MANAGER@SMITH.BITNET.UUCP (10/02/87)
Thanks to Art McClinton, Robert McQueen, Selden Ball, Dick Barry, and Tom O'Toole for their answers and advice. Briefly, my original question concerned how rigid DEC's specifications for air temperature and humidity in machine rooms were. This is a summary of the answers that I received. While there does not exist a temperature such that everything below it is ok and everything above it is death (obviously), pretty much everyone agrees that higher machine temperature means a higher failure rate, all other things being equal. This is because higher temperatures accelerate the breakdown of logic in IC's. A quote from Selden Ball: > The hotter you run your machine room, the more problems you will have. > The hotter an IC is, the faster the molecules migrate, > and the sooner its logic gate structures become unreliable. The whole purpose of controlling your machine room environment (i.e., air temperature) is to control machine temperature. Obviously, these are not the same, nor is there some kind of exact ratio. If, for example, the machine ventilation is less than it should be, cooling will be less effective. Most people recommend that the room be kept within spec, and that steps be taken to ensure an adequate airflow. In addition, Art and Dick brought up the issue of the RATE of temperature variation, which I confess I had not thought of. Quotes as follows. Art: > In addition to the humidity specification, you left out the temperature > variation specification. DEC does not like to see more that a 1 degree > per hour variation in room temperature. Exceeding the temperature variation > specification causes problems with magnetic media devices. Dick: > ...Also, reduced failure rates can almost be assured if the > temperature/humidity values of a computer room are held reasonably constant. > If your equipment can maintain a + or - 2 deg. spread, all the better. As Art pointed out, air conditioning is cheap...certainly less than the cost of a CPU. Someone told me once that are not covered for damage that results from "environmental failures", so that's what it would come to. Again, thanks to all. Mary Malmros Smith College MANAGER@SMITH.BITNET