New VEGA 64 amdkmdap crashes/resets

[SyncViews]

I got a PowerColor Radeon RX VEGA 64 Red Devil from this site and after installing getting game crashes or short lockups with amdkmdap in event viewer on Win 10 64bit. Still waiting on a new monitor and it was late so just tested at 1080p, 60Hz with a few light games like Factorio and Cities:Skylines, didn't even increase the settings.

Happens every 5-15 minutes and the event comes in pairs a few seconds apart:
"Display driver amdkmdap stopped responding and has successfully recovered."

Powered using two separate cables from an EVGA 850W SuperNOVA G2. The rest of the system had been running fine with a GTX 1060 for some months.

Used the DDU program from wagnardsoft in safe mode and installed a previously downloaded win10-64bit-radeon-software-adrenalin-edition-18.9.1-sept12.exe from amd.com.

The AMD Settings tool seems to quit on recovery, so got a screen from GPU-Z, temps looks fine and very low load except during the lockup.


I am thinking this looks like a defective GPU but anyone have any other ideas?

 

[Panos]

use DDU. Wipe the drivers in Windows safe mode. Install latest drivers 18.9 and let us know how it goes.

test the card on all 3 pre defined settings also. (power save, balanced, turbo).
use the driver overlay to observe clocks and temps.

let us know how it goes.

 

[lowdog]

I'd install the 18.5.1 drivers first as they were the last set that seem to install correctly and then test the card. The 18.9.1 drivers are dodgy and best avoided from my experience with them.

 

[SyncViews]

use DDU. Wipe the drivers in Windows safe mode. Install latest drivers 18.9 and let us know how it goes.

Used the DDU program from wagnardsoft in safe mode and installed a previously downloaded win10-64bit-radeon-software-adrenalin-edition-18.9.1-sept12.exe from amd.com.

Only seems to be a 18.9.1 or a 18.5.1, so used 18.9.1.

I'd install the 18.5.1 drivers first as they were the last set that seem to install correctly and then test the card. The 18.9.1 drivers are dodgy and best avoided from my experience with them.

Will try the 18.5.1. Is amdkmdap crashes known with 18.9.1?

test the card on all 3 pre defined settings also. (power save, balanced, turbo).
use the driver overlay to observe clocks and temps.

I tried the other settings in AMD Global Settings, didn't seem to change anything and reverted itself to balanced. Does the overlay show anything different from GPU-Z? I couldn't get the ctrl-shift-O thing to display (Alt-R did only on Desktop, not in game...) and with the settings app quiting every time didn't help and not sure if that breaks the overlay.


EDIT: Crashed Window's while repeatedly starting Factorio to try and get the overlay to work: VIDEO_SCHEDULER_INTERNAL_ERROR "
0x00000119 (0x0000000000000005, 0xffff9e8cff5cb000, 0xffff9e8cffd36ac0, 0x00000000000175ec)."

 

[SyncViews]

Reinstalled 18.5.1 cleanly after safe mode DDU and did some basic testing today without issue.

Still concerned though, while a bad driver could cause crashes, if AMD let such a thing past QA would that not be 1000's of people complaining? Worried going to have issues again later soon as anything changes...

 

[Panos]

Reinstalled 18.5.1 cleanly after safe mode DDU and did some basic testing today without issue.

Still concerned though, while a bad driver could cause crashes, if AMD let such a thing past QA would that not be 1000's of people complaining? Worried going to have issues again later soon as anything changes...

When was the last time you installed windows?

 

[lowdog]

Reinstalled 18.5.1 cleanly after safe mode DDU and did some basic testing today without issue.

Still concerned though, while a bad driver could cause crashes, if AMD let such a thing past QA would that not be 1000's of people complaining? Worried going to have issues again later soon as anything changes...

This worked for me after a clean uninstall and then DDU in safe mode - installed 18.8.2 and then did a express update to 18.9.1 but there was no option to install Radeon Relive (I had to run the amddvr64.msi from previous unpackaged 18.7.1 to install Relive as it's the last driver that had the Relive installer in it) and all was good.

