What if someone told you that more than 10% of the dollars you spent powering your servers did no useful computing work? This sounds wasteful, however, that 10% is spent spinning the air movers that remove the heat generated through power conversion and powering of silicon and peripherals.
As a thermal and acoustic architect for servers itâ€™s my goal to reduce that 10%, but the electrical energy going into a computer is converted to heat. That heat must be removed to ensure components stay within their temperature limits ensuring data integrity and long term reliability.
For years and years the focus was on improving performance, even if that 10% sometimes pushed up to 20% in extreme cases. In todayâ€™s environment, performance requirements must now be balanced with the power required to create that performance. This change has driven a wide array of silicon features that can create that balance but the overall server cooling design must adapt to take advantage of those features while using the most cooling-efficient thermal components.
The session ETMS002, â€˜Server Cooling Design Optimization for Low Power Consumption', at the upcoming IDF will provide answers demonstrating how servers are becoming more 'cooling-efficient' while ensuring that performance can be maximized. Cooling tradeoffs based on board layout, heat sink selection and usage of silicon thermal management features will be discussed and quantified with regard to their impact on potential power savings.
Whether you are concerned with server design itself or with becoming more informed on purchasing decisions, this IDF session will enable you to understand the cooling and thermal management implementations that will save energy and reduce total cost of ownership.