Random freeze ( Not BSOD ) while doing almost nothing.

[MrRenko]

• OS - Windows 11 Pro
• x64
• Windows 10 Pro
• Age of system (hardware) : +/- 5 years except graphic cards
• Age of OS installation less than a year - have you re-installed the OS? No
• Intel Core i7-8700K @ 3.70
• GeForce RTX 3070ti OC edition
• MotherBoard - Asus ROG Strix Z-370G Gaming
• Power Supply - Seasonic SSR-650RM 650W
• System Manufacturer ?
• Exact model number (if OEM or laptop) ?
• Laptop or Desktop? Desktop
   

  Hi,
   

  I got a new GPU two weeks ago. Changed my 1080ti to a 3070ti.
   

  I removed my previous Nvidia drivers and other nvidia apps ( didnt use a DDU tool ). I installed the one suggested on TUF Asus website not the latest.
   

  Since this day I got weird random crash while doing almost nothing. I'm watching something on YouTube ( with Firefox ) and the screen crash and 10-15 seconds later the computer reboot. No bluescreen, just a big freeze of everything.
   

  I witness 2 freeze but I think there was more since I got 5 minidump file in my Windows/minidump folder, all are dated after the purchase of the card.
   

  Usually the same bugcheck analysis header.

  INTERNAL_POWER_ERROR (a0)
  
  The power policy manager experienced a fatal error.
  
  Arguments:
  
  Arg1: 0000000000000608, A driver has attempted to transition a component to idle without
  
  a preceding active request.
  
  Arg2: ffffab81b698d568, POP_FX_COMPONENT component
  
  Arg3: 0000000000000000, POP_FX_COMPONENT_FLAGS component condition
  
  Arg4: 0000000000000000

   

  But the details is different from files to files.
   

  PROCESS_NAME is firefox.exe, or NordVPN.exe or dwm.exe or svchost.exe or HsTsr.exe
   

  Also the MODULE_NAME is secnvme or nvlddmkm
   

  Didn't change anything else than the GPU in the last year. Never had this kind of bug before except the fact that my computer usually never go in sleep mode, 3 out of 4 times he shut down when he is supposed to go in sleep ( even before the new GPU ).
   

  PSU is 650W gold Seasonic, I use a calculator and whole computer not supposed to go over 530w in peak.

  
  Got 3 HD, 1 Samsung SSD PRO 1TB as a gaming drive,1 intel SSD 600 gb for system, 1 Western Digital 2TB ( storage )

  
  I have test the new card with 3D Mark on stress test multiple time, no prob at all.
   

  I'm on Windows 11. Hibertate is not enable ( I read a lot of INTERNAL_POWER_ERROR (a0) related to hibernation )

  
  Do you think I need to send back the graphics card ? Get a new / stronger PSU or it might be something else ?

  
  Thanks a lot !

  
  Minidump share : https://www.mediafire.com/folder/a6f0l2xssky2g/Minidump
  
  P.S. : There was no crash since 2023-07-19 and not I got another one today an hour ago.
  
  It just freezed while watching a video on YouTube and nothing else. 

  
  
  GPU Tweak III was open when the pc crash, I took a picture with my phone. I also include the picture.

Minidump-20230726-1445.txt

[Agall]

11 minutes ago, MrRenko said:

• OS - Windows 11 Pro
• x64
• Windows 10 Pro
• Age of system (hardware) : +/- 5 years except graphic cards
• Age of OS installation less than a year - have you re-installed the OS? No
• Intel Core i7-8700K @ 3.70
• GeForce RTX 3070ti OC edition
• MotherBoard - Asus ROG Strix Z-370G Gaming
• Power Supply - Seasonic SSR-650RM 650W
• System Manufacturer ?
• Exact model number (if OEM or laptop) ?
• Laptop or Desktop? Desktop
   

  Hi,
   

  I got a new GPU two weeks ago. Changed my 1080ti to a 3070ti.
   