 

[SyncViews]

When was the last time you installed windows?

About 9 months (was Direct Win 10 install, not upgrade from 7 or 8). Was meaning to try a clean install as well, but not had the time as lots of software. No other system issue suggesting a messed up OS though.

This worked for me after a clean uninstall and then DDU in safe mode - installed 18.8.2 and then did a express update to 18.9.1 but there was no option to install Radeon Relive (I had to run the amddvr64.msi from previous unpackaged 18.7.1 to install Relive as it's the last driver that had the Relive installer in it) and all was good.

Will try the older driver versions. But only issue you had was missing Relive, no stability issues? I can't recall if it prompted me for a restart or not, but if not I did it manually.

 

[SyncViews]

Was away for a bit but fresh windows install seemed OK on 18.5.1, about to risk trying the new versions again. Until a regular Window's update (1803, build 17134.345) then it crashed the driver even sat on the desktop. Tried system reset, reinstall again, but no fix this time, not sure on Win10 it's possible to revert to an old update state to test since a new install goes to the latest. Can only think its the card. Tried swapping the 1060 back in and it works fine (just a DDU, not clean Window's).

 

[ltron]

What power supply are you using and are using separate power cables for each connector on the card?

 

[SyncViews]

What power supply are you using and are using separate power cables for each connector on the card?

Powered using two separate cables from an EVGA 850W SuperNOVA G2.

 

[Panos]

Was away for a bit but fresh windows install seemed OK on 18.5.1, about to risk trying the new versions again. Until a regular Window's update (1803, build 17134.345) then it crashed the driver even sat on the desktop. Tried system reset, reinstall again, but no fix this time, not sure on Win10 it's possible to revert to an old update state to test since a new install goes to the latest. Can only think its the card. Tried swapping the 1060 back in and it works fine (just a DDU, not clean Window's).

By default windows 10 update to their own driver even if you have newer version. Easiest way to do so is to use ddu as it stops them to do so by default.

Use it again now to wipe the nv drivers. Put the vega and install latest drivers. Let us know how it goes.

 

[SyncViews]

I used DDU after the Windows Update that made things worse (so ~1 week old OS) to ensure it was on 18.5.1, and that is what AMD settings showed. I did not use DDU on the fresh reinstall I did (install windows, login, install driver I downloaded, restart, login, bsod or driver-reset within a couple of minutes on desktop), but it showed the desired version when I checked. I assume even if Window's some how stealth installs the driver over a manual one, it shows the right version in settings?

And if Windows update had a broken one, would expect an outcry when that goes to almost every existing customer?

 

[Panos]

I used DDU after the Windows Update that made things worse (so ~1 week old OS) to ensure it was on 18.5.1, and that is what AMD settings showed. I did not use DDU on the fresh reinstall I did (install windows, login, install driver I downloaded, restart, login, bsod or driver-reset within a couple of minutes on desktop), but it showed the desired version when I checked. I assume even if Window's some how stealth installs the driver over a manual one, it shows the right version in settings?

And if Windows update had a broken one, would expect an outcry when that goes to almost every existing customer?

IMHO you have issues that nobody else has. Especially since those are dependent on Windows version also.

 

[Davanius]

Windows updates have been screwing up my builds lately. This week alone has been a nightmare with both my 1070ti and Vega 56 with stuttering moving apps and the web browsers around the desktop. Even gpu z won't read the cards correctly and has to be restarted often.

 

[spikerules]

Have you tried a clean boot?

 

[melmac]

Faulty Card. RMA it.

 

[zophiel]

I had this on the wifes pc. 18.9.3 drivers, win 7, vega 56 (2 different ones), i5 4460. Just replaced the psu for a new evga 850G3 from an evga 550w one just incase it was not powering right. Though it has been fine for ages until the last couple of weeks.

even swapped out the 56 for the one in the kids pc, its still the same. Just getting her to go through some of her games to see what happens now with no clocking on cpu or gpu. Then will try rolling back the drivers and see what happens with those, if she has any more problems.

seems pretty crappy that its the same with both of the 56 gpus.