ITherm 2012

Abstract:  With installed footprints growing rapidly, data centers are requiring larger percentages of total utility capacities (electricity, water and sewer) internationally, as well as becoming a larger percentage of companies costs. In this talk, Noteboom will compare different generations of internet data centers and demonstrate the successful progression of improved efficiencies in each, as well as presenting current challenges and goals going forward.  He will also demonstrate process of building data centers that cost less from both from a capital and operational standpoint-- all while saving utilities, lessening environmental impact and delivering more processing power per watt.

Biography: Scott Noteboom serves as Yahoo!'s Sr. Director of Data Center Engineering & Operations. Since joining Yahoo! in 2005, he has served as chief design architect and a founder of the companies data center self construct / operate initiatives. He also managed over 10x growth of the companies data center / compute operating footprint, leading teams that installed and support multi hundred thousand hosts. Scott's duties include managing all aspects of the data center lifecycle-- from design, construction, operations, to decommissions.  His passion centers on shaping, understanding and managing initiatives related to large scale data center / compute platforms, and shaping more efficiency into those platforms. Scott has published numerous papers and articles, innovated multiple technology patents and has been a speaker on the subjects of high performance, efficient large scale data center / compute platforms. He has more than 15 years experience in the engineering and design of data centers and internet platforms.

Prior to joining Yahoo!, Scott was Director of Data Center Operations at AboveNet, managing the worlds 2nd largest data center collocation footprint at the time. He also founded several companies in the Internet Service Provider (ISP,) and Voice over IP (VOIP) space-- one that sailed and one that failed. He began developing published software and running a dial-up bulletin board service (BBS) in his teens.

Abstract: The ubiquitous trend towards a connected, digital, always-ON lifestyle is driving the development of electronic devices that have smaller form factors, higher performance requirements, along with constraints on power consumption and cost. The result is a constant upward pressure on the amount of heat that must be removed and also the heat flux density at the silicon, package and system level. Although this general trend holds for applications in the computing arena, gaming console arena, consumer electronics arena as well as in the portable electronics arena, the relative constraints on the cost, size and complexity of the thermal solution, power available for cooling, acoustic considerations, thermal design power vary widely between the different applications.

For example, in portable consumer electronics, the problem of dissipation of heat is compounded by devices that are increasingly smaller, flexible, closer to the body, and requiring ever higher processing power to perform the multitude of functions driven by convergence. Heat dissipation in high-end laptops that are also ultra-portable adds considerable expense in the design and manufacture of these devices. A rough rule of thumb until recently was that processors exceeding the 40-W range result in a cost penalty of approximately a dollar for every additional watt (Tiwari et al. 1998) in computing applications. The heat dissipation problem is further exacerbated in portable electronics because active cooling methods such as fans, refrigeration, etc. carry a very severe cost and battery power requirement penalty. There is also a desire to keep portable electronic products cooler than mobile computing products because of their proximity to exposed skin, which is sensitive to heat.

Traditionally, the electronics community has tried to address thermal management primarily at the system level �Eusing various kinds of cooling fans and heatsink combinations to regulate the temperatures inside computer enclosures.  Removal of heat was not glamorous and almost an afterthought to be addressed primarily by packaging engineers at the system level.  But higher circuit densities, higher levels of integration and advances in 3D packaging are resulting in chips that run so hot that the risk of not being able to make a usable product using the latest Si advances can become a very real show-stopper. 

This has galvanized new research into chip designs, novel materials and thermal solutions that could allow electronics to run much cooler and remove the roadblocks presented by thermal management.  For instance, the thermal design power of microprocessors ten years ago was less than 20W while today’s CPUs dissipate more than 100W of heat.  This problem can be much worse if hot spots are present within the chip, and heat fluxes higher than 1000 W/sq. cm will need to be overcome.  Likewise, although 3D packaging architectures with chip and package stacking promise higher integration and lower cost, they are difficult to cool because the thermal path is much longer than in 2D packaging.  The fear is that in the worst case, any theoretically expected performance increase will be left unrealized if thermal issues aren’t addressed.

