I have it set up like this:
CUDA manager kicks it with fans at 100% whenever the card reaches 80 degrees. If, for any reason at all, this does not lower the temp sufficiently, it starts clocking down at 81 degrees. And if that still does not help, at 100 degrees CUDA manager terminates the mining proces. I'm toying with lowering that last treshold to 90 degrees, but honestly, even the second line of protection was never needed, as my card only reaches the 80 degrees when I have had my windows closed for a couple of days, or perhaps one day when it is warm summer weather outside, or when it is really warm summer weather outside when I do have my windows open.

Normally it fluctuates between 73 degrees with fans around 45% during the night with windows open, to 78 with fans at 62% in the evening, when I have had my windows closed since the morning. But it is pretty cool weather, with a max temp of 18 degrees during the day, and during last hot period of a few weeks ago, I had to downclock my card a bit, as it was hitting the 80 degrees to often, and the fan was getting annoying.

As for the VRM temps, MSI afterburner doesn't measure that, but I never had issues, so I am assuming that's fine.

You can always google the given safe temps by the manufacturer and aim for that at max. I myself run a Gigabyte windforce 3x GTX 770, and I'm a part-time miner, I do mostly the same as you, and since my computer is pretty responsive during mining, even with the CPU assisting (i7-4770k), I can still do my university stuff, nicely optimized, or those that aren't big resource hogs, flash games and youtube vids, so I do that during mining as well.

its for GPU my VRM run at 98 most of the time.

98 is way too high for the VRM (well I think so... can't monitor mine on the 780ti)

Do you mine everything basically 'solo'? or is this just the way that you calculate.

I have never tried solo mining... but it would be so nice to not have to really deal with all the pools.

I have 6x 750ti up and running as of last night, so would you think that these numbers would be straight conversions (as estimates) of performance? Or is there some level in solo mining where you will never get your own block done?

Thanks for any input, I always find your charts useful.

Dont count me on it. If your vrm is high and shouldnt be then maybe an RMA? Mine is watercooled so my temps never hit 40'C

what is the temp in idle ? Could be also some bad calibration of the sensor ?
Or some default with the cooler...

I will check and update here the idle temps

Now after 5 mins of idling GPU is at 40C and VRM is at 58C. My system is stable until now and anyways I am getting a new rig in few days. At peak GPU was 78C and VRM was 95C. Although I must mention temps here are around 33-34C.

Since we're on the topic of temps,my card's doing something odd. I start cuda/ccminer and then OC the card. Fine,the core clock goes up and stays there as long as the mining continues. Then there's a small connection error and the MSI stats of all the clocks and temp drop to 0. On restarting,the mem clock goes back to the OC'd value but the core clock is underclocked,ie,even lesser than it's original clock. Why is this?