Best Graphics Card Repair Business in the United States?

[HeyItsChris]

Best Graphics Card Repair Business in the United States?  Preferably in Missouri but anywhere is okay as long as prices aren't too high & quality of work is high.

 

I need to have my Powercolor Red Dragon AMD Radeon RX Vega 56 looked at.  It gets hot, fans spin but no post no detection.

 

Thanks

 

PS - Do any of you guys know how to repair GPU's?

[Avacado]

Northwest Repair is the only one I know off

[HeyItsChris]

Where are they located, may you leave me more details?  Thanks

[Avacado]

23 minutes ago, HeyItsChris said:

Where are they located, may you leave me more details?  Thanks

https://www.youtube.com/channel/UCvDDbj3qJYg-5FSPFWJxqMw

Edited January 12 by Avacado

[pio]

3 hours ago, Avacado said:

https://www.youtube.com/channel/UCvDDbj3qJYg-5FSPFWJxqMw

Yup, Northwest Repair can get it done for you.

I'll be honest, you're probably going to spend more to repair it than you would buying a used Vega56 maybe.  HOWEVER, you'd have a fully functional card by repairing, NWR is really good at what he does.  So..........good luck?  Could search around on ebay and see if you can score a MI25 assuming @LabRat hasn't bought them all yet.  

[LabRat]

On 12/01/2024 at 14:01, pioneerisloud said:

Yup, Northwest Repair can get it done for you.

I'll be honest, you're probably going to spend more to repair it than you would buying a used Vega56 maybe.  HOWEVER, you'd have a fully functional card by repairing, NWR is really good at what he does.  So..........good luck?  Could search around on ebay and see if you can score a MI25 assuming @LabRat hasn't bought them all yet.  

Yes... I accidentally influenced the market buying my 6 MI25s.

I paid ~$40-55 ea.

 

Think of the MI25 flashed into a WX9100 as a Vega FE 16GB w/ a single* mDP 1.4(a?) output and a couple less power phases.

I use RdN.ID's MPT workaround driver for finite p-state control using OverDrivenTool. 

 

*I have confirmed that an MST hub allows multiple display outputs (no freesync, tho). You can also pass-thru the output over PCIE to a low-end Navi 2x or 3x card, for AFMF in Win11(easier to assign raster-rendering card)

 

Given time, they might get back to ~$50ea, and are some of the 'beefiest' Vega 10 cards left in the secondary market.

(All MI25s come with 16GiB of Samsung HBM2.

Every example I've had does 1050mhz-1100mhz+ on the HBM.)

 

Edited January 14 by LabRat

[HeyItsChris]

So the card I thought was bad actually started working sort of.  It sort of works but has issues, anyone know why?  Powercolor Red Dragon Vega 56 on an eGPU setup hooked up to my amd ryzen 3550h mini pc with the Oculink EXP GDC BEAST v10.1 to m.2 nvme slot working well.  Just need to know why its not working lol

Edited January 26 by HeyItsChris

[LabRat]

17 minutes ago, HeyItsChris said:

So the card I thought was bad actually started working sort of.  It sort of works but has issues, anyone know why?  Powercolor Red Dragon Vega 56 on an eGPU setup hooked up to my amd ryzen 3550h mini pc with the Oculink EXP GDC BEAST v10.1 to m.2 nvme slot working well.  Just need to know why its not working lol

Sounds like it finally got out of its safemode or w/e. That's good. 

I've had similar issue(s) w/ a maxcloudon x4 riser, that I was powering w/ a seperate 1200W server PSU (mining surplus parts):
It was a 'PITA' to get recognized+installed @ first, and then was 'hiccuping' and 'stuttering' at a very-precisely repeating interval.