  I removed my previous Nvidia drivers and other nvidia apps ( didnt use a DDU tool ). I installed the one suggested on TUF Asus website not the latest.
   

  Since this day I got weird random crash while doing almost nothing. I'm watching something on YouTube ( with Firefox ) and the screen crash and 10-15 seconds later the computer reboot. No bluescreen, just a big freeze of everything.
   

  I witness 2 freeze but I think there was more since I got 5 minidump file in my Windows/minidump folder, all are dated after the purchase of the card.
   

  Usually the same bugcheck analysis header.

  INTERNAL_POWER_ERROR (a0)
  
  The power policy manager experienced a fatal error.
  
  Arguments:
  
  Arg1: 0000000000000608, A driver has attempted to transition a component to idle without
  
  a preceding active request.
  
  Arg2: ffffab81b698d568, POP_FX_COMPONENT component
  
  Arg3: 0000000000000000, POP_FX_COMPONENT_FLAGS component condition
  
  Arg4: 0000000000000000

   

  But the details is different from files to files.
   

  PROCESS_NAME is firefox.exe, or NordVPN.exe or dwm.exe or svchost.exe or HsTsr.exe
   

  Also the MODULE_NAME is secnvme or nvlddmkm
   

  Didn't change anything else than the GPU in the last year. Never had this kind of bug before except the fact that my computer usually never go in sleep mode, 3 out of 4 times he shut down when he is supposed to go in sleep ( even before the new GPU ).
   

  PSU is 650W gold Seasonic, I use a calculator and whole computer not supposed to go over 530w in peak.

  
  Got 3 HD, 1 Samsung SSD PRO 1TB as a gaming drive,1 intel SSD 600 gb for system, 1 Western Digital 2TB ( storage )

  
  I have test the new card with 3D Mark on stress test multiple time, no prob at all.
   

  I'm on Windows 11. Hibertate is not enable ( I read a lot of INTERNAL_POWER_ERROR (a0) related to hibernation )

  
  Do you think I need to send back the graphics card ? Get a new / stronger PSU or it might be something else ?

  
  Thanks a lot !

  
  Minidump share : https://www.mediafire.com/folder/a6f0l2xssky2g/Minidump
  
  P.S. : There was no crash since 2023-07-19 and not I got another one today an hour ago.
  
  It just freezed while watching a video on YouTube and nothing else. 

  
  
  GPU Tweak III was open when the pc crash, I took a picture with my phone. I also include the picture.

Minidump-20230726-1445.txt 26.64 kB · 0 downloads

If you havent full uninstalled Nvidia drivers and overclocking software then reinstalled, then I'd recommend ensuring you're not running into a CPU limitation from background tasks. The 8700k should be fine in background tasks, but its definitely a substantial limitation for an RTX 3070ti in most games in 2023.

 

A 650W Seasonic should be fine, but you can always test it by running the GPU at ~85% TDP and see if the issue occurs. That's usually a reliable way of testing that, and/or going even lower to like 50% if need be.

 

Forcing PCIe 3.0 in the UEFi might also help, since an 'auto' feature may try to negotiate higher.

[MrRenko]

5 minutes ago, Agall said:

The 8700k should be fine in background tasks, but its definitely a substantial limitation for an RTX 3070ti in most games in 2023.

I was not playing anything. Actually I never had any crash while playing a game. The 5 now 6 times happen when I only use the pc to listen to some music on YouTube.

Current CPU usage with the same stuff open : 4%, memory 21% ( of 48 GB ) and 0% of the GC.

As I said, I run multiple time the stress test on the new card and no issues at all... I run every Stress Test on 3D Mark back to back for 3 hours without any issue...

And sometimes when I watch a video everything stop, the colors of the screen get weird and after 15-20 seconds, the pc just reboot....

[Agall]

6 minutes ago, MrRenko said:

I was not playing anything. Actually I never had any crash while playing a game. The 5 now 6 times happen when I only use the pc to listen to some music on YouTube.

