GPU Core Clock continually slows down while gaming

[Godlock]

I've been having this probelm ever since I got the Witcher 3 back when it came out. Whenever I tried to run the game without any prgorams running like firestorm or afterburner my game would play for maybe 5 minutes at most before crashing. I then tried running the game with afterburner and upon crash I found that right before the crash the game had reached a power limit of 139% (I have it set to stock 100% I don't overclock, none whatsover ,on any of my parts). I then used firestorm to set my power limit to 100% and still got crashes. Then I tried setting the power limit lower and ended up making the game playable, that is if I cap the framerate. If I set the framerate to 60 fps; however, I end up being able to play for a couple of minutes at 60fps before the framerate gradually goes down. Upon examination I notice in firestorm that my GPU core clock speed lowers as I play . The GPU temperature is at a very decent 70 degrees.
I have tried everything I can think of, graphics drivers, force constant voltage, increasing the power limit nothing seems to work.
It seems to me that the more intensive the game the more the graphics card will do this, I don't know why as it only does this with firestorm, in afterburner it will crash no matter what. Maybe firestorm is doing this to prevent it from drawing over the power limit? But why is it doing that in the first place?
My specs are
1xGTX 970 4GB GPU
Fx-8350 CPU
1 1TB SATA 2 Hard Drive
1 256GB Sata 2 Hard Drive
Gigabyte Ga 78 LMT motherboard
8GB ddr3 1333 MHz
EVGA SuperNOVA b2 750W PSU (Should NOT be a problem, actually just got this as I thought it might fix the problem, which it didnt)
One last thing Before I started playing Witcher 3 I ran very intensive games with this setup just fine including Crysis 3(Max settings) Skyrim with tons of graphics mods including realvision ENB and Dying light. But now when I try to run those games they exhibit the same behavior.

 

[NevaScurred]

Go into Nvidia Control Panel->Manage 3D Settings->Program Settings tab.

Select the game executable from the dropdown or click Add and locate the game .exe.

Scroll down to Power Management Mode -> set it to Prefer Maximum Performance.

If you do this for every game your clock speed will max out and stay maxed out while the game is running.

 

[Godlock]

Go into Nvidia Control Panel->Manage 3D Settings->Program Settings tab.

Select the game executable from the dropdown or click Add and locate the game .exe.

Scroll down to Power Management Mode -> set it to Prefer Maximum Performance.

If you do this for every game your clock speed will max out and stay maxed out while the game is running.

Just tried it. Sadly it didn't work stayed above 1000 for about 30 seconds then started getting lower and lower. When I cap the framerate to 30 the Core clock stays steady but I'd rather not have to do that for all my games.
Something else that I noticed in afterburner is that the Core Clock slowly declines as the Power limit fluctuates between 0 and 1(Max).

 

[Reaper_7799]

Never used firestorm but is it also an overclocking utility? If it is, you could try to uninstall all of those programs and just reinstall one. I know when I had precision x and afterburner installed, I think they sent conflicting settings to my card because it was acting funky and someone suggested to delete one of them. That was even with precision turned off but it still sends it's settings and such. After that it has worked fine. Just throwing a suggestion.

 

[Godlock]

Never used firestorm but is it also an overclocking utility? If it is, you could try to uninstall all of those programs and just reinstall one. I know when I had precision x and afterburner installed, I think they sent conflicting settings to my card because it was acting funky and someone suggested to delete one of them. That was even with precision turned off but it still sends it's settings and such. After that it has worked fine. Just throwing a suggestion.

Just tried this, sadly didn't work either. But I'm appreciative of any steps I can take at this point so thanks.
The really odd thing about this is that the clock goes down even though everything else is normal the temps and power supply % usage are all in good shape. I really don't understand what's going on.

 

[NevaScurred]

Go into Nvidia Control Panel->Manage 3D Settings->Program Settings tab.

Select the game executable from the dropdown or click Add and locate the game .exe.

Scroll down to Power Management Mode -> set it to Prefer Maximum Performance.

If you do this for every game your clock speed will max out and stay maxed out while the game is running.

