[SOLVED] Weird GPUs Behaviour

Jan 26, 2021
5
1
15
Hi all!
I've have encountered a problem that I've never seen before nor I've never heard of, and I'm calling for help after days spent at troubleshooting this.

I have received three Asus Strix RX 570 4GB to test from a friend, and I begin to test them on my test bench which has the following components:

MOBO: ASRock B460M-ITX/ac
CPU: Intel i3 10100f
RAM: 2x 8GB G.Skill TridentZ RGB @ 3000Mhz (they run at stock 2166mhz)
SSD: Kingston M.2 A400 SSD 240GB
PSU: Cooler Master V750 - 750w
OS: Windows 10 Pro 64 bit

Before installing and testing the first RX 570, I've used DDU to remove any Nvidia drivers (I had GTX 1050 Ti on the test bench up 'til that moment), then switched off the PC, removed the GTX 1050 Ti, mounted the first RX 570 and turned on the PC.
Everything works great, video signal is there, so I proceed to install the latest drivers and run a stress test (heaven) for about 30min , and everything works fine.
I proceed to test the second RX 570, I remove the first and mount the second, turn on the PC, everything lights up but I got no video signal. I try both the DVI and the HDMI ports with different cables (not the DP since I don't have a DP cable) but still no signal.
I proceed then to test the third one, but the results are exactly as the second one.

So here my conclusion is simple: only the first card is working.

I give back the two non-working cards to my friend and later he tries them on his PC, and they BOTH work perfectly fine.
So he brings them again to me and we test them togheter on my test bench, but again the results are the same: only the first card works on my PC and the other two give me no video signal.

What I have tried so far:
  • Using different cables to exclude the problem there, and they all works fine on my other gaming rig or my TV.
  • I tried completely removing the drivers with DDU between the removal of the working and the installation of the non-working RX 570, and as a precaution I have removed also any Nvidia and Intel graphics drivers, but still the issue persisted.
  • I tried mounting again the GTX 1050 Ti (even without removing the AMD drivers) and everything works, I also tried mounting a GTX 1060 6GB and I still got video signal.

What I have noticed:
  • When I start the PC with one of the "non-working" RX 570 the fans seems to spin at 100% (both the GPU fans and also the stock Intel cooler fan) constantly, I'm not sure if they are at 100% because they are both really quiet
  • When the PC is on with one of these "non-working" RX 570 I can only turn it off by switching off the power supply directly, the power button doesn't seem to work. The reset button instead works
  • When I take off one of the "non-working" RX 570 and mount back the only one that works, I get video signal, but windows greets me with this message, after which I just click restart and the PC restart and then works perfectly fine.

This last point makes me think that it might be a windows 10 related issue? But I really don't know at this point.

Sorry for the long post, but I wanted to give as much information as possible.

Thanks in advance to whoever will help me!
 
Solution
Hello and welcome newcomer!

The message that windows greets you it's meant to happen after 3 failed attempts, so you can have additional options, like repairing windows, or booting to safe mode, going to BIOS, etc...

I would have come to the same conclusions with you with these results.

I would also check the following: Latest windows updates, BIOS, chipset drivers and also see if all 3 570s have the same vBIOS. It won't hurt if you also reset CMOS.
Another thing to try is to use a spare HDD/SSD by installing windows there and only having that plugged in. That will eliminate any windows issues if there are any.
Hello and welcome newcomer!

The message that windows greets you it's meant to happen after 3 failed attempts, so you can have additional options, like repairing windows, or booting to safe mode, going to BIOS, etc...

I would have come to the same conclusions with you with these results.

I would also check the following: Latest windows updates, BIOS, chipset drivers and also see if all 3 570s have the same vBIOS. It won't hurt if you also reset CMOS.
Another thing to try is to use a spare HDD/SSD by installing windows there and only having that plugged in. That will eliminate any windows issues if there are any.
 
Solution
Jan 26, 2021
5
1
15
Thank you for the swift reply!

