XFX Amd Radeon VII, No Display, vBios flash doesn't work and Code 43...

[mlladen996]

@eidairaman1 @XFXSupport
Hi guys, so yesterday i opened up the card and did some testing to see if there is anything obvious that may be causing the card to malfunction.
So what i found is that the gpu chip is getting hot when the pc turns on and i checked the vrms and they do have voltage as you can see on the photos i attached. So i do not know where to look next for the problem.

 

[eidairaman1]

Try Contacting zettabit.lab or @buildzoid from actually hardcore overclocking on youtube or request repair from @XFXSupport (pay them all for their service)

 

[heyhant]

@eidairaman1 @XFXSupport
Hi guys, so yesterday i opened up the card and did some testing to see if there is anything obvious that may be causing the card to malfunction.
So what i found is that the gpu chip is getting hot when the pc turns on and i checked the vrms and they do have voltage as you can see on the photos i attached. So i do not know where to look next for the problem.

Did you find any solution? I have a sapphire one with the same problem! Thank you

 

[mlladen996]

Hey @heyhant I couldnt find the issue with the card and i sold it.

 

[eidairaman1]

Did you find any solution? I have a sapphire one with the same problem! Thank you

Contact sapphire

 

[heyhant]

No luck there! No more sapphire products for me! Now, for the pcb troubleshooting it might be a short in some part on the pcb. Are there any suggestions? I concluded that hynix memory radeon viis are unreliable and low quality cards and even if you treat them with special care they will fail quickly. So frustrating

 

[arkyson]

Hello everyone,
thank you for this tread...
I have the same exact issue as @mlladen996 describes in his initial post.

I have a Powercolor Radeon VII (also second hand)

GPUz shows the exact same, LITERALLY IDENTICAL (ie, finds card, bios version, ROPs/TMUs, shaders...but pixel/texture fillrate unknown, memory size 0mg, clock speeds 0mhz)
I can add that on GPUZ v2.33....it additionally shows HBM2 (Micron) ....as oppose to just HBM2 as in @mlladen996 initial post

Also getting Error code 43 in windows (when i put an older nvidia card in PCIe slot 1)
I have tried all the DDU/driver updates etc as described above...
I have also tried @eidairaman1 ATIflash methods. The card is recognized by both the windows & dos versions of ATIflash, and has accepted very bios I have tried (all from this website). The card came with bios v105, and its the same story when I flash v106

With just the Radeon Vii in PCIe slot 1, I can not get any video output...
fan & lights on the video card come on...

Lenovo D30 Workstation (type 4223)...is the computer/motherboard I am trying to install this card in...
it has a Bios POST code LCD built into the motherboard --> shows system hangs on POST code B2...

from what I've googled, post code B2 usually represents VGA check failed...and interestingly when I put my other older nvidia card in, I can see the motherboard cycling throught the post codes, and momentarly pauses on B2, then the screen come on and the computer continues through various other post codes until it settles in AA when working...

I've also checked out Eli Tech on youtube (he's awesome) and did the same as @mlladen996 , opened the card up....
the main gpu chip does get hot, and I can verify i get the same voltages on the various phases and chips... so the board seems okay (?)

I called AMD... they mentioned some workstation mainboards don't work well with AMD chipsets and suggested I ask Lenovo (waiting on a response...)

In the Bios I noticed various PCI settings:
- PCI latency timer
- VGA palette snoop
- PERR# generation
- SERR# generation
- PCI Express Settings --> ASPM support, Link Training Retry, Link Training Timeout (uS), and unpopulated links...

Wondering if any of these might have a role??

Also going to try to take the card to a repair shot or best buy to see if I can test it on a newer, known to be AMD compatible, motherboard....

If anyone has any tips or suggestions please do respond...

 

[Caring1]

Lenovo tend to white list hardware so certain things won't work.