Just tried it. Sadly it didn't work stayed above 1000 for about 30 seconds then started getting lower and lower. When I cap the framerate to 30 the Core clock stays steady but I'd rather not have to do that for all my games.
Something else that I noticed in afterburner is that the Core Clock slowly declines as the Power limit fluctuates between 0 and 1(Max).

I read the other response as well. Are you looking at your clock speed and temps through the OSD or minimizing the game? Stupid question I know. What does you cpu usage and temps look like?

I also second the idea of uninstalling everything but afterburner. Make sure all OC settings are zero except power limit.

Other things I tweak in Nvidia CP are optimize for single display and vsync off.

If that doesn't work maybe use DDU to completely wipe your display driver and install only physx and graphics driver (3d stuff only if you use it).

 

[Godlock]

I read the other response as well. Are you looking at your clock speed and temps through the OSD or minimizing the game? Stupid question I know. What does you cpu usage and temps look like?

I also second the idea of uninstalling everything but afterburner. Make sure all OC settings are zero except power limit.

Other things I tweak in Nvidia CP are optimize for single display and vsync off.

If that doesn't work maybe use DDU to completely wipe your display driver and install only physx and graphics driver (3d stuff only if you use it).
[/quotemsg]
sorry I meant to reply to the other guy. But yes I did monitor most of this. I just tested with both cpu and gpu usage and the findings are a bit odd. While the core clock is stable for a couple of minutes the gpu AND cpu usage are both around %70. However, after a couple of minutes my GPU usage spikes to 100% (seemingly for no reason, I just stand in a field rotating the camera a bit) and my CPU usage drops to about 50%. Ive tried uninstalling the display drivers many times as well as rolling back to an older driver.

 

[beekermartin]

I would install GPU-Z and monitor the sensors. Maybe it will show something no one has mentioned yet.

 

[Reaper_7799]

Try downloading the free version of 3dmark firestrike and try running it. Make sure you have your osd turned on and showing cpu usage, gpu usage, power % and temps for both of them. Let us know where each of those are at. Shouldn't take but about ten minutes to run it.

http://www.futuremark.com/support/downloads

EDIT: I know I see your temps are running at 70 but try it with this software and also what are your cpu temps? Are you sure you might not be getting some kind of processor bottleneck?

You also might try turning up the resolution and try playing a game at 1440P with ultra settings using DSR. This will make your graphics card do more of the work and might show that it is your processor or at least see if it will push your gpu to 100%.

And when you do run firestrike, put a link here to the online page it brings you to after it's done so we can see the results. It shows graphics score and physics score which is your processor.

 

[Eggz]

This sounds like a cooling issue to me. Check for dust on both your CPU and GPU cooler.

What CPU cooler do you have? Can you also say a little more about the airflow in your case?

 

[Godlock]

Ran Firestrike and the GPU usage kicked to 99% very fast (seemed to make sense due to detail) and the Power % was around 90 and the CPU was around 40% for the first two tests the CPU usage for the CPU test was around 95% and for the combined test all the values were normal(GPU usage fell to 55% and a bit above sometimes) though, if the test had been longer the clock speed WOULD have lowered as I ran a Heaven Test (All settings as high as they'll go) and the issue sprang up again. However, I have some sensor information (dont know what most of it means).
For Heaven:
The highest value for the Memory controller load was 52%
Bus interface load was 5%
VDDC hit 1.1250 V max
Memory used hit 1267 MB
Highest temp was 79 degrees (Though I've run games at this temperature before with no issues not alot but sometimes)
Power consumption hit 91%

if any of this helps

http://www.3dmark.com/3dm/7661010?

 

[beekermartin]

I believe when it hits 80c it will then began to down clock the GPU core. I would run the test again with the fan speed set to 100% to see if it can keep the temps under 80c at all times.

 

[Eggz]

Assuming your computer guts are clean and not caked with dust, the best way to test if you have thermal limitation from heat output is to eliminate airflow as an issue.

Go get a house fan. Open up the side panel of the case, turn the fan on high, point it right into the case, and then repeat the same things that caused problems before. Does it still slow down?