I abandoned that project but in retrospect, I've since ran across a few factors that you might look@/address:

1. AmernimeZone/RdN.ID drivers(Specifically for Vega 10 cards) 
offer specific configs in the wizard for "eGPU". If there were not registry/low-level settings that needed 'tweaked' for higher-latency and higher-noise eGPUs, they wouldn't be in the wizard. (I highly recc. trying them out, in troubleshooting this issue)
2. Ground Loops, Uneven ground planes and Inducted Interference.
Even connected to the same power supply, you may have differences in resistance/potentials to ground, if only from long(er) cables (read: antennas). This may/will impact PCIe's signal stability and readability (for lack of a better term).

3. Looks like you may not be alone, with that PARTICULAR eGPU Riser.

Stability problem with EXP GDC Beast V9.0B M.2 and dual HDMI

EGPU.IO

---

Hi folks, A few weeks ago I received my beast with M.2 for an exernal GPU setup. I paid about 90 euros for it and I was very excited to give it a tr...

 

Edited January 26 by LabRat

[HeyItsChris]

Mine uses Oculink, not the dual hdmi version.  Mines the newest one, the v10.1 released like less than 30 days ago.

 

I tried NimeZ drivers.  I hope together we can figure this out.

[pio]

26 minutes ago, HeyItsChris said:

Mine uses Oculink, not the dual hdmi version.  Mines the newest one, the v10.1 released like less than 30 days ago.

 

I tried NimeZ drivers.  I hope together we can figure this out.

 

People in the thread Labrat linked to earlier recommended this extension:
https://www.aliexpress.com/i/2255799918873144.html?gatewayAdapt=4itemAdapt

Have you tried the GPU in a different machine yet to rule the riser out completely?  Have you tried other risers?  The card is being recognized, kinda.  You can also check the windows error reporting I believe for any PCIe errors?  If there's PCIe errors, its the riser probably.

Edited January 26 by pioneerisloud

[LabRat]

39 minutes ago, HeyItsChris said:

Mine uses Oculink, not the dual hdmi version.  Mines the newest one, the v10.1 released like less than 30 days ago.

 

I tried NimeZ drivers.  I hope together we can figure this out.

There's now two diff versions of the installer, w/ similar functionality


Make sure you're using the FlexArch Option


If using the CLI, I'd be trying the flexarch "eGPU" option


Then, select [10] (failing that, [3]).

(select your Vega card, and any other devices Snappy Driver Installer says are avail)

This is as far as I can go w/o altering my currently working 6500XT+MI25/WX9100 setup.
Hopefully, that'll've gotten you in the right direction, at least.

 

  

12 minutes ago, pioneerisloud said:

 

People in the thread Labrat linked to earlier recommended this extension:
https://www.aliexpress.com/i/2255799918873144.html?gatewayAdapt=4itemAdapt

Have you tried the GPU in a different machine yet to rule the riser out completely?  Have you tried other risers?  The card is being recognized, kinda.  You can also check the windows error reporting I believe for any PCIe errors?  If there's PCIe errors, its the riser probably.

I think his issue *MIGHT* be related to the fact his miniPC already has integrated graphics. That's what the "FlexArch" option(s) in RdN.ID drivers are for. 

Which, reminds me:
His card may already be 'working' but Windows 10 will not render to it, w/o the application itself being configured to address and render on the card.

I'm not a big fan of Windows 11, but it works 100% smoothly and user-configurable for switching between and rendering-across GPUs.

 

Edited January 26 by LabRat

[HeyItsChris]

Thanks well i did test in another 2 systems no post no detection nothing so i let it sit & i tried it on the egpu setup and its sort of working so maybe it had to completely dry out?  i got it wet lol

 

So now ill try flexarch drivers.

Edited January 26 by HeyItsChris

[pio]

2 minutes ago, HeyItsChris said:

i got it wet lol

 

[LabRat]

6 minutes ago, HeyItsChris said:

maybe it had to completely dry out?  i got it wet lol

 

What kind of wet? Coolant-filled loop, soda, rain? lol

Asking, as diff liquids will 'solvent' and transport dust under board components,
and/or leave residues behind that can interfere with function in odd ways.

