Is my overclock causing this?

[MindArlekin]

Some time ago I bought this:

https://www.gnr.co.uk/product/afox-geforce-gtx1060-6gb-h2-geforce-gtx-1060-6gb-gddr5_af-af1060-6144d5h2.htm

It certainly not best 1060 out there, but, considering ongoing mining craze at a moment of buying, it cost me almost MSRP, when next cheapest option price was almost double that I paid. Plus, 2 year warranty is also a good thing.

Card, in my opinion, had insufficient cooling, but I modified it in "Roadkill / zip tie style", and, after this, i considered applying daily overclock on a card.

I had some hints on what capable card is, because i messed around with overclock shortly after i bought it. So, i applied already known parameters, and proceeded to stability testing.

And here I have an issue.

I expect, that if my overclock isn't stable, that stress testing software (Heaven in my case) will crash with artifacts (8 bit like) almost instantly after applying overclock to the card. But in this case I may have Heaven looping from 15 to 30 minutes without crashing, but with some occasional stuttering and/or some strange artifacts like colored transparent glare flashes or dots. Then Heaven may just freeze or instantly crash. If i use Furmark, instead of Heaven, it just simply works, without crashes, and if i spot some artifacts (like some red pixels in a middle of a donut), I know, that i just was too greedy with memory clocks.

If I play games, they may also freeze up, or crash, but it can happen in first five minutes, or after two hours of playing. In graphics intensive games, like Ghost Recon Wildlands, for example.

 

Honestly, I don't have much experience in overclocking gpus. I have broad experience with cpu/ram, but it my first time, when i have powerful enough gpu to consider it overclocking potential viable. So, can someone share knowlege about proper testing and validating gpu overclock?

[Majestic]

Define how you "modified" the card?

[Rainbow Dash]

Here's what I've learned from overclocking Pascal GPUs, Unigine Heaven, and Super Position will push the GPU Boost 3.0 clocks a little lower than what you have the offset some odd reason and that makes it seem stable. But as soon as you run a game the GPU Boost 3.0 will push the clock to the max you have on the offset, which then causes the GPU to crash. So the only way to find out what the max GPU Boost clock you get from the offset is to use GPU-Z and click on the PCI-E Render Test:

Then go to the Sensors and you will see what your max boost clock would be, for example, my GTX 1050TI will boost up to 1923MHz max, and when I run Unigine Heaven or Super Position it would run the GPU Boost up to 1875MHz - 1911MHz so it doesn't crash when you use the benchmark. The only way to actually test for stability with the GPU clock is to run a game with low to medium settings so the Boost Clock will keep at the max speed constantly.

 

For memory clocks you can run MSI Kombuster's 4GB VRAM burn test until you see artifacts or the program crashes.

[MindArlekin]

This how it looks out of the box:This what radiator is in the card:

this is "new" radiator from MSI gtx 760:

This how card looks now:

 

[MindArlekin]

I installed Kombustor and run the memory burn stress test.

After i stop test via pressing Esc button, or after program crashes i experience stutters in a system.

Feels like when some software loads all memory/hard disk. But if i open task manager or other monitoring software, i see this:

Seems like gpu core somehow remains under load even after test is done/crash and Kombustor and Afterburner is closed.

 

[Rainbow Dash]

20 minutes ago, MindArlekin said:

I installed Kombustor and run the memory burn stress test.

After i stop test via pressing Esc button, or after program crashes i experience stutters in a system.

Feels like when some software loads all memory/hard disk. But if i open task manager or other monitoring software, i see this:

Seems like gpu core somehow remains under load even after test is done/crash and Kombustor and Afterburner is closed.

Right click on the task manager and enable the GPU and GPU Engine values so you can see if the program has hung.