно горячий
I don't see anything suspicious, even under the microscope. Power transistors work along with their gate drivers as you can see on the scope. The problem comes from the brain, from the Digi+ ASP1106.

People say this chip is similar to the RT8877 or RT3667BE. There are PDF manuals for them. There is one point called "Quick Response":

Код: Выделить всё
Quick Response
The VDD controller utilizes a quick response feature to
support heavy load current demand during instantaneous
load transient. The VDD controller monitors the current of
the VVDD_SENSE, and this current is mirrored to internal
quick response circuit. At steady state, this mirrored
current will not trigger a quick response. When the
VVDD_SENSE voltage drops abruptly due to load apply
transient, the mirrored current flowing into quick response
circuit will also increase instantaneously.

Whatever it is, it may add pulses but I have never seen such behavior before so I posted images from my oscilloscope. Maybe someone have seen at least double pulses on GPU VRMs.

It looks like a PWM pumps inrush power in response to the big load which is not present. The whole board is quiet, rather cool on the thermal camera. GPU is only warm with such a small radiator as seen on one image above. I always put this small radiator wit a small fan on top of the crystal and it is warm under windows. I don't stress test it like that of course but Windows desktop and tserver work for memory testing.

Do we have real documentation for ASP1106 ?
Roberts3333 [ 08 янв 2022, 12:38 ]
Заголовок сообщения:  Re: Asus RX 470 4 GB Mining - блок питания невероятно горячий
Pandemonium6 писал(а):
Do we have real documentation for ASP1106 ?

ASP1106 = ISL95712
mia [ 08 янв 2022, 12:51 ]
Заголовок сообщения:  Re: Asus RX 470 4 GB Mining - блок питания невероятно горячий
Try contact to aleksros...http://vlab.su/viewtopic.php?f=48&t=87343
Pandemonium6 [ 08 янв 2022, 15:52 ]
Заголовок сообщения:  Re: Asus RX 470 4 GB Mining - блок питания невероятно горячий
Roberts3333 писал(а):
ASP1106 = ISL95712

Thank you for this tip.

It means that PWM, in my case, works constantly in "FIGURE 7. R3™ MODULATOR OPERATION PRINCIPLES IN LOAD
INSERTION RESPONSE" https://www.renesas.com/us/en/document/ ... -datasheet page11.

So I have possibility that there is something wrong with conditioning circuit before COMP, FB, VSEN pins or the chip got crazy. I have to make some measurements.
Pandemonium6 [ 09 янв 2022, 13:01 ]
Заголовок сообщения:  Re: Asus RX 470 4 GB Mining - блок питания невероятно горячий
Finally I repaired it. It was really the behavior of this "LOAD INSERTION RESPONSE" or "Quick Response" circuit. I reflowed all components around PWM along with the DIGI+ chip itself and no reaction, no change.
Finally after days of contemplation I looked deeper by the microscope, from an angle and it looked like one or two pins marked on the image below did not have enough tin. I soldered all suspiciously looking joints an all four sides.

Now the VRM is arctic cold and GPU does not go black under load and of course pulses on the chokes are as expected 0, 90, 180 and 270 degrees. So remember to use two channels of the oscilloscope when you have instability problems. You must ensure that PWM gives the proper pulses and they are shifted properly.

Thank you everybody who wanted to help :-)
Posted after 9 minutes 10 seconds:
mia писал(а):
Mechanical damage, cracked tracks?

In general it was the closest answer but these pins are so small that I did not expect it :-)
Posted after 1 minute 15 seconds:
Roberts3333 писал(а):
ASP1106 = ISL95712

This was extremely helpful because the timing diagram pointed me to the "LOAD INSERTION RESPONSE" circuit.
Страница 1 из 1 Часовой пояс: UTC + 4 часа
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group
http://www.phpbb.com/