Vega 64 shuts down under load

[Tegneren]

Hi

I have a MSI Vega 64 Wave (Liquid cooled) that has recently just started to shut down when put under load (eg. games, folding, benchmarks). When this happen I get a signal loss on the monitors and all the power LEDs on the GPU goes dark (there should be at least one glowing as long as its powered on). The pc continues to run and I still get sound from videos and such, but no picture on the monitors, and I have to restart the PC to get picture back.

It works just fine with just regular non-demanding use such as watching videos and internet browsing.

Sometimes it works for a while during games, but mostly it just shuts right off.

 

Both Windows, the Radeon software and other drivers is fully updated. The GPU is connected to PSU with two seperate 8-pin connectors, and i have checked that everything is fully connected and seated in the respective sockets.

 

I recently upgraded the CPU and MB from a R7 3700X w/ MSI X370 Gaming Pro Carbon to what is noted below, but the system ran fine for weeks after that.

 

The temperatures is fine with around 50-60C for the GPU and 60-70C for the memory under heavy load (the times where it actually works), and the full system draws around 650W at full tilt (measured at the wall), so I should think a 850W PSU would be enough.

After  this started happening I thought the Liquid metal i had on the GPU might have spilled and causing trouble, so I opened it up and replaced the liquid metal with regular paste. There was no sign of spillage, and replacing it didnt fix the issue. I also replaced the pads

(side note: I'll never using liquid metal again. its such a pain to use and clean up, and didn't give any noticeable temperature difference. So basicly not worth it)

 

Anyone have any Idea whats going on or how to fix it?

 

My system:

Windows 10 Pro. 20H2 build 19042.746

Ryzen 7 5700X

ASUS ROG Strix B550-F GAMING

MSI RADEON RX VEGA 64 WAVE 8G

4x G.Skill Flare X 8GB 3200MHz CL14 (@3600MHz CL16)

Samsung 960 EVO 250GB M.2 NVMe 

NZXT Kraken X61 AIO cooler

CORSAIR TX 850M 80+ Bronze PSU 

Monitors:

Samsung C27HG70

Samsung PC-skjerm LS27E591CS

[lambdabambda]

Is your card overclocked? Maybe it could be unstable

[Tegneren]

I had the power slider on 50% and memory on 1000MHz (up from 945MHZ stock). This has been fully stable for years, and it didn't help to reset it.

I also tried to underclock it, but it didn't help.

[-iSynthesis]

I've had something similar happening with my Vega 56 I believe. It shut down both monitors and the fans ramped up to 100%, temps were fine seemingly.

For me it was the Hot Spot temp (it's shown in hwinfo). While the GPU temp was around 70 or 75 the Hot Spot went up to 118 and shut down afterwards.

 

I used a method of tensioning the cooler that Igor's lab used. Basically you want to put quite a bit of paste on the GPU (The die on vegas is really large and sometimes uneven).

Then you mainly tension the upper two screws of the core (The one further away from the PCIe slot) and once they are completely screwed in continue with the other ones.

This took my Hot Spot temp down a good 20° and now it does not crash anymore.

 

Before doing all of this report Hot Spot temps first obviously, might be just fine and be another problem.

[Nena Trinity]

It might be your power supplies over current protection getting triggered, happened to my Cooler Master PSU so I had to replace with with a EVGA G2... UwU

 

At the time I had FX-8370 & RX Vega 56

[Tegneren]

2 minutes ago, Nena Trinity said:

It might be your power supplies over current protection getting triggered, ...

Wouldnt that shut down the hole computer?

[Nena Trinity]

1 minute ago, Tegneren said:

Wouldnt that shut down the hole computer?

Try DDU remove all Nvidia, AMD & Intel drivers, reboot and re-install a driver that is known to be stable. Not always latst driver! UwU

[Tegneren]

16 minutes ago, -iSynthesis said:

Before doing all of this report Hot Spot temps first obviously, might be just fine and be another problem

I just ran a furmark stress test on remote desktop (im at work right now), and as you can see the hot spot temp is fine.

It crashed btw. or at least the picture in teamviewer froze.

[Nena Trinity]

1 minute ago, Tegneren said:

I just ran a furmark stress test on remote desktop (im at work right now), and as you can see the hot spot temp is fine.

It crashed btw. or at least the picture in teamviewer froze.

I suspect faulty drivers or it might be the card... Re-install windows is a option here?

[Tegneren]

1 minute ago, Nena Trinity said:

I suspect faulty drivers or it might be the card... Re-install windows is a option here?

I will try DDU'ing and install an older driver

The windows install is just a week or so old, so i'm gonna hold off on that. Although the windows installer couldn't find the driver for the on-board network card on my MB during installation for some reason.

[Tegneren]

1 hour ago, Nena Trinity said:

Try DDU remove all Nvidia, AMD & Intel drivers, reboot and re-install a driver that is known to be stable. Not always latst driver! UwU

DDU complete and installed version 20.11.2 of the radeon software (I actually think i had the "optional" driver 21.2.1 installed before)

Running a Furmark stresstest whit my "standard" OC atm. Currently 20 minutes in. Seems fine so far with hot spot temp of 79C.