After your suggestions I made a few more tests:

  • I've added a speaker to the motherboard and it makes 5 short beeps when the problem occur
  • I tried removing the SSD from the bench test and the issue was exactly the same, no video and 5 short beeps
  • I cleared the CMOS but the problem remained, no video and 5 beeps
  • Then, I mounted one of the non-working RX 570 on my gaming PC (ASRock B450M steel legend, R7 3700X, 2x8GB Viper Patriot at 3600MHZ, Kingston A2000 nvme 500GB SSD, Gigabyte P650B 650w) after removing any graphics drivers and I got video signal. But here it gets interesting:

Now that I have video with one of the non-working cards I proceeded to install the most recent AMD drivers (installation that started and finished without issues) and:
Tried to run Heaven, which gave me this error message.
I then checked with GPU-Z and I see this, while MSI Afterburner show this and if I tried opening the Radeon control panel i get this.

Then, I installed the working RX 570 and got the very same information/error messages from GPU-Z, Afterburner, Heaven and the Radeon software.
I then removed again the drivers with DDU and restarted the PC, re-installed again the AMD drivers and run Heaven, which went smoothly without errors or issues.
I then checked GPU-Z and got this while MSI Afterburner showed me this and I were able to open the Radeon control panel without issues.

So, what I see is that the non-working card is recognized by both GPU-Z and MSI Afterburner, but the first shows GPU clock and memory at 0mhz, and the second shows only the name of the card, but no driver version or memory and clock information.
While the non-working card display everything correctly.

At this point I understand that the problem actually lies with the cards, but what problem is it? Can it be solved? Because they are all 1 year old card and (especially given the period) it would be a great waste to just toss 'em!
 
Yeah it seems the GPUs are the issue, but you also mentioned five beeps. Five short beeps indicate CPU problem. I am not sure why you got them since the system runs fine with a known working GPU.

If they are only 1 year old, then warranty is still in effect. You should RMA them.
If for some reason you don't want or can't RMA them, I would try flashing the correct vBIOS.
You can find each GPU's vBIOS here:
 
Jan 26, 2021
5
1
15
From what I see they have both the same vBIOS (015.050.002.001.00000 - Revision "EF") so why should I flash e new vBIOS on the problematic cards?

p.s. I can't RMA them because the previous owner (from which my friend bought them) opened them and removed the warranty stickers.
 
From what I see they have both the same vBIOS (015.050.002.001.00000 - Revision "EF") so why should I flash e new vBIOS on the problematic cards?

p.s. I can't RMA them because the previous owner (from which my friend bought them) opened them and removed the warranty stickers.
Why would he remove the warranty stickers? Did he open them open and messed with them? Why would he sell them if they are only 1 year old? Did he used them for mining? Which brings me to the answer of the vBIOS. If they were used for mining, they would have been flashed with a mining bios and reflashed with the original afterwards. The reflashing might went wrong or the cards were already bad from the mining usage.

I would ask for a refund for them.
 
Jan 26, 2021
5
1
15
Yes that is exactly what I though, that they were used as mining cards. Also because my friend payed them about 60% of what you find them right now (in my area), but when I asked to my friend some info regarding the seller he said the he only know that the seller sold them as "1 year old non tested RX 570" and he bought them just for the low price and the possibility to re-sell them for a decent profit. He don't want to bother with trying to give them back to the seller, and for me it actually changes nothing, the only thing I know is that he will give me one for free if I solve the problem, so 🆒

I'll try to reflash the vBIOS then, if it works yay, if it doesn't at least I tried. What would you suggest? Should I download the same vBIOS from the working RX 570 and use that, or use one from TPU?

Edit:
Also, thank you for you help I really appreciate that
 
No problem mate, that's why I am here!

So there is no way of being sure that they are only one year old. They might have been used for 3 years and finally gave out after all that time mining.

That's what I thought about the vBIOS, you have nothing to lose since it does not properly work either way. As for what to use, I believe that TPU is the safest bet BUT, since you have 2 problematic cards, try the first one with the vBIOS from TPU and if it works, awesome... If it doesn't, use the vBIOS from the working 570. Or 1 from TPU and one from the working 570...
 
Jan 26, 2021
5
1
15
Good news! Reflashing the vBIOS fixed it! Now works perfectly and I can run any stress test without a single issue.

Thank you again for the support man, maybe I would have given up if it hadn't been for your help :)
 
  • Like
Reactions: dotas1