MSI B450/RX580 signal loss as soon as I launch any game (with GPU-Z logs)

[belouch31]

Hello guys,

I have been struggling for two full days on this and i'm begging for any help.

Issue:
I 'built' my first computer 1.5 year ago (see config below). It works smoothly for any non gaming stuff (even photo editing) and I actually never did game much on it. But now I finally had time to, and when i launch ANY game, the screen loses signal, and the whole computer is crashed, the CPU and GPU fans spin at about 30% and never stop until I manually force-shut down the computer. If I try any GPU testing tool (like UserbenchMark or FurMark), they detect the GPU, but as soon as I launch any test, I get the same issue (signal loss, fans spinning, computer crashed).
Note: I have no integrated GPU in my CPU and my desktop work goes smooth. I can even do photoediting in gimpshop at high resolution.

Configuration:

• Motherboard: MSI Tomahawk Mortar b450
• BIOS Version/Date: American Megatrends Inc. 1.00, 7/16/2018
• CPU: AMD Ryzen 5 2600
• Memory: 16GB GDDR5
• GPU: Sapphire Radeon RX 580
• VBIOS: 113-1E3870U-O49
• Driver: 26.20.15019.19000
• PSU: Seasonic M12 620W
• OS: Windows 10 Pro x64 v1909 (OS Build 18363.836)

What I tried:

• Updated GPU drivers (using AMD Radeon Adrenalin 2020 Edition)
• Updated MSI drivers (using MSI Live Update 6)
• I tried 2 different HDMI cables
• I tried 2 different screens-I tried to change the power limit to -30% or +30%
• I tuned the GPU to -50%
• I tuned the VRAM to its lowest possible value
• I tried launching games in 800x600 with lowest details possible
• I removed and reput the GPU in the Motherboard
• I replaced the PCIe cable from PSU to GPU with a new one
• I had issues with my Windows 10 Home which was stuck on v1803. Since everything I found on the internet didn't work for me, I thought it could be the reason, so I bought and installed a new Windows 10 Pro v1909 on another SSD. It did not fix the issue.

Note: I never did any overclocking on either CPU or GPU.

If anyone cares to help, I have 3 GPU-Z logs from three different tests (attached):
-launch a game low resolution
-launch a game high resolution
-launch a GPU stress test

-Also a screenshot of GPU-Z when I do photoediting at high resolution (no issue there)

I am litteraly begging for any help. Please keep in mind I am not super tech savvy I tried to provide all the relevant information but please let me know if I forgot anything.
Thank you so much for any help on this!
Belouch

 

[Regeneration]

Can you run games / 3D apps in a window (not full screen)? If you press on the [?] in GPU-Z near to Bus Interface what happens?

 

[belouch31]

Can you run games / 3D apps in a window (not full screen)? If you press on the [?] in GPU-Z near to Bus Interface what happens?

• Games in window mode: same issue unfortunately
• If i click on the [?] next to Bus interface, it proposes me to do a render test. As soon as I click OK, the same issue triggers
  
  Edit: I attached a screenshot of the GPU details on GPU-Z

 

[Regeneration]

There are 3 possibilities here:

1. Defective graphics card
2. Defective power supply
3. PCIe power cable is not plugged correctly.

Check if the PCIe power cable is plugged correctly on both ends.

Seasonic PSUs are considered the finest in the industry. So its most likely the graphics card.

Try it in another PC if possible just to be sure, and then contact the company responsible for warranty service / RMA.

 

[belouch31]

Thank you Regeneration.
Unfortunately it's not the 3rd one, since I've triple checked that and I even replaced the PCIe power cable between the PSU and GPU with a new one.
I'll try to see if I can find anyone to lend me a PC to stick my GPU in or to try with their PSU

 

[Regeneration]

If you have done it already, try to uninstall AMD drivers, perform a driver cleanup with DDU from safe mode, and then install the latest driver from AMD's website.

Probably not going to help since software problems will result in a error screen rater then lockups and black screens. This is most likely to be hardware issue.

 

[belouch31]

If you have done it already, try to uninstall AMD drivers, perform a driver cleanup with DDU from safe mode, and then install the latest driver from AMD's website.

Probably not going to help since software problems will result in a error screen rater then lockups and black screens. This is most likely to be hardware issue.

Thanks, I just tried that with DDU in safe mode and new drivers, but your gut feeling was right.
Unfortunately I couldn't find any other PSU to test, so i contacted the seller to do the warranty process but it will take ages.
In the meantime, i did more tests, and the more i do and the less it looks like a GPU issue (or even a PSU issue) to me (but I don't know much...).
For example, I did a very unconventional stress test today where I opened 10 high resolution files in photoshop and about 70 youtube videos plus a bunch of other stuff, while monitoring with GPU-Z.
No crash occured and the weird thing is I managed to get each CPU-Z monitored variable higher than the value I observe when the crash occurs.

