Be watchful if you have a GeForce card–it might not be entering low power mode

Today I received a chat message from a friend, linking me to a forum post where a MSI GTX 970 card owner was complaining about the high idle temps of his new graphics card. It is a known fact that MSI decided not to spin the fan on Twin Frozr V cooler until the GPU temps exceeded 65C, just like the cooler on Asus Strix cards.

As soon as I read the post, I knew what the problem was. It was a driver issue. I have observed the exact same thing before, where the GPU and memory clocks would not drop below the base values at idle. The last time I resolved this issue by cleaning up the drivers using Display Driver Uninstaller (DDU for short) and reinstalling the newest drivers. Well, they were exactly the same version as I had installed – so there was not bug with the driver, if you installing it fresh.

I was worried that my Palit GTX 670 would be acting the same way, thus I opened up GPU-Z and went into the sensors tab hoping to check what clocks the GPU was running at. I am afraid to say that the clocks were running at base clocks – or the overclocked base clocks.

Here is what was shown on GPU-Z

I am thankful to the guy who brought this issue to my attention because I would have paid a huge power bill being unaware of this issue. By the way, this happened once, and I ended up paying about JPY 2,000 more for electricity.

I quickly ran DDU, which mentioned that there was a newer version than the one that I had on my hard drive. Hence I downloaded the latest version and ran it. It booted the system into the Safe Mode and asked me to select the operations it should perform, and I selected the ones that made sense. I apologize for not noting down which options I selected. It then removed the drivers and automatically booted the system back in to Windows. I remember vaguely that one of the options I selected was to disable automatic NVidia driver download via Windows Update. I am unsure if it would disable downloading of all the drivers using Windows Update or just the NVidia drivers, but it did not matter as I would always download the drivers manually.

Then I downloaded the latest driver (344.11) from NVidia website (the same version as I had previously installed) and installed them. As usual, I installed the bare minimum – that is the driver (which cannot be unchecked from the list), GeForce Experience and PhysX driver. I also checked “Perform a clean installation” option in the installer.

After the driver was installed, I loaded GPU-Z once more and noticed that it had fixed the issue. The clocks were down, the voltage was down and the temps were down as a result. By the way, I did not have to reboot to make it work. I just worked as soon as I installed the driver.

I know that the temps are still kind of too high for a GTX 670 that uses a non-reference cooler. I think I need to clean the cooler and reapply thermal paste. But the stupid screws Palit have used to mount the cooler cannot be unscrewed with the tools I have at hand and I had been postponing it all this time.

I did not restore the overclock that I had before, but I hope it would not bring back the issue. I will let you know if something breaks when I overclock the card as well as the display. If I do not get back, that means everything is fine.