-Decrease room ambient temperature via air conditioning, which in turn decreases case ambient temperature. The latter is always warmer.
Case ambient affects all cooling devices at about a 1 to 1 ratio. For example, if someone else has the exact same hardware as you, but their case ambient is 22C and yours is 30C, then operating temperatures of your devices can be expected to be about 8C higher.
-Increase fan rpm to try and force more air throughout system.
-Cap frame rate to monitor refresh rate. This method decreases the amount of power used.
-Decrease gpu power limit to 90% or so. This method doesn't do anything if the gpu isn't using up to that level of power to begin with.
What about undervolting? Is that no longer a thing, with modern GPUs?
The problem I've observed with this method - at least on Nvidia gpus - is that the Gpu Boost algorithm makes it difficult to validate stability of an undervolt. Like, if you're trying to find the lowest possible UV for your card, good luck with that.
It can freely change settings if it runs a few degrees warmer than usual, or it detects instability in the manual settings, which is the opposite of what you'd want it to do during such testing.
It doesn't crash like a bad cpu overclock does either, and can keep chugging along like nothing's wrong. You'd have to closely watch graphs to see if it's hiding anything.
I've been advocating using the power limit slider and bumping up the core clock a little because of it.