Graphics Cards

[mgr83]

Hello all

First time on this forum - please be kind lol

I've been running a R7 since launch at stock frequencies with a stable undervolt (using Heaven) at 962mv.  All has been OK until about three weeks ago.  I'm now having frequent (hard) crashes with Wattman settings being restored on restart. 

The crashes mainly occur when the PC is idling on desktop.  The screen goes black and the GPU fan picks up.  The PC is unresponsive. All I can do is power off and restart. 

I've tried:

Resetting back to Wattman default settings

Using two separate power cables to PSU

Uninstalling and reinstalling Adrenaline Driver

Looking on the forums some people have suggested the problem is too little volts at the lower power states.  I'm currently running Wattman at 962mv across all power states.  This has helped but the system still fails eventually. 

It maybe unrelated but I've also noticed the GPU loading is highly variable during gaming.  This has not lead to a good gaming experience, particularly in VR due to frame drops.  I've attached a pic to demonstrate.  Temps are all good, esp. with undervolt.

Is anyone else having the same problem? Does anyone have a solution? Thinking about RMAing the card as I've had about enough of it

Thanks in advance

David

FX8370 @ 4.5GHz (Yes I know - i'm waiting for Ryzen 3k)

Asrock Extreme 3 board

16GB DDR3 @ 2400MHz

PSU rated at 750w

Various SSDs

Windows 10

Driver version 19.5.2

[ajlueke]

Hello David,

Just to clarify, do you also notice the crashing when you are running at stock settings? 

[mgr83]

Hello

Thanks for the reply. I've reset system settings bk to default on several occasions and it is worse.  Presumably because of the voltage curve and lower voltages at the bttm end of the curve. 

Can crash in seconds of startup at default. Strange how it was OK until 3 or so weeks ago.

David

[ajlueke]

Try stripping out the AMD drivers with the clean install utility from AMD.  When you load back into windows 10, does it still crash sans drivers?

You can then reinstall your chipset drivers followed by the adrenaline drivers, testing the desktop stability each time.

[amite]

Nice overclock - looking at your system and all the hardware your power supply might be clipping. Might want to reduce the load a  bit and see if problem persist.

[jdbruce]

Ok I had to create an account just to make this post. This issue has been ongoing since last year, see this 7 page thread: Default Radeon WattMan settings has been restored due to unexpected system failure 

Also this issue isn't just restricted to Windows 10 or the RX series. A few months ago I decided to update the graphics drivers on my Windows 7 machine running with a HD 7950 on a single monitor setup. It is relatively older machine which I use mostly for work, I was hesitant to update the drivers since it was an older card and was running smoothly, but eventually the update notice got the best of me.

A few weeks later my problems begun, I started to notice visual artifacts appearing on the screen and I got the WattMan error, it got worse over a period of a few weeks until the card completely died. I had experienced a similar problem in the past and was able to fix it by giving the exposed board on the card a good clean so I tried that again, but it didn't seem to resolve the issue this time.

In the end I gave up and decided the card probably died from natural causes, but it seems like that's probably not true, had I not updated my drivers my card would likely still be fine. In any case, I bought a new Gigabyte RX 570 4GB card a couple of weeks ago and installed it in the same machine, and it's been working smoothly, until today. I just had visual artifacts the WattMan message saying my settings have been reverted to stock.

I did a Google search and found many more threads about this issue than the searches I did when my 7950 died, so it seems quite clear there's a driver issue. My first question would be, shouldn't the voltages be set to stock to begin with? Are the new drivers overlocking cards by default and if so is that really a good idea? If my new RX 570 dies I wont be very happy and you can bet I'll be getting a refund.

Reading through that long thread I noticed someone mentioned that "What I find out is that every third party program that messes with Voltages or refers to Fan rpms etc, has something to do with the wattman problem." and I feel like this could be on the right track, I have some Windows 7 gadgets on my desktop which monitor my CPU and GPU usages and temperatures, and Windows 10 has a more complex hardware monitoring system which may be why the issue is mainly affecting Windows 10 machines.

[mgr83]

Aww thanks mate. 

I've read the forum, I'm having the exact same problem! There are a few suggestion from Matt @ AMD and I shall try those.  Then I shall try an uninstall using DDU followed by a phased reinstall as suggested by Ajlueke above.  I will also uninstall Overdrive and MSI afterburner just in case they are incompatible.  A weekend job I think! I'm not going to reinstall Windows as I'm due to build my Ryzen 3k build in July (fingers crossed). 

I will let you know how I get on!

Thanks

[mgr83]

Hello, further to the above I have today:

Fixed some corrept system files

Troubleshooted the windows update utility

Updated Windows version 1803

Deleted AMD Overdrive and MSI afterburner

Removed VGA drivers using VDU

Installed AMD driver only ie not audio or Adrenalin

Annnnnd....

It is still crashing!

Think is, it was stable using windows basic graphics driver so gotta be either the card or driver.

Am really not happy as currently have a £650 paper weight!

Guna put the R7 in my new Rzyen 3000 build in July.  If it's still the same I'm going to RMA it. 

Thanks for your help tho.

AMD - SORT OUT YOUR SH*T!

[ajlueke]

If it is stable using the basic Windows driver, doesn't that rule out the card?  Seems the Adrenaline drivers aren't playing nicely on your system for some reason.

[paloro]

Wattman issues, and posts about same, have been abundant. No real solution has ever been discovered.

[pokester]

I can tell you in 3 of my older cards 2 HD 7xxx and one R9 380x. I had to set my power limit to maximum, to get these cards stable in the new drivers. All these cards had always worked fine before these last few drivers. Luckily I knew to try this because my RX 580 has always needed me to do this. If you have not tried this, you might see if it helps your situation or not. I couldn't even watch video in a browser without doing this. I am going to try this latest driver 19.6.1 that just came out and see if it helps. Many are reporting this driver fixes multiple issues. We will see.