If not, you have a thermal limit; otherwise, it's something else and you'll have to keep digging. Try the fan test because it's easy and might give you an answer with very little effort.

 

[Godlock]

Just did the fan test on high problem still sprang up (also didn't cool the GPU any still stayed around the levels it had before). I got a new case pretty recently NZXT Phantom 410 and got some extra fans I have 2 exhaust fans on the top and one on the back and two intake fans one on the side and front. The GPU very Rarely hits 80 degrees (hasn't done so once while I've been monitoring it) and when the problem springs up the temperature is only around 69-71 degrees.

 

[Reaper_7799]

This goes along with eggz also but try this too. Go into afterburner or whatever you're using now and create a custom fan curve. This is just to test...set it to turn on at least 80-90% or 100% at say 60 degrees or lower and then try running heaven or try running your game. It's gooing to be loud but maybe for some reason your card is messed up and is throttling at the wrong temperatures or something.

EDIT:....okay if you're at 71 degrees it shouldnt be doing any throttling. Mine throttles but it's set at 83 degrees to throttle and I don't get to 83. I did until I set custom fan curve though.

 

[Reaper_7799]

Did I miss or what are your cpu temps when all this is happening?

 

[Godlock]

Just tried setting a fan profile and instead of the core clock going down the game just crashed before that ever happened. GPU-Z shows that the power consumption reached 108% and the VDDC reach a new high of 1.2120V. which sounds real bad.

 

[Reaper_7799]

Just tried setting a fan profile and instead of the core clock going down the game just crashed before that ever happened. GPU-Z shows that the power consumption reached 108% and the VDDC reach a new high of 1.2120V. which sounds real bad.

You might have to warranty the card, how old is it? It's sounding like it's some kind of internal issue with the bios or something maybe.

EDIT: Try one more thing too, I'm running out of suggestions. Go into afterburner and you know how the power limiter and the temp limit is keyed together to move up and down. click on the little key link then and it will unlock it so you can move them seperately. Keep your stock voltage and raise the temperature level from whatever it is set too and raise it to say 90 degrees. You aren't even hitting 80 so it wont matter but some cards throttle slightly even before you reach the temperature you have selected to start the throttling. Keep your regular fan curve but watch your temps.

What is that temp setting at now?

That evga b2 is a pretty decent psu too so I don't think it's psu issue but you could try running the other pcie cable to your card instead just to rule it out.

 

[Reaper_7799]

Try this last thing also, you might have but adjust only the power limiter to 110%. Don't increase voltage or anything else and see if that does anything.

 

[Godlock]

Sadly, I figured out the issue, the hard way.

It was a hardware issue and my gpu exploded while doing more tests (exaggeration, but I smelled burning) dont know what happened, bad luck I guess. I never overclock r do anything dangerous. Also incredibly unluckily is that the warranty doesn't apply to me because I didn't register soon enough. Just got this card in November.....sigh.

 

[Reaper_7799]

Sadly, I figured out the issue, the hard way.

It was a hardware issue and my gpu exploded while doing more tests (exaggeration, but I smelled burning) dont know what happened, bad luck I guess. I never overclock r do anything dangerous. Also incredibly unluckily is that the warranty doesn't apply to me because I didn't register soon enough. Just got this card in November.....sigh.

That's a bummer will they really not warranty it unless you are registered? Maybe they will make an exception but I made sure to register mine just in case. Sorry about that.

 

[NevaScurred]

Sadly, I figured out the issue, the hard way.

It was a hardware issue and my gpu exploded while doing more tests (exaggeration, but I smelled burning) dont know what happened, bad luck I guess. I never overclock r do anything dangerous. Also incredibly unluckily is that the warranty doesn't apply to me because I didn't register soon enough. Just got this card in November.....sigh.

Contact the manufacturer and tell them what happened, most times registration is mandatory for an extended/lifetime warranty. I don't remember what brand your card was but any of them should have a 1 year out of box warranty that doesn't require registration...even if it does the card is < 1 year old.