A while ago I started a thread about weird temperatures on idle. I thought that this is caused by driver issues and after I've installed a new driver, the temps were down again. However, after I did a Windows re-install (inb4 haters: no, not a virus or any BS like that. Mirror's Edge PhysX gets broken by Mass Effect 1 and none of the described fixes worked for me. Blame EA for not being able to figure out their $h!t so that one of their games messes up another!), the temps were weird again.
But now, I believe I've found a real solution and I want to share my experience with the community, as I've seen some people having the same problem on nVidia forums as well as on some other websites.
Basically, my idle temps are around 33-35C, but before I've seen them around 40-45C with 0% GPU usage. There's a setting in nVidia control panel:
Selecting "Adaptive" will make sure your card stays cooler on idle (30-35C). However, there's a couple of downsides to it that "Performance" mode (idle temps 40-45C) doesn't have.
1) You might experience rare crashes in Chrome or Firefox if using hardware acceleration. (Can't say anything about Opera, IE or [strike]Safari[/strike], I don't know if they have hardware acceleration).
2) Same rare crashes might occur on the desktop.
3) The crashes never occur while gaming, however, because the card works at full frequency and the crashes occur because of some half-assed frequency nVidia cards are using on idle with Adaptive performance mode.
The reason for the crash is Win7's TDR feature (Vista has it, too). It detects the GPU not responding and "restores" the driver, during which your screen goes black for ~5 seconds and then you get the notorious "Driver stopped responding and was successfully restored" message. By default, the GPU response timeout is set to ~2 seconds in the registry. I've read about people recommending to increase it to 8 seconds but there's a mixed response - for some it fixes the problem, for some it doesn't. I believe that the effectiveness of this fix and frequency of the occurrence of the issue also depend on your card's frequencies - i.e. OC'd/non-OC'd, how high compared to stock, etc.
The problem occurs on both reference and OC'd cards, that's for sure, but it looks to me like the TDR fix works better for the reference card owners.
In any case, you've got two things here. Want lower idle temps and occasional crashes - select Adaptive mode. Want stable experience and don't mind 10C higher temps - got for Performance.
Hopefully this will help some nVidia card owners I've read about a similar issue with AMD cards - the driver restoration, possibly also caused by TDR - but I'm not aware of any power mode settings for AMD cards that might fix that, since I never had an AMD card. If someone knows if there's such a thing, feel free to post it below!
But now, I believe I've found a real solution and I want to share my experience with the community, as I've seen some people having the same problem on nVidia forums as well as on some other websites.
Basically, my idle temps are around 33-35C, but before I've seen them around 40-45C with 0% GPU usage. There's a setting in nVidia control panel:
Selecting "Adaptive" will make sure your card stays cooler on idle (30-35C). However, there's a couple of downsides to it that "Performance" mode (idle temps 40-45C) doesn't have.
1) You might experience rare crashes in Chrome or Firefox if using hardware acceleration. (Can't say anything about Opera, IE or [strike]Safari[/strike], I don't know if they have hardware acceleration).
2) Same rare crashes might occur on the desktop.
3) The crashes never occur while gaming, however, because the card works at full frequency and the crashes occur because of some half-assed frequency nVidia cards are using on idle with Adaptive performance mode.
The reason for the crash is Win7's TDR feature (Vista has it, too). It detects the GPU not responding and "restores" the driver, during which your screen goes black for ~5 seconds and then you get the notorious "Driver stopped responding and was successfully restored" message. By default, the GPU response timeout is set to ~2 seconds in the registry. I've read about people recommending to increase it to 8 seconds but there's a mixed response - for some it fixes the problem, for some it doesn't. I believe that the effectiveness of this fix and frequency of the occurrence of the issue also depend on your card's frequencies - i.e. OC'd/non-OC'd, how high compared to stock, etc.
The problem occurs on both reference and OC'd cards, that's for sure, but it looks to me like the TDR fix works better for the reference card owners.
In any case, you've got two things here. Want lower idle temps and occasional crashes - select Adaptive mode. Want stable experience and don't mind 10C higher temps - got for Performance.
Hopefully this will help some nVidia card owners I've read about a similar issue with AMD cards - the driver restoration, possibly also caused by TDR - but I'm not aware of any power mode settings for AMD cards that might fix that, since I never had an AMD card. If someone knows if there's such a thing, feel free to post it below!