 

[arkyson]

Lenovo tend to white list hardware so certain things won't work.

seriously?!?... these grimmy koonts go out of there way to whitelist/blacklist certain hardware?!?...what utter BASTARDS!!

thanks @Caring1, i'm gonna call Lenovo and chew them out...i'll tell them I'm selling all my lenovo crap over this and making a youtube video exposing this nonsense

I do see some posts out there about modding vbios to bypass whitelists...anyone know if this has been done before on the Radeon VII (can't seem to find much out there on the Radeon VII, but lots for AMD RX's)

thanks again

 

[roller5435]

Hi,

Did anyone find a solution to the Code 43 condition for the Radeon VII. My Radeon VII, after a reboot, is now exhibiting this condition and I'm hoping that all is not lost. Any guidance/solutions that have been discovered would be greatly appreciated and it appears after searching the internet that there are others that are experiencing this issue with the Radeon VII. Thank you.

 

[jesdals]

On a Skylake platform I had to enable cpu iGpu to boot proberly

 

[dialab]

As i have made some tests with Radeon VII code 43 cards and also removed some details. For me it seems that AMD failed with mcu layout this time. It comes open from center in pcb and that cause code 43 in most cases. Seems also logic now why they have stoped produce it. Will post some pictures soon, where can see why it is so.

Here can be seen how curved mcu itself it is. If we add BGA ball in center with same dimension then in edge, then there will be almost 0,3mm distance of air. So center ball must be bigger, then it probably will have contact.

 

[unimagine0124]

I bought 2 Radeon VIi from seller in China for about 10% lower cost than brand new market price about USD580 each. 1 of them posted and working on an older board with i7 3770K and running with Corsair TM-650W with no issues for 1 month plus and once I swapped with the 2nd unit I purchased, things got messy. 2nd unit did not post right away and have to go thru just about everything except replacing out the main chips which is not possible as I think there no such chips and may need a specialized BGA rework machine too.
Plug in an RX580 at another PCIE slot and booted with no issues. Windows detected but with error code 43 for Radeon VII. Successfuly flashed latest Asrock UEFI BIOS but same code 43. Windows updated to latest and also several fresh installs, not helped. After swapped the 2 unit with the single working Radeon VII, now both units got code 43.
Does anyone find a fix for these ?

Bought these exactly 1year ago, but just could not find anything on these during that time. There are more people getting these issues now but still no real fix for it. Recently tried brand new 2000W PSUs and several other motherboards but only not tried with latest 8th and 9th generation Intel and AMD motherboards to see if it's compatibility issues or just plain GPU hardware issues. Anybody successfully repaired these?

 

[Vlad62]

Hallo, have EXACTLY the same problem with my Sapphire Radeon VII card!
I successfully re-flashed it, but it still shows the error code 43 in Device Manager and GPU Z shows memory size 0 and GPU clock 0.

mlladen996 did you succeeded in making the card working?​

 

[gK_hun]

As i have made some tests with Radeon VII code 43 cards and also removed some details. For me it seems that AMD failed with mcu layout this time. It comes open from center in pcb and that cause code 43 in most cases. Seems also logic now why they have stoped produce it. Will post some pictures soon, where can see why it is so.

Here can be seen how curved mcu itself it is. If we add BGA ball in center with same dimension then in edge, then there will be almost 0,3mm distance of air. So center ball must be bigger, then it probably will have contact.

Hi!

So i would like to join the club with a card what does the same. I think it happened when i shut the machine off, and it got some cold air, and the chip must have been curved like on the pictures.

@dialab do you think it can be re-balled? Or is it impossible? I'm not an expert in re-balling, but if it is possible, i would like to repair it, i love this card

Thank you in advance

 

[dialab]

Hi!

So i would like to join the club with a card what does the same. I think it happened when i shut the machine off, and it got some cold air, and the chip must have been curved like on the pictures.

@dialab do you think it can be re-balled? Or is it impossible? I'm not an expert in re-balling, but if it is possible, i would like to repair it, i love this card

Thank you in advance

It can be reballed, but currently it seems handwork with very big time waste as there are also capacitors below mcu and no stencil for this mcu yet (havent found, but maybe exist).
I wonder people who buy right now such cards with price over 500€ and much more

 

[delshay]

You can place the solder balls by hand with tweezers. It's a pain but it can be done. I soldered in A CPU (BGA) in complete manually mode using my fully digital soldering dispenser. This was more an experiment as I wanted to push my soldering skills to the limit & it worked. Would not try it on a Vega VII chip as it needs to much patient, but placing the balls by hand I recommend doing this over a few days & not in one go if you are unable to find the stencils.