[Mark Kaine]

2 hours ago, Tegneren said:

I just ran a furmark stress test on remote desktop (im at work right now), and as you can see the hot spot temp is fine.

It crashed btw. or at least the picture in teamviewer froze.

I think something is wrong with these numbers... AMD GPUs have higher Hotspot temps if you just sneeze in their direction...

 

 

like others I recommend to DDU and install known to be good drivers. 

 

I also suspect your PSU getting old from all the overclocking btw, also worth checking imo. 

 

 

edit: 

1 hour ago, Tegneren said:

Seems fine so far with hot spot temp of 79C.

yeah, that's more realistic

[-iSynthesis]

50 minutes ago, Mark Kaine said:

I think something is wrong with these numbers... AMD GPUs have higher Hotspot temps if you just sneeze in their direction...

 

 

like others I recommend to DDU and install known to be good drivers. 

 

I also suspect your PSU getting old from all the overclocking btw, also worth checking imo. 

 

 

edit: 

yeah, that's more realistic

I believe there was an old undervolt being applied. One of the above pics showed a max core voltage of 1.044V, stock is 1.25 I believe. You can undervolt a good amount but mine wasn't terribly stable below 1050...

That'd also explain the low temps

[Tegneren]

21 minutes ago, -iSynthesis said:

I believe there was an old undervolt being applied. One of the above pics showed a max core voltage of 1.044V, stock is 1.25 I believe.

I haven't done anything with the voltage. 

Running Furmark (same settings as from the picture earlier) now the core voltage is at an average of 0.940V at stock speeds and 1.060V when OC (+50% power limit and 1100MHz memory)

And its seemingly stable

 

1 hour ago, Mark Kaine said:

think something is wrong with these numbers... AMD GPUs have higher Hotspot temps if you just sneeze in their direction...

The temps in the picture are after just a few seconds before it crashed in a very cold apartment, so I dont think they were incorrect necessarily.

 

But anyways it seems to run fine after a DDU. Thanks to all of you

I will let you know if the issue returns

[Tegneren]

Ok. it seems the issue didn't go away after all. 

It can work fine for a while after a driver reset (DDU/reinstall), but after som time under load it returns.

 

i suspect she has put in her retirement notice, but I'm still interested to know if its something I can do about it.

 

I actually managed to snag me a RTX 3070 from a online store today and it will arrive in a couple of days, when I checked yesterday there wasn't a single consumer grade graphics card available in the entire country between $200 - $1000, but when i checked today there was 24 ASUS 3070 TUF OC in stock, so I pulled the trigger. When I had completed the order there was none left in stock. 

I have been meaning to replace it for a while anyways

 

So it will be replaced and then i can find out if it actually was the GPU or something else that was the issue.

 

 

[Falkentyne]

7 hours ago, Tegneren said:

Ok. it seems the issue didn't go away after all. 

It can work fine for a while after a driver reset (DDU/reinstall), but after som time under load it returns.

 

i suspect she has put in her retirement notice, but I'm still interested to know if its something I can do about it.

 

I actually managed to snag me a RTX 3070 from a online store today and it will arrive in a couple of days, when I checked yesterday there wasn't a single consumer grade graphics card available in the entire country between $200 - $1000, but when i checked today there was 24 ASUS 3070 TUF OC in stock, so I pulled the trigger. When I had completed the order there was none left in stock. 

I have been meaning to replace it for a while anyways

 

So it will be replaced and then i can find out if it actually was the GPU or something else that was the issue.

 

 

I was getting this black screen crash on my Vega 64.  It preceded the HBM memory dying (color vertical bars in BIOS with some text visible, smaller bars in windows).

Before I got the black screens, I got random sporadic slowdowns in Fortnite, like from 120 FPS to 10 FPS for a few seconds, then back up to 120 FPS.  Eventually that led to black screens.  And then color bars

 

Lucky I had already ordered my 3090 FE at that time and it was being shipped, so I only had to suffer with a backup R9 290X for a few days.

[Tegneren]

2 hours ago, SavageNeo said:

your psu is pretty old to this point. Might want to replace it.

 

I will try that if the issue persist with the new GPU

[PcTechGuy8989]

On 2/2/2021 at 2:17 PM, Tegneren said:

Hi

I have a MSI Vega 64 Wave (Liquid cooled) that has recently just started to shut down when put under load (eg. games, folding, benchmarks). When this happen I get a signal loss on the monitors and all the power LEDs on the GPU goes dark (there should be at least one glowing as long as its powered on). The pc continues to run and I still get sound from videos and such, but no picture on the monitors, and I have to restart the PC to get picture back.

It works just fine with just regular non-demanding use such as watching videos and internet browsing.

Sometimes it works for a while during games, but mostly it just shuts right off.

