Page 18 - ASME InterPACK 2017 Program

Technology Talks

Interfacial Imperfection Effects on Thermal Boundary Resistance in Electronic Materials
And Devices

Typical solutions to thