• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Broken rx 5700 xt red devil - technical help

Stretschi

New Member
Joined
Feb 23, 2021
Messages
4 (1.33/day)
Hello,

One of my 5700 xt red devil broke last week - bios is flashed for mining and because of issues with fans - was running fine 3 Month these settings (low power settings - no overheating).
The PC won´t start up - so identified 1 of 4 cards was causing error - placed it alone, placed it in another PC - could not install the driver by failing at 90% all time (differnet drivers) - bluescreen "mashiene thread exeption" -HW not identified as RX5700 XT - amdvbflash could not find the driver - got it back to the rig, pluged PCI-E on after win startup (driver on the system) - could start flashtool amdvbflash and interact:

asic Navi10 ?(Slave)? -flash SOC15 SPI - romsize 40000 - test fail (a running card: no "slave", different flash and romsize 100000)

fail Error 0FL01 when flashing original bios - ROM unlock worked

bios switch to original OC - same - flashing aborting with "Failed to read ROM" - tryed a lot of things - different verisions of amdvbflash (294 304), forced flash and so on.

Any ideas what to do with it? Earasing the Manipulated ROM by shorting the EEPROM - is it even possible in this state? I think the card is wrecked, but not because of my bios-update/mining on low power!? Wanted to flash it back to original BIOS... Have electrical skills and can open the GPU, would like to send it back to Powercolor because the card is like 6 month old - but I think they find the bios - so would like to repair it or get it replaced without risk - please help a bit - found one guy with similar issues whose leg of a 8-pin network chip/transistor was broken (no rx 5700 xt - another card!) - could perform repairs if anybody has informations how i could identify the problem or solve the warranty issue.

Thank you very much - maybe someone had a similar case.
 
Joined
Feb 23, 2021
Messages
6 (2.00/day)
Location
Latvia
System Name _build://brain/conscious/mind/analitycs/hi-tech/electronics/
Processor inDigON8840488
Motherboard ..//model/unknown/experimental_mk57/
Cooling ..//natural/male_mk63/
Memory :infinity:
Video Card(s) ..//retina/gen_IV/blue/model_281
Storage 10*10Gpx
Power Supply ..//bio-tech/converters/stage_V/model_937/
Mouse ..//bio-tech/limbs/arms/model_542/
Hello,

One of my 5700 xt red devil broke last week - bios is flashed for mining and because of issues with fans - was running fine 3 Month these settings (low power settings - no overheating).
The PC won´t start up - so identified 1 of 4 cards was causing error - placed it alone, placed it in another PC - could not install the driver by failing at 90% all time (differnet drivers) - bluescreen "mashiene thread exeption" -HW not identified as RX5700 XT - amdvbflash could not find the driver - got it back to the rig, pluged PCI-E on after win startup (driver on the system) - could start flashtool amdvbflash and interact:

asic Navi10 ?(Slave)? -flash SOC15 SPI - romsize 40000 - test fail (a running card: no "slave", different flash and romsize 100000)

fail Error 0FL01 when flashing original bios - ROM unlock worked

bios switch to original OC - same - flashing aborting with "Failed to read ROM" - tryed a lot of things - different verisions of amdvbflash (294 304), forced flash and so on.

Any ideas what to do with it? Earasing the Manipulated ROM by shorting the EEPROM - is it even possible in this state? I think the card is wrecked, but not because of my bios-update/mining on low power!? Wanted to flash it back to original BIOS... Have electrical skills and can open the GPU, would like to send it back to Powercolor because the card is like 6 month old - but I think they find the bios - so would like to repair it or get it replaced without risk - please help a bit - found one guy with similar issues whose leg of a 8-pin network chip/transistor was broken (no rx 5700 xt - another card!) - could perform repairs if anybody has informations how i could identify the problem or solve the warranty issue.

Thank you very much - maybe someone had a similar case.
Hey, man,

Am assuming you've wrecked the bios soc-8 chip, right?
Just google for a similar model in picture, look what type of ic is it and solder one back with the correct bios taken from the web - this should work for warranty if you clean-up PCB after soldering.

Anyway, if you have some more questions, just ask them.
 

Stretschi

New Member
Joined
Feb 23, 2021
Messages
4 (1.33/day)
Ok, assuming it is the Bios-EEPROM - both Chips (Dual) wrecked in one step is nearly impossible (there is a hardware-switch routing through 1-1 bios-chip - this switch should be burned if both chips got a short at once). So If the prob is on this electrical line, I think something burned on the way to the chips - could figure that out, but I would have to open the GPU and loosing warranty eighter way. For example: Short in the line before the switch - burnes one BIOS-EEPROM - I have switched the Bios - so a perma short to supply-voltage would have wrecked the second EEPROM too (you could see that visualy - thats not the case). If they are intelligent, this is not the way to develope a PCB - short to ground or OL are possible (all before the bios-switch).

