Frequent BSoD for ntoskrnl.exe, dxgmms1.sys when watching yt, twitch

[Arc]

The stress test on the GPU is passed, but the GPU is failing on other things. So the issue may not be the GPU itself.

There is no finite reason to suspect the RAM, but a failing memory may make anything fail, and may disguise as any other error. Better you test your RAM modules for possible errors, if not already done during the issue.
How to Test and Diagnose RAM Issues with Memtest86+
Run memtest for at least 8 passes, preferably overnight.

And how old the build is?

Is the computer hot? Report us the heat of the computer after a couple of hours of your normal/stressful usage. Upload a screenshot of the summery tab of Speccy.

 

[enzym]

The thing is that I actually replaced both the GPU as well as RAM completely because of those BSoDs.
The complete system is from October 2009.

Feb 2013 - RAM replaced entirely by a 4x4 GB set to avoid dual channel issues (checked voltage in BIOS, too).
Jun 2011 - GPU (MSI AMD Radeon HD 6950 Twin Frozr II) was bought to replace a 2x SLI of Radeon HD 5770. I specifically chose a single card of a name brand like this to avoid problems with the dual monitor configuration that crossfire gave me as well as problems with airflow and to have good cooling. I had installed Arctic Cooling Accelero L2 Pros on the ATIs previously to see if temperature might be a problem.
My CPU has a Scythe Katana 3 sitting on top of it that was purchased in the same round as the Arctic GPU coolers, to make sure good cooling wasn't an issue there either. None of that helped to remedy the BSoDs.

I ran multiple instances MemTest for over 8 hours before coming to this forum - it turned up no errors.
I also ran MemTest86+ from a USB flash drive for 2 passes, which took several hours by itself as it would be with 16 GB. Intended to complete the recommended 8-9 passes, but it seems that that would take at least 3 days of continuous testing, and since the RAM was new I decided against it for the time being.

Furmark stressed the GPU and quickly reached a stable 79°C, which it didn't exceed even after 15 minutes.
This is Speccy in idle (image).

I went ahead and ordered a new system minus GPU/RAM, including OS and SSD even though we failed to identify the cause. Maybe that will lead to some insights, whether BSoDs persist or not.

What really confuses me (and I can't put into words how much it does) is that the crashes only seem to occur when I'm watching livestreams.

 

[Arc]

Memtest86+ should be run 8 passes.

Idle temps are good, but during gaming or otherwise GPU usage, the temperature might be increasing. Otherwise there are no real reason of crashes there.

Can you dismantle the PC and clean the inside carefully, specially the air vents and the fan grills? And while re-seating, apply new thermal paste the the CPU.

Check if the Power Supply Unit (PSU) supplying adequate power to the computer or not.
eXtreme Power Supply Calculator
At least 30% extra is safe.

As another test, if you use the onboard graphics, does it crash?

 

[enzym]

We already checked Wattage on page 3. It came out at 372 of 550 supplied W.

I cleaned the PC and re-applied thermal paste after finding out that BSoDs continued even with the MSI card a while ago, too. I made sure to use the right amount (less is more) and saw a 5°C drop in temperatures across the board. No errors after stressing the system by running multiple benchmarks then... (furmark, prime95, unigine)

I'm not aware my board (GA-MA790XT-UD4P) has onboard graphics. How do I switch to it?

Maybe now you understand why this problem has been so frustrating for me. It was no joke when I said that I've been trying to figure this out for months before I turned to you guys. And even with everything we've tested here, we're still at square one.

 

[Arc]

I have requested Vir Gnarus to have a look at your thread. His knowledge and efficiency is supreme. May be he has a way out of the situation.

 

[Vir Gnarus]

I see bit flips in the CPU registers in the crash dumps, which are causing the crashes by accessing memory using corrupted memory addresses. Unfortunately, that doesn't really narrow things down much, it just tells us there's a hardware problem.

I'm having difficulty figuring out what you've done so far by reading through the thread. What hardware tests have you done so far? What hardware have you swapped out? Unfortunately, without the right tools, hardware diagnostics usually end up going the hardware swapping route, which evidently from what you've said you are now in the process of doing. Just as a warning, though: purchasing multiple new parts at once increases the chance of complicating things by buying a dud part, which will only throw us a curveball by presenting similar or its own set of problems. The most optimal way of going at this would be to have a reliable system available to test the purchased parts in, but I understand if that is not available. Hence why sometimes it's just best to send it to a local repair shop as they have the tools to do just that, and moreso.

CONTEXT:  fffff880095630f0 -- (.cxr 0xfffff880095630f0)
rax=fffffa800f5177c8 rbx=fffffa800f5177c0 rcx=f[B][COLOR=Red]7[/COLOR][/B]fffa800f5177c8
rdx=fffffa800d01ebf0 rsi=00000000014a2e00 rdi=fffffa800f616640
rip=fffff800030c7ccb rsp=fffff88009563ad0 rbp=00000000014a2e70
 r8=0000000000000000  r9=0000000000000000 r10=fffffa800cb045d0
r11=00000000001f0003 r12=fffff88002fd5180 r13=fffffa800f5177c8
r14=fffffa800f4e09d8 r15=fffffa800f5177c8
iopl=0         nv up ei pl zr na po nc
cs=0010  ss=0018  ds=002b  es=002b  fs=0053  gs=002b             efl=00010246
nt!KiInsertQueue+0xab:
fffff800`030c7ccb 48894108        mov     qword ptr [rcx+8],rax ds:002b:f[B][COLOR=Red]7[/COLOR][/B]fffa80`0f5177d0=????????????????
Resetting default scope