Peak values of test 1 (many photoshop/youtube videos opened)
GPU Clock: 698.8
Memory clock: 1000.0
GPU Temperature: 47°C
GPU Load: several peaks including some at 100%
Memory controller load: several peaks including some at 100%
Memory used (Dedicated): 1250
Memory used (Dymanic): 119
GPU chip power draw: mostly stable at 28.4, several little peaks at 30-30.5 and then a few very punctual and very crazy values like 95232 or 4196645.5 (??)
RAM went up to 15800 aha (thanks chrome)

--> NO CRASH

Peak values of test 2 (any rendering test even a non-stressfull one)
GPU Clock: 300
Memory clock: 300
GPU Temperature: 33°C
GPU Load: 0
Memory controller load: 1%
Memory used (Dedicated): 201
Memory used (Dymanic): 17
GPU chip power draw: a flat 28.4

--> CRASH...


So the last two questions I have left before giving up entirely are:
-do these results tell you or give you any hint on anything?
-I have never flashed the BIOS of my motherboard so it is still the orginal one from when I bought it. Newer versions are available now. Is there any chance this issue could be caused by that? Should I take the risk of updating the BIOS?

Thanks a lot for any help

 

[HD64G]

Latest AMD chipset drivers and UEFI version for the MSI board could solve such problems IF the latest drivers for the GPU don't resolve the issue.

 

[holyprof]

Probably this is not the cause, but the GPU pulls 75W from the motherboard, so if the motherboard has power deliveyr problem, it can cause reboots or crashes. I had a problematic motherboard that worked ok with integrated GPU, but crashed or rebooted every time the GPU was asked to do something in 3D (i.e. opening a game or 3Dmark).
Tested with 3 different GPUs: Asus GT210, MSI GT720, EVGA GTX960 - all gave me the same problem, but worked ok in any other PC.
My conclusion was that it's a motherboard problem. Later i knew it really was, because the motherboard died 3 months later (the PC was working OK with its integrated GPU).

Edit: I also swapped the power supply and it didn't help.

 

[belouch31]

I tried with latest GPU drivers (actually some new ones were available today, it really got my hopes up aha) and also with latest AMD chipset drivers.

Holyprof: thanks for sharing, that means there is one more possible suspect component aha, not good for me...

I decided that before toying around with flashing the BIOS of the MB (I was told you can lose your warranty?) I will buy a used R9 390 GPU.
That way if when i switch them the R9 works, it will mean the issue was the RX580 so i'll send it to RMA and i'll have the R9 to use in the meantime.
If switching it doesn't solve it, I will probably just set the whole thing on fire

Thanks for the help guys, i'll let you know how it goes when I receive the R9...

 

[Caring1]

Flashing the Bios shouldn't void warranty on the Motherboard.
Update that, then try another X16 slot if your Motherboard has one, even an X8 will work fine.

 

[belouch31]

OK in a last desparate attempt I tried:
-Updating BIOS of the motherboard
-Deassembling and reassembling every single component of the computer
It did not fix it.
I wanted to try your suggestion also Caring1 but my MB only has one other PCI slot, i tried to put the GPU there but because of its size, it gets stuck by other stuff like the P+ P- Power/reset plugs and other stuff and i can't plug it in
Unless anyone has another idea that hasn't been mentioned i'll stop fiddling with it and wait for the R9 390 I bought to get here Thanks again all for your advices!

 

[holyprof]

Yeah, better stop messing with it. I expect that the R9 390 will help pinpoint the problem. Hopefully, it's a GPU problem, not motherboard.

 

[Regeneration]

If there are 0 problems with the PC except of when 3D clocks are engaged then the graphics card is just defective.

Something the problem is just simple like that. SeaSonic M12 series are known to be great PSUs.

Compatibility issue between VGA and motherboard will result in a failure to POST.

The card is detected, runs at PCIe 3.0 x16, it just cannot enter 3D clocks.

 

[belouch31]

Hey guys,
Just to give you some closure here. I received my R9 390, plugged it in, updated drivers and I can now run GPU stress tests and games
So indeed, it seems my RX-590 was dead.
Thanks again a huge lot for the help!

 

[TheOathMan]

Hey guys,
Just to give you some closure here. I received my R9 390, plugged it in, updated drivers and I can now run GPU stress tests and games
So indeed, it seems my RX-590 was dead.
Thanks again a huge lot for the help!

Hello, I came across your post while I'm doing some digging, because I encountered exactly the same issue with my used RX580 4g. The seller told me it's tested and it's like new. So I took his words for it.

However, if you still have your card that you deemed defective or dead. There's one last thing that you could try to fix it which I'm going to do very soon. Change the thermal grease with a quality one. Make sure it's applied and it's covering every spot. I feel like if a thermal grease isn't applied properly, and there are still spots that are not covered or still exposed, it could get overheated. The senser might not be able to detect it which will eventually cause crashes, blackscreen etc.. while getting perfectly normal gpu temperature reading.
This solution have solved at least one GPU I came across. I'm hoping this is gonna fix mine.
Good luck