You're other alternative is "custom preforms" which is even better, but is very expensive.

As i have made some tests with Radeon VII code 43 cards and also removed some details. For me it seems that AMD failed with mcu layout this time. It comes open from center in pcb and that cause code 43 in most cases. Seems also logic now why they have stoped produce it. Will post some pictures soon, where can see why it is so.

Here can be seen how curved mcu itself it is. If we add BGA ball in center with same dimension then in edge, then there will be almost 0,3mm distance of air. So center ball must be bigger, then it probably will have contact.

Never seen capacitors on the same side as the solder ball before, I'm impress, but it looks like a knightmare to solder-in due to where the capacitors are located.

 

[Vlad62]

It can be reballed, but currently it seems handwork with very big time waste as there are also capacitors below mcu and no stencil for this mcu yet (havent found, but maybe exist).
I wonder people who buy right now such cards with price over 500€ and much more

In my personal case I was sure, that I would get a warranty repair or replace and was very surprised, that it is not possible... I had many cases, when the manufacturer replaced the broken devices (sometimes even with the newer models).
What is also interesting, is that many people write they got error 43 after swapping the card or so. How this could be connected with the theory of defective soldering?
And if defective soldering is a case, why would not the firms just replace these cards (especially in cases like mine) with the newer models like rx 5700 or rx 6700? This would be a great advertising and relief for the customers, who are not to blame for AMD defective manufacturing process...

 

[dialab]

In my personal case I was sure, that I would get a warranty repair or replace and was very surprised, that it is not possible... I had many cases, when the manufacturer replaced the broken devices (sometimes even with the newer models).
What is also interesting, is that many people write they got error 43 after swapping the card or so. How this could be connected with the theory of defective soldering?
And if defective soldering is a case, why would not the firms just replace these cards (especially in cases like mine) with the newer models like rx 5700 or rx 6700? This would be a great advertising and relief for the customers, who are not to blame for AMD defective manufacturing process...

For 5700xt they even cannot find replace, so how they should do it for Radeon VII?

 

[Vlad62]

For 5700xt they even cannot find replace, so how they should do it for Radeon VII?

If they have a shortage of new cards, they could open or rent a service center to repair the broken ones and reball all Radeon VII professionally. Why not?

 

[dialab]

If they have a shortage of new cards, they could open or rent a service center to repair the broken ones and reball all Radeon VII professionally. Why not?

First of all will be nice if they share some info, but so far nothing. They just stop produce it and thats it. Ofcourse if they able to sell parts like mcu itself, there many cards can be fixed.

 

[Vlad62]

First of all will be nice if they share some info, but so far nothing. They just stop produce it and thats it. Ofcourse if they able to sell parts like mcu itself, there many cards can be fixed.

They suppose to have some amount of parts or whole cards to replace the defective ones, right? And the defective ones also could provide some amount of parts (some might have problem with memory, some with mcu, some with something else). Why do the just reject any service - it is a puzzle for me...

 

[dialab]

They suppose to have some amount of parts or whole cards to replace the defective ones, right? And the defective ones also could provide some amount of parts (some might have problem with memory, some with mcu, some with something else). Why do the just reject any service - it is a puzzle for me...

Radeon VII have all in one, MCU contains also memory. There no separate memorys like 5700 etc.

 

[Vlad62]

Radeon VII have all in one, MCU contains also memory. There no separate memorys like 5700 etc.

If so, how would you as a specialist evaluate the possibility of repairing MCUs of Radeon VII cards you have with error 43, if you have necessary info and equipment?
And one more question. How would you explain the fact, that GPU Z shows 0 memory and 0 GPU Clock for Radeon VII? Does it necessarily mean that MCU is defective or it can be another reason also for such data?

 

[dialab]

If so, how would you as a specialist evaluate the possibility of repairing MCUs of Radeon VII cards you have with error 43, if you have necessary info and equipment?
And one more question. How would you explain the fact, that GPU Z shows 0 memory and 0 GPU Clock for Radeon VII? Does it necessarily mean that MCU is defective or it can be another reason also for such data?

This means Your Windows driver not loaded correctly for some reason. It may cause by several things at GPU side, even if some power is out of range or any other hw issue in GPU side.