I´m working in hardware developement - so a few educated guesses:

- If Powercolor really wants to find the manipulated Bios - they will, even if it is in structures with missing datasets. The question is always how deep they are doing defect analisys.
- There are so many measurement-contacts on the GPU itself, they have all possibilitys when a real !not! automated troubleshooting begins.
- 2 wrecked Bios-Chips in one step is very unlikely, if you don´t see a big fat burned area around them. Bond or manufacturing defekts are around 0.0XX % normally - if the chip manufacturer has some Probs on the line - max 1% - no way both chips defect at once.
- If I delete the EEPROM through a short on the Chip - (if they have good people) - they will understand, that it was a manipulation and deny the warranty because its not so easy to really delete EEPROMS through a defect - virtual or real EEPROM - no difference
- I will never change a chip on a automated-manufactured PCB without any traceability - they will look and see it ;) - even if my skills be reaaly good in soldering with good equipment. It would be better chance to just send it in and hoping for some "known" issue after the automated test (no further investigation)


Do I get the card back, when powercolor denies my warranty? Then I would try to send it in and repair the card afterwards when they send it back...?
I´m preatty sure the game is over with that card - could be lucky sending it in...
 
Joined
Feb 23, 2021
Messages
6 (2.00/day)
Location
Latvia
System Name _build://brain/conscious/mind/analitycs/hi-tech/electronics/
Processor inDigON8840488
Motherboard ..//model/unknown/experimental_mk57/
Cooling ..//natural/male_mk63/
Memory :infinity:
Video Card(s) ..//retina/gen_IV/blue/model_281
Storage 10*10Gpx
Power Supply ..//bio-tech/converters/stage_V/model_937/
Mouse ..//bio-tech/limbs/arms/model_542/
Ok, assuming it is the Bios-EEPROM - both Chips (Dual) wrecked in one step is nearly impossible (there is a hardware-switch routing through 1-1 bios-chip - this switch should be burned if both chips got a short at once). So If the prob is on this electrical line, I think something burned on the way to the chips - could figure that out, but I would have to open the GPU and loosing warranty eighter way. For example: Short in the line before the switch - burnes one BIOS-EEPROM - I have switched the Bios - so a perma short to supply-voltage would have wrecked the second EEPROM too (you could see that visualy - thats not the case). If they are intelligent, this is not the way to develope a PCB - short to ground or OL are possible (all before the bios-switch).

I´m working in hardware developement - so a few educated guesses:

- If Powercolor really wants to find the manipulated Bios - they will, even if it is in structures with missing datasets. The question is always how deep they are doing defect analisys.
- There are so many measurement-contacts on the GPU itself, they have all possibilitys when a real !not! automated troubleshooting begins.
- 2 wrecked Bios-Chips in one step is very unlikely, if you don´t see a big fat burned area around them. Bond or manufacturing defekts are around 0.0XX % normally - if the chip manufacturer has some Probs on the line - max 1% - no way both chips defect at once.
- If I delete the EEPROM through a short on the Chip - (if they have good people) - they will understand, that it was a manipulation and deny the warranty because its not so easy to really delete EEPROMS through a defect - virtual or real EEPROM - no difference
- I will never change a chip on a automated-manufactured PCB without any traceability - they will look and see it ;) - even if my skills be reaaly good in soldering with good equipment. It would be better chance to just send it in and hoping for some "known" issue after the automated test (no further investigation)


Do I get the card back, when powercolor denies my warranty? Then I would try to send it in and repair the card afterwards when they send it back...?
I´m preatty sure the game is over with that card - could be lucky sending it in...
So are you certain that any of the chips are burned, did you check for short to ground?
If the chip is dead then engineers would not even bother getting something off them - only in some pedantic true conspiracy working investigation on a particular matter...
The easiest way to try your luck is just to do to the seller and say that you or someone else didn't know that you cannot pull the switch on a working card.
 

Stretschi

New Member
Joined
Feb 23, 2021
Messages
4 (1.33/day)
So are you certain that any of the chips are burned, did you check for short to ground?
If the chip is dead then engineers would not even bother getting something off them - only in some pedantic true conspiracy working investigation on a particular matter...
The easiest way to try your luck is just to do to the seller and say that you or someone else didn't know that you cannot pull the switch on a working card.
!Without dismounting the backplate and the cooler!: All is fine. I have a clear look on the two separate Bios-chips and they are looking normal. No smell, no defect to see on the whole card.
I didn´t check anything else HW-related, because of the danger of loosing the warranty. I still see a chance of getting a new card.

I will try to send it in without deleting the EEPROM, because I think I will not be able to flash on that card - changing an EEPROM-chip is risky as well - I will just try it like it is. I used the original bios of the card and manipulated it, so let´s hope the best.

