Frequent and random appearing BSOD 'internal power error'.

The AMD driver that was installed before the BSOD troubleshooting was 8.14.1.6489
The current AMD driver is 8.14.1.6463
It is unclear the driver that had been installed after first using DDU and before the current installation.
It may take some time to evaluate the driver with typical computer activities.
The graphics hardware and driver can be stress tested.
It will be important to monitor temperature, freezing , display artifacts etc. and abort testing as needed.
Aim to test for 1 hour.
Read this link an make note of the warnings before starting the test: FurMark - GPU Stress Test Performance Maintenance Tutorials

Code:

Event[719]:
  Log Name: System
  Source: Display
  Date: 2017-11-27T18:06:13.764
  Event ID: 4101
  Task: N/A
  Level: Warning
  Opcode: Info
  Keyword: Classic
  User: N/A
  User Name: N/A
  Computer: Antec
  Description: 
Display driver amdkmdap stopped responding and has successfully recovered.

Update: Still haven't got any BSOD. :)

But, my favorite game Dungeons2 still freezes during the game. Now, it already happens while loading a new or saved game, so I can't play at all now.

So, yesterday I updated the graphics driver manually from package "amd-catalyst-15.7.1-win10-64bit" to the driver in package "amd non-whql-win10-64bit-radeon-software-crimson-relive-16.2.1-sep20" from within device manager. It is the newest driver the AMD site has available for this graphics card.

the driver now says:
Attachment 166333

I tried to play dungeons2 again yesterday, but the same happens. The game freezes while loading a saved or new game. Once I got it to run, but it froze after a few minutes. The temperature of the graphics card isn't elevated at all. (around 55 degrees.)

This evening, I did the FurMark test for 1 hour. All seemed OK, see the screenshots below.

after 20 minutes:
Attachment 166334
Attachment 166335
Attachment 166336

after 1 hour:
Attachment 166337
Attachment 166338
Attachment 166339

I also tested another game (War for the Overworld, Minecraft) the past days, and I could play these games for hours without problem.

Finally, I run DM_collector this evening and uploaded the zip file.

What should be my next step? Start with the driver verifier steps, or e.g. reinstall dungeons2 first...?

Run Memtest86+ version 5.01 for 8 or more passes.
Memtest86+ - Advanced Memory Diagnostic Tool
Memtest86+ - Advanced Memory Diagnostic Tool
Memtest 86+ version 5.01 is not run by time but by passes.
The more the passes the better the test.
It sometimes takes passes in the 20s to find problems with RAM / DIMM /MB
The computer has 8 GB RAM
There is 2 Gb on each DIMM.
Memtest86+ version 5.01 testing takes approximately 1 -2 hours / GB RAM
So plan most of the testing overnight.
Just 1 error is a fail and you can abort testing.
Then test 1 RAM module at a time in the same DIMM for 8 or more passes.
Use a camera or a smart phone camera to take a picture and post an image into the thread.
Memtest86+ - Advanced Memory Diagnostic Tool

step 4) There is 1 newer bios available: F9b.
see: https://www.gigabyte.com/Motherboard...-10#support-dl

As this is a beta BIOS , I decided (in the past) not to update to this version. Do you agree, or should I update it to F9b?

(Still no BSOD :) and I can play dungeons2 again, but it still freezes regularly during playing)

Yes F8 is the latest non-beta and that is the optimal choice.
Let's check the hardware (memtest86+ version 5.01 followed by CPU testing) and then move to Windows driver verifier if there is no freezing or bsod during CPU testing.

For Memtest86+ version 5.01 all RAM should be tested at one time as to optimize the testing.
Memtest86+ - Advanced Memory Diagnostic Tool
Memtest86+ - Advanced Memory Diagnostic Tool
The testing is done not by time but by passes.
The more the passes the better the test.
Sometimes it takes passes in the 20s to find malfunctioning RAM.
Testing RAM modules at the same time allows testing of multichannel capability of the motherboard and RAM.
Just 1 error is a fail and you can abort testing.
Then to differentiate good from bad RAM modules the modules are tested 1 at a teim in the same DIMM for 8 or more passes each.
A substantial portion of the testing can be done overnight.
When the testing has completed 8 or more passes use a camera or smart phone camera to take a picture and post and image into the thread.
Testing 1 RAM module at a time is not needed if all RAM modules pass when tested together. Again the more the number of passes the better the test.
Testing RAM modules 1 at a time is only needed to differentiate good from bad RAM modules or for determining good from bad DIMM.

here the result of the first testrun last night. 17 hours, 9 passes, 0 errors.
Attachment 167780

Test the CPU: Prime95 - Stress Test Your CPU Performance Maintenance Tutorials
Aim for > 3 hours testing and record test duration and maximum temperature.
Make note of the warnings in the link and abort testing as needed.
Please use the Microsoft snipping tool to post images into the thread.
Take Screenshot in Windows 10 General Tips Tutorials

Post an updated zip now and post a zip whenever there is a new BSOD:
BSOD - Posting Instructions - Windows 10 Forums

Pending the results of Prime95 plan to use Windows driver verifier:
Driver Verifier-- tracking down a mis-behaving driver. - Microsoft Community
Enable and Disable Driver Verifier in Windows 10 Performance Maintenance Tutorials