I read an interesting article about Google running servers above recommended ambient temperatures. This is contrary to popular recommendations propagated on the Internet and other sources. So, this summer I placed a headless server in my garage and witnessed internal temperatures of ~140-154F and ambient room temperatures above ~110F. I set the thermal shutdown temperature to be > than 160F, so the server wouldn't shut down. This particular server is about 5 years old, so I didn't have much to lose. The interesting thing is that it survived just fine over the summer (including the supposedly too cold ~25F low last night). This particular server also runs a real load of > 50% CPU 24/7 handling real-time video processing. Some other factors to note is that there's zero air flow and ambient temperature and humidity are variable in this environment, so this definitely isn't your typical server room. The graph below denotes average high and low temperatures in Parrish, FL:

So you are probably asking yourself why I would do such an experiment. Well, when you look at the cost of power and cooling it usually exceeds the hardware costs. My goal was to go above and beyond what Google has done with it's data centers and really torture this server and see if it would survive while saving me on cooling costs. If it survives into this summer then my application server will be the next victim. All this saves me from cooling the heat discharged from these servers.

Note the internal temperature while the ambient temperature is in the low 60s (F).

Ambient temperature currently in the low 40s (F).

Hopefully my experiment will pay off for me personally (my server also has AMD® Cool'n'Quiet technology, so the fans run only as required), but you have to wonder how many companies are wasting energy based on bad information.