Thank you for the help, maybe someone else has another idea, but I think it´s a HW-defect.


Personal note:
The best thing is, that a lot of AMD GPU have probs on the nofan option or the fan controll - you can just fix that by editing the bios and loosing warranty :D. My cards just worked for more than 6 hours stable after deactivating the nofan option, deinstalling the adrenalin-driver and using Afterburner without overclocking-adrenalin and fancurves (Fans fix on 70% (mining) - solving problems but noisi). So DDU the driver and let Win install just the minimal driver, manipulating the Bios fan settings, etc.
So I´m no person using warranty-options even if I´m responsible for a defekt, but these AMD-cards are longterm not useable without bios editing - so whats the point in loosing warranty this way - just sucks! And imagine - which amount of the AMD GPU are manipulated because they are crashing 24/7 - 50%?! WTF - Im a bit pissed, I had like 30 GPUs in my life and none got a defect before they were like 3 years old... eighter none had such driver-issues.
 
Joined
Feb 23, 2021
Messages
6 (2.00/day)
Location
Latvia
System Name _build://brain/conscious/mind/analitycs/hi-tech/electronics/
Processor inDigON8840488
Motherboard ..//model/unknown/experimental_mk57/
Cooling ..//natural/male_mk63/
Memory :infinity:
Video Card(s) ..//retina/gen_IV/blue/model_281
Storage 10*10Gpx
Power Supply ..//bio-tech/converters/stage_V/model_937/
Mouse ..//bio-tech/limbs/arms/model_542/
!Without dismounting the backplate and the cooler!: All is fine. I have a clear look on the two separate Bios-chips and they are looking normal. No smell, no defect to see on the whole card.
I didn´t check anything else HW-related, because of the danger of loosing the warranty. I still see a chance of getting a new card.

I will try to send it in without deleting the EEPROM, because I think I will not be able to flash on that card - changing an EEPROM-chip is risky as well - I will just try it like it is. I used the original bios of the card and manipulated it, so let´s hope the best.

Thank you for the help, maybe someone else has another idea, but I think it´s a HW-defect.


Personal note:
The best thing is, that a lot of AMD GPU have probs on the nofan option or the fan controll - you can just fix that by editing the bios and loosing warranty :D. My cards just worked for more than 6 hours stable after deactivating the nofan option, deinstalling the adrenalin-driver and using Afterburner without overclocking-adrenalin and fancurves (Fans fix on 70% (mining) - solving problems but noisi). So DDU the driver and let Win install just the minimal driver, manipulating the Bios fan settings, etc.
So I´m no person using warranty-options even if I´m responsible for a defekt, but these AMD-cards are longterm not useable without bios editing - so whats the point in loosing warranty this way - just sucks! And imagine - which amount of the AMD GPU are manipulated because they are crashing 24/7 - 50%?! WTF - Im a bit pissed, I had like 30 GPUs in my life and none got a defect before they were like 3 years old... eighter none had such driver-issues.
Well in this case if they both are perfectly fine there is one other option - to re-program the chip without even soldering the thing... It is called external chip programming using some of EEPROM programming hardware and software combined, just need to find a way how to wrap thin wire around the legs(pins) on the chip and flash needed bios this way.

There are videos explaining step by step on youtube how to safely re-program and even under-volt for less consumption and more stability including options for fan curve or profile.
I mean if you take care of your stuff and don't let it run full throttle all the time then it will serve you longer, right?

- Comparing AMD Adrenaline drivers to Nvidia's old-school driver - that alone buys me totally to buy products from AMD, they show that they really care for their customers and constantly deliver every time a new product comes out for how-long now - since Zen/Polaris, right?

Thanks, Stretschi - for sharing your insights in this area!
 

Stretschi

New Member
Joined
Feb 23, 2021
Messages
4 (1.33/day)
Ok, thank you, just looked into it - maybe I´ll get an external programmer and try that, if that is not working, at least I know the chips are wrecked or not
 
Joined
Feb 23, 2021
Messages
6 (2.00/day)
Location
Latvia
System Name _build://brain/conscious/mind/analitycs/hi-tech/electronics/
Processor inDigON8840488
Motherboard ..//model/unknown/experimental_mk57/
Cooling ..//natural/male_mk63/
Memory :infinity:
Video Card(s) ..//retina/gen_IV/blue/model_281
Storage 10*10Gpx
Power Supply ..//bio-tech/converters/stage_V/model_937/
Mouse ..//bio-tech/limbs/arms/model_542/
Ok, thank you, just looked into it - maybe I´ll get an external programmer and try that, if that is not working, at least I know the chips are wrecked or not
You are welcome, my friend!
Let me know if this was somewhat useful info and would love to hear more from you about your progress in this issue you are facing.
 
Top