Sponsors: AMD, CISCO, NSF, MARCO IFC

The performance density of modern electronic systems (including desktop and mainframe computers and routers) is severely limited by the temperature rise at hotspots. Hotspots are regions of especially high computational density at which heat fluxes can exceed several kW/cm2. The hotspot problem is particularly important and challenging for complex 3D circuits integrating memory, sensing, and photonic functionality.

Stanford has been focused on hotspot cooling since the late 1990’s with contributions on targeted microfluidic cooling and concept advancement for 3D circuits. The latest realization of this is shown in the second figure on this webpage, which is a schematic of a modular microfluidic cooling strategy for 3D circuits and systems. At present, we are focused on the coupling between hotspots and temperature distributions, including the development of experimental methods for measuring the temperature fields in practical devices. We use a diffraction-limited infrared microscope system for thermometry integrated with a fluidic convection system that allows chip operation at targeted power levels. In related research, we are studying the coupling between transient, localized heating on a chip and active chip-integrated cooling using either microfluidics or thermoelectrics. This work shows explores the limits in dimension and timescale of local regions of computational density at which the hotspot temperature rise may be aggressively suppressed with targeted cooling.

In 2001 and 2002, this group authored series of patents that were the foundation for a major startup company, Cooligy Inc, which developed novel and sold microfluidic heat sinks and specifically targeted hotspot cooling as a deliverable. This company supplied heat sinks for Apple desktop computers in 2005 and 2006. Cooligy was acquired by Emerson in 2005.

Picture1_2
IR-transparent microfluidic heat sink mounted on AMD Thermal Test Vehicle (TTV)
Picture2_2
Cross-section of IR-transparent microfluidic heat sink.
Picture3_0
Inverse solution thermal transfer function.
Picture4
Inverse solution method applied to temperature profile (above) to resolve underlying heat flux profile (bottom). Note: results shown for previous lab title.