Forcing GTX 680M clocks to stay at overclocked speeds?

Hi, I'm using a Clevo P150EM with the NVIDIA GeForce GTX 680M.

Recently, I have been putting a lot of time into a game of which the GPU utilization is not optimal. It hovers between 40-60% at best and can go as low as 30%. I tried overclocking the GPU so that I would get more FPS out of the game but strangely the Core Clock keeps fluctuating between the default 3D clocks of 758MHz and my overclocked speed of 1000MHz. It seems like the GPU driver does not allow the card to stay at overclocked speeds if the GPU usage isn't high enough.

I've gone ahead to set the card to 'Prefer Maximum Performance' for the game .exe in the NVIDIA Control Panel and it only works partially. The Core Speeds will now no longer fall below 758MHz but it still fluctuates between that, 1000MHz and anything in between.

Is there a way to force it to stay at 1000MHz at all times? I've tried to use NVIDIA Inspector to do so but it doesn't seem to work. It does overclock, but again it will not maintain the overclocked Core Speed. I've tried lowering the overclock and increasing it to around 1050MHz but that doesn't seem to be the thing that is causing the GPU driver to fall back to 758MHz randomly and then clock back up at times.

Can anyone help?

 

have u monitored ur gpu temps during that overclock? make sure ur gpu isnt just trying to stay alive by throttling while ure trying to push it over the brink

 

What game is it? Tbh it sounds like the game is really badly optimised and isn't using the full potential of the graphics I can bet it isn't supporting multi core CPU either.... Is it a ubisoft game? Because if it is its definitely the game not the machine.

 

It's staying around 79-81 degrees so I don't think it's that. Anyway it downclocks automatically at above 90 degrees usually.

It's Euro Truck Simulator 2 which should be optimized for multi-core and I'm even running a 64-bit .exe version of the game.

 

Its the game:

http://www.tomshardware.co.uk/forum/126040-13-euro-truck-simulator-usage

 

Hmm that might be the case but is there really no way to just force the drivers to keep the core clocks at max overclock? Something like tricking it into thinking that the overclock speeds are the default 3D speeds?

 

you could just start the rendering test in gpu-z and let it run in parallel to the game doesnt take up much gpu power but should keep the clocks at the levels theyre supposed to be

 

You could force the P-State with nvidia inspector.

 

that, of course, would be the wiser course of action to take

 

Well I thought so too but I did not want to make you feel bad

 

no worries, i can take being overruled / outsmarted by the meaker

 

Thanks guys for the suggestion. I've tried using this but it doesn't seem to be able to keep the GPU clocks at overclock levels still:

nvidiaInspector.exe -setBaseClockOffset:0,0,210 -setMemoryClockOffset:0,0,500 -setVoltageOffset:0,0,0 -setTempTarget:0,0,87 -setGpuClock:0,2,968 -setMemoryClock:0,2,2300 -forcepstate:0,2

Am I doing something wrong?

 

What temperature are you running at when the clocks go down?

 

About 81 degrees. Sometimes even lower. I'm not sure if the forcepstate number and the second number for setGpuClock and setMemoryClock are correct.

 

What happens if you scale the core speed back 30 mhz or so?

 

You mean something like this?

cd C:\Users\Oscar Lee\Desktop\NVIDIA Inspector\
nvidiaInspector.exe -setBaseClockOffset:0,0,180 -setMemoryClockOffset:0,0,500 -setVoltageOffset:0,0,0 -setTempTarget:0,0,87 -setGpuClock:0,2,938 -setMemoryClock:0,2,2300 -forcepstate:0,2

I've tried that and it still doesn't stick to max clocks. It does go there sometimes, but it just doesn't want to stay there, it'll just continuously seek between 758MHz (default) and the overclocked speed.

 

Hmm this doesn't work on all games I've tested. Maybe the command line isn't doing its thing.

 

Can you post up some logs?

 

What kinds of logs?

 

GPU-z sensor tabs or msi afterburner showing usage, clocks etc.

A log to file from GPU-Z would work too.

 

I've attached a GPU-Z log below. You will see that it was quite stable at the overclocked speed (968MHz) for quite a while until 11:54am where it suddenly decided to clock down. You can see that temperature is not the cause too.

https://dl.dropboxusercontent.com/u/22096700/GPU-Z%20Sensor%20Log.txt

 

My guess would be your VRMs are struggling to keep up on the card. What pads have you used and what thicknesses?

 

Oh silly me, it seems like -forcepstate should be 0,0 instead of 0,2. It works now.

However, I'm facing another problem. When I was using MSI Afterburner to set the overclocks, I could do +250MHz for the Core and +450 for Memory. However, using the NVIDIA Inspector command line overclocking, anything above +100 for Core and +75 for Memory would soon cause the GPU to either reset, the laptop to entire freeze or BSOD. Why is this happening? I've noticed that using the Inspector to overclock causes the Voltage to shoot up to 1.000V for short periods of time whereas using MSI Afterburner, it will be a gradual increase according to the clocks. For example, overclocking by any amount in Inspector would cause the Voltages to go to 1.000V occasionally while using Afterburner even with a +250 Core overclock will not cause the voltage to exceed 0.9875V at any point in time. How do I solve this?

 

Since you have forced the P-State that is the max clock and max voltage domain.

 

I'm sorry but I don't quite understand what you mean. Does this mean I can't use forcepstate to overclock any higher? If so, are there other methods to overcome this?

 

It means the clock and voltage wont change as that's what the different P-States are, the different clock vs voltage settings.

 

Correct me if I'm wrong, what you're saying is that the clock speeds are linked to the voltages through this method and so I can't go over those numbers since they'll hit the Voltage cap? If so, would flashing a VBIOS with a higher voltage increase the limit?

 

It should just be maxed all the time, what happens if you then set a clock in msi afterburner?

 

I've tried that. The driver crashes when a 3D application is run above that +100 value, though no freeze or BSOD.