IIRC, the gen. rule of thumb for a LC leak-damaged card, is a full bath in 91% isopropyl alcohol, and then actively drying for a day, or passively drying for 3-5+ days (depending on temps and humdity)

Edited January 26 by LabRat

[pio]

4 minutes ago, LabRat said:

IIRC, the gen. rule of thumb for a LC leak-damaged card, is a full bath in 91% isopropyl alcohol, and then actively drying for a day, or passively drying for 3-5+ days (depending on temps and humdity)

 

This would be a fantastic start if you hadn't done so already.  Like, immediately, before you hurt the poor Vega.    Even if its just water, just water WILL evaporate just fine yes.  It's, like Labrat mentioned, moreso about any dust or literally anything else that the water could've drug in with it.  Get the card 110% clean, 120% dry, and put back together and repasted / repadded.  And then give it a go.  And firstly, I'd say try it in a regular PC / bench setup whatever without the riser first.  Make sure the card itself actually is working good first before you try it in the risered extension.  Process of elimination.  

[HeyItsChris]

I just dont have a rig laying around in the place that im at currently to test.  

 

This is what I got so far ... errors with flexarch 23.9.2

 

Edited January 26 by HeyItsChris

[HeyItsChris]

It had a problem starting...

[LabRat]

16 minutes ago, HeyItsChris said:

It had a problem starting...

Did you use DDU to clean up all the GPU drivers, first (in safe mode)? I've gotten that error on:
1. Corrupt Drivers / Bad in-place driver install/update
2. Forced wrong driver
3. Bad vBIOS flash

4. an actually damaged card (ESD and/or phys damage, in my experience)

You'll have to rely on the Msft basic display device driver for your APU; it needs removed too.
I recc. you use the GUI installer, it allows you to select both APU and eGPU driver, concurrently.

Edited January 26 by LabRat

[HeyItsChris]

Okay ill test it again & as far as how i got it wet with water just trying to clean the thing it got really dusty in friends pc.  Now im testing rx480 and its working fine so there is definitely an issue with the vega 56.  Should I test it again?

[HeyItsChris]

Testing it again, the vega 56 powercolor red dragon.  Here is what I got so far...  It can see the core temp & hotspot temp of the gpu core diode in gpuz and hwinfo sensors and a couple other things but not everything shows up.

[HeyItsChris]

It says it started in the event viewer but still no dice.

 

 

[LabRat]

20 hours ago, HeyItsChris said:

It says it started in the event viewer but still no dice.

 

 

I'd try a game or something 3D that can be told which GPU to render on. 

 

On my mixed-dGPU build, more than a few times I've had it just decide not to recognize the video-out on either the 6500XT or WX 9100, or I get 'device cannot start' on the 6500XT.

(Full Power Cycling usually fixes it, and I've yet to figure out what triggers it, and it rarely happens on Windows 11, oddly more often on 10)

 

IIRC, Doom: Eternal is a launch option added to the steam shortcut.

Other games often have an option in the game/config or a launch option too.

[HeyItsChris]

I guess I forgot to mention these pictures...

 

Edited January 27 by HeyItsChris

[LabRat]

32 minutes ago, HeyItsChris said:

I guess I forgot to mention these pictures...

 

Yup. That might've been important. 

 

 

There's either corrosion/debris under components and on traces/contacts.

or

 

In my (swiss cheese) experience, those artifacts are 'typical' of VRAM failure in non-HBM cards*.
Considering, the level of integration of the VRAM on Vega... Hope it's option A.

TBCH, I think you've got a 'parts card' and/or a 'wall hanger'.

*The pattern could be indicative of a video-output issue too, but the card not being properly recognized and having repeated issues even starting, makes me think VRAM/related.

 

Edited January 27 by LabRat

[HeyItsChris]

I think I have dried up soap in between pins still and possible under the what mPGA or BGA or whatever you call it?  The Silicon chip, underneath it where the pins are might be dirty from when I used hand soap to clean it the first time (what an idiot i am) but its a lesson learned.  I think I can still fix it just need the right stuff & a bit of time.