I've been toying around with my 1070 ti FTW  and wanted to ask for some advice if it is possible to stabilize my clocks that work in benches/gpu stress tests but not in game.
 
Stable overclock 1070 ti:
 
power target 120%
Temp target 92c
GPU offset: 150
Memory clock offset: 350
Volt: 100%
K boost: on
 
Result: Stable in games/various benchmarks stress tests/0 artifacts after 30 minutes and a Unigine superposition score of 15620 on medium
This is done with EVGA precision.
 
Card runs on 2050 core clock, usually clocks down to 2025 and in some games 2012 depending on the fan speed.
Haven't ran in any issue after a few days of use.
 
 
 
Did another with MSI after burner while keeping precision off and this was the result
 
power target 120%
Temp target 92c
GPU offset: 200
Memory clock offset: 200 ((anything above 200 crashes with msi overclock))
Volt: 100%
 
Result: Stable in benchmarks/gpu stress tests/ 0 artifacts and a score of 15760 in unigine superposition.
That was after running tests for 2 hours+.
 
The moment I boot Overwatch it crashes with in 10 min of gameplay.
 
 
 
EVGA precision test:
 
power target 120%
Temp target 92c
GPU offset: Up to 190 everything stable in what ever sort of bench/GPU stress test.
Memory clock offset: 350
Volt: 100%
K boost: on
 
Same story, the moment I boot Overwatch and launch a game  it crashes with in 10 minutes, even if I put the offset on 155.
 
 
Has this card just hit it's core clock cap?
All fun seeing it run on 2100 core clock in benches and getting higher min/average/max FPS along with the scores.
But I can't get it stable in games it is pretty useless.
 
Would different drivers/clean windows 10 install (had a 680GTX in there up to recently) or anything else help to get it stable at 2100 core or is this just it?
 
 
 
 

I've uninstalled all Nvdia drivers through cccleaner, wiped out all nvidia folders manually after.
With the latest nvidia drivers I used a clean install.
However I did not use DDU or wiped out the registry.
 
I'll reinstall everything and see.
Tempted to just do a clean Windows 10 install, as much of a pain as that can be haha.
 
I see it might be Overwatch related as well, mainly used that game as a stability tester.
I'll try a few other games with the 190 offset and see how they run.
Will report it back here later.

Thank you, that one was helpful.
 
 

I've done the DDU/regi cleaner and tried it again.
The card pretty much crashes in various game once I pass the 150 offset, doesn't seem to matter if I lower volt/power or keep mem clock on 0.
It doesn't seem to want to go above the 2050 core clock, memory stays stable after 500 overclock as it keeps increasing the min/avarage/max FPS +scores.
 
Maybe new drivers will allow a small push but I think this about it, bit of an average overclock if I'm led to believe the posts on the net.
 

I've been testing out your suggestions:
 
- Max power and Max temps with volt down results in a  1873 clock, in game it runs 60 degrees.
This is stable.
 
I've ran EVGA's OC tool, results are these.
 
- Long scan 30-60 min crashes no matter what I do.
- Manual scan crashes, same story.
- Short scan results in these numbers.
 
Power target 100
Temp target 83
GPU clock 114
Mem clock 0
 
1999 GPU clock
Moving power target up.
Clock 2012
Moving volt to max
2025
 
All of these are stable in every game/bench so far.
 
Next up is
 
Power target 120
Temp target 92
GPU clock 150
Mem clock 350
Volt maxed
 
Result 2050 clock
4353 mem clock
All of these are stable in every game/bench so far.
 
Last one.
 
Attempt to get GPU clock stable at 2062
 
- Attempt 1: Achieving it by maxing power slider, GPU clock 175 and upping the volt.
It just crashes after a while.
Upping the volt just results in faster  crashes as it adds a higher core clock.
 
- Attempt 2: Reducing power slider and upping the volt to get the 2062
Crashes
 
- Attempt 3: Lowering power slide/max temp slider and upping the offset to get 2062.
Still crashes.
 
 
I don't think the card needs a RMA, though past 150 offset it doesn't seem to matter if you add more volt or not.
It just doesn't want to get stable.
Temperatures between out of the box and the 2050 clock are not much different either.
 
We're talking about a very small FPS gain, but it just bites me that I can get it stable at 2100 in every bench/stress test I could find with 0 artifact.
Yet in games it just crashes.
 
 
 

You could try using afterburner. But like you said, were talking minimal FPS gain. So why stress your card further than it wants to go for 1-3 fps gain, if that, and potentially causing the card to give up and die?  I'd just max the power and temp target as I said before and then overclock the memory to the max it can without artifacts or crashes and leave the gpu core clock at +0mhz.

Precision xoc is very very buggy. Msi afterburner is known to be better. Why evga hasn’t fixed their software yet, no one knows. Been a long time since even a minor update for anything.

Because on Pascal, you gain more benefits from higher memory speed, especially when you let GPU Boost 3.0 do its thing with the core clock speed.  I have noticed less scores in 3dmark firestrike when fiddling with core clock speed, than I have fiddling with memory clock speed.