Another unpredictable constraint that is often overlooked is government regulation forcing OEMs to market energy efficient devices to drastically cut the energy consumed and ameliorate global warming trends.

The tools available to the thermal management engineer to address these issues can be primarily categorized as either passive thermal management techniques or active thermal management techniques.  The passive techniques exploit conduction (heat spreader, thermal interface materials and greases), natural convection (finned heat sinks, ventilation slots, liquid immersion cooling), radiation (paints, coating, and mechanical surface treatments) and phase change (phase change materials, heat pipes, vapor phase chambers).  Active thermal management techniques that are exploited include forced convection (fans, nozzles), pumped heat exchangers, and refrigeration (vapor compression, thermoelectric coolers).  These general tools can be further categorized as in the macro, micro and nano level solutions. For example, one promising approach is to incorporate materials with high thermal conductivity into the chip as specialized heat spreaders to mitigate hot spots.  Candidate materials include diamond (thermal conductivity of 1kW to 2kW per meter per K) and carbon nanotubes (3kW to 3.5kW per meter per K) or graphene (single atomic layer of graphite atoms in honeycomb lattice).  Researchers at Intel demonstrated success with micro-scale Bi2Te3 thermoelectric coolers to eliminate hot spots in the chip when placed between the chip and heat spreader.  Another approach being investigated is the use of quantum dots (silicon nanowires comprising of silicon and silicon germanium).  Yet another approach is to generate power either from the system’s waste heat or by harnessing thermal energy from the environment.

The field of thermal management of electronic devices has entered a very dynamic and fast moving phase and attracting the interest of materials scientists, chip designers, nano physicists in addition to the usual mix of mechanical and packaging engineers.  The talk will illustrate the diverse and vibrant approaches at the cutting edge of computing, consumer and portable electronic device thermal management.

Abstract: Consumer demand for ever-smaller, ever more powerful electronic devices has put manufacturers in a “cooling quandry.�EThey need the thermal efficiency of air cooling without the acoustic downside of fans. Smaller fans need to work harder and run louder to move as much air as large ones, which makes them far from ideal for cooling projectors, ultrathin laptops and other devices that require quiet, effective thermal management in a small space.

Other traditional alternatives don’t work either. Passive cooling can’t keep up with the thermal output of a modern CPU, for instance. Water cooling is bulky and prohibitively expensive.

So what is the alternative? This presentation will discuss applications for a revolutionary technology called silent air cooling, which uses an electrical field to generate airflow. It will also discuss products that can make do with existing cooling technologies. Namely:

• iPods, iPhones and other electronics with low heat output �Epassive cooling is OK
• server farms �Eliquid cooling is acceptable
• desktop and laptop computers, projectors, cable set-top boxes: air cooling required, silent air cooling preferable
• LED lights �Epassive cooling insufficient to dissipate heat generated by LEDs internal electronics; LED fixtures need a compact, quiet technology to cool sufficiently; silent ACT is the ideal candidate

Biography: Liam Goudge has served as a Senior Vice President of the Company since March 2005. As Senior Vice President, Silent Air Cooling (previously Emerging Markets and Technologies) Mr. Goudge is responsible for development of the company's innovative thermal management solutions. From 1994 to 2005, Mr. Goudge worked at ARM Holdings plc, where he held a variety of roles in licensing, sales, product marketing and business development. Most recently, Mr. Goudge served as ARM's director of worldwide business development, where he was responsible for guiding corporate strategy and development, as well as merger and acquisition activity. Prior to ARM, Mr. Goudge held managerial positions in marketing at Texas Instruments, in both France and England, where he worked on wireless products for the early GSM cellular market. Mr. Goudge holds a Bachelor's degree with honors in electronic engineering and French from the University of Nottingham, England, and a postgraduate diploma in marketing from Cambridge College of Marketing