 

Both Windows, the Radeon software and other drivers is fully updated. The GPU is connected to PSU with two seperate 8-pin connectors, and i have checked that everything is fully connected and seated in the respective sockets.

 

I recently upgraded the CPU and MB from a R7 3700X w/ MSI X370 Gaming Pro Carbon to what is noted below, but the system ran fine for weeks after that.

 

The temperatures is fine with around 50-60C for the GPU and 60-70C for the memory under heavy load (the times where it actually works), and the full system draws around 650W at full tilt (measured at the wall), so I should think a 850W PSU would be enough.

After  this started happening I thought the Liquid metal i had on the GPU might have spilled and causing trouble, so I opened it up and replaced the liquid metal with regular paste. There was no sign of spillage, and replacing it didnt fix the issue. I also replaced the pads

(side note: I'll never using liquid metal again. its such a pain to use and clean up, and didn't give any noticeable temperature difference. So basicly not worth it)

 

Anyone have any Idea whats going on or how to fix it?

 

My system:

Windows 10 Pro. 20H2 build 19042.746

Ryzen 7 5700X

ASUS ROG Strix B550-F GAMING

MSI RADEON RX VEGA 64 WAVE 8G

4x G.Skill Flare X 8GB 3200MHz CL14 (@3600MHz CL16)

Samsung 960 EVO 250GB M.2 NVMe 

NZXT Kraken X61 AIO cooler

CORSAIR TX 850M 80+ Bronze PSU 

Monitors:

Samsung C27HG70

Samsung PC-skjerm LS27E591CS

Hello there, I'm Josh. 

I'm going to attempt my best at helping you resolve this issue.

Before I get any furthur into instructing you I need a few questions answered:

1. Is your card overclocked?
2. Have you checked for overheating?
3. Have you checked whether the coolers have any dust on or in them?
4. Are you willing to undervolt using a third party software recommended by thousands of tech specialists?

Looking forward to a reply,
Josh,

[Tegneren]

12 hours ago, PcTechGuy8989 said:

Hello there, I'm Josh. 

I'm going to attempt my best at helping you resolve this issue.

Before I get any furthur into instructing you I need a few questions answered:

1. Is your card overclocked?
2. Have you checked for overheating?
3. Have you checked whether the coolers have any dust on or in them?
4. Are you willing to undervolt using a third party software recommended by thousands of tech specialists?

Looking forward to a reply,
Josh,

Hi

Some of this is answered earlier, but:

 

1. I have been running a OC most of the time, using MSI After burner I had power limit increased with 50% and memory speed set to 1000MHz. But running on stock speed  doesn't help, neither does decreasing core clockspeed below stock.

2. Yes. Temperatures are fine. you can see that if you read furter up this thread.

3. No signifikant dust build-up.

4. I have tried some undervolting in After Burner, but never been able to get it to run stable.

[Jeppes]

What bios is flashed on that card? There is just no way stock voltage is under 1V. It is 1.2V.

[Tegneren]

On 2/7/2021 at 2:48 PM, SavageNeo said:

your psu is pretty old to this point. Might want to replace it.

I dont think the PSU was the problem either. I just got my new GPU (RTX 3070) and it has been running stable with folding and some gaming for a day now.

 

21 hours ago, Jeppes said:

What bios is flashed on that card? There is just no way stock voltage is under 1V. It is 1.2V.

Its the original bios. Or at least I haven't changed it. I bought it used(for quite cheap at the time). The previous owner didn't seem like the guy to mess with bios and such and had only owned it for a relatively short time.

It has dual bios, but the other setting only gave me higher power consumption and heat output and less stability. Only tried it once.

 

And regarding the voltage. It only was this low when using 720p resolutinon with no AA in Furmark. When i upped it to 1080p or higher it was around 1.2V.

 

 

I think I have to write it of as a lost cause at this point, but I'm going to try a bit more.

[Spotty]

Thread cleaned. Stay on topic and no personal bickering and attacks.

• Ensure a friendly atmosphere to our visitors and forum members.
• Encourage the freedom of expression and exchange of information in a mature and responsible manner.
• "Don't be a dick" - Wil Wheaton.
• "Be excellent to each other" - Bill and Ted.
• Remember your audience; both present and future.

[Nena Trinity]

4 hours ago, Tegneren said:

I dont think the PSU was the problem either. I just got my new GPU (RTX 3070) and it has been running stable with folding and some gaming for a day now.

 

Its the original bios. Or at least I haven't changed it. I bought it used(for quite cheap at the time). The previous owner didn't seem like the guy to mess with bios and such and had only owned it for a relatively short time.

It has dual bios, but the other setting only gave me higher power consumption and heat output and less stability. Only tried it once.

 

And regarding the voltage. It only was this low when using 720p resolutinon with no AA in Furmark. When i upped it to 1080p or higher it was around 1.2V.

 

 

I think I have to write it of as a lost cause at this point, but I'm going to try a bit more.

Maybe the owner before him tried to do something such as installing a custom BIOS or tweaked BIOS to gain more perforamnce? In that case I would see if there is a BIOS switch on the card?