Current CPU usage with the same stuff open : 4%, memory 21% ( of 48 GB ) and 0% of the GC.

As I said, I run multiple time the stress test on the new card and no issues at all... I run every Stress Test on 3D Mark back to back for 3 hours without any issue...

And sometimes when I watch a video everything stop, the colors of the screen get weird and after 15-20 seconds, the pc just reboot....

I used to run into the same issue on my 7950x3D and RTX 4090 system which appears to have issues negotiating PCIe versions. Random restarts/BSODs while only doing 'background' tasks like watching youtube, but also issues on POST/BOOT where I'd either get no display and/or only get a display several seconds into BOOT.

 

Cards can supposedly negotiate lower PCIe versions to save power while in a c-state, so forcing PCIe 3.0 might solve the issue similarly to mine.

[JLssg4]

Would 'performance' power plan/settings act as a workaround-fix?
Seems to be all hinged on power management and 'transition to idle', so just never let it idle
(sorry I don't know a real solution)
 

[MrRenko]

Just now, JLssg4 said:

Would 'performance' power plan/settings act as a workaround-fix?
Seems to be all hinged on power management and 'transition to idle', so just never let it idle
(sorry I don't know a real solution)
 

Was not close to be in idle, I set it to like 5-6h and I switch the songs every 10-15m.

[JLssg4]

7 minutes ago, MrRenko said:

Was not close to be in idle, I set it to like 5-6h and I switch the songs every 10-15m.

Your SS shows 450MHz clock and 9% usage on the GPU.  Maybe multiple contextual definitions of 'idle' at play here, but I think of that as being in an idled/low power mode.  What am I missing? 
[For clarity the setting I'm referring to would be in Nvidia Control Panel -> Manage 3D settings -> power management mode.  And again, not entirely a solution as you'd be basically removing it's idle/low power mode...more of an accessible toggle to hone in on the issue and maybe gain a little control before a proper solution is found.]

[MrRenko]

2 hours ago, JLssg4 said:

Your SS shows 450MHz clock and 9% usage on the GPU.  Maybe multiple contextual definitions of 'idle' at play here, but I think of that as being in an idled/low power mode.  What am I missing? 
[For clarity the setting I'm referring to would be in Nvidia Control Panel -> Manage 3D settings -> power management mode.  And again, not entirely a solution as you'd be basically removing it's idle/low power mode...more of an accessible toggle to hone in on the issue and maybe gain a little control before a proper solution is found.]

Should switch from Normal to Prefer maximum performance ?

[MrRenko]

2 hours ago, Agall said:

I used to run into the same issue on my 7950x3D and RTX 4090 system which appears to have issues negotiating PCIe versions. Random restarts/BSODs while only doing 'background' tasks like watching youtube, but also issues on POST/BOOT where I'd either get no display and/or only get a display several seconds into BOOT.

 

Cards can supposedly negotiate lower PCIe versions to save power while in a c-state, so forcing PCIe 3.0 might solve the issue similarly to mine.

Do you know which setting I should change in UEFI Bios ?

There is 

Advanced\System Agent (SA) Configuration\DMI\OPi Configuration
   > DMI Max Link Speed : Currently at Auto , can switch to GEN 1 or GEN 2 or GEN 3


Advanced\System Agent (SA) Configuration\PEG Port Configuration
> PCIEX16/X8_1     X16
    > PCIEX16/X8_1 Link speed : Currently at Auto, can switch   to GEN 1 or GEN 2 or GEN 3
> PCIEX8_2         Not Present
    > PCIEX16/X8_1 Link speed : Currently at Auto, can switch   to GEN 1 or GEN 2 or GEN 3


Advanced\PCH Configuration\PCI Express Configuration
> PCIe Speed : Currently at Auto, can switch   to GEN 1 or GEN 2 or GEN 3


Or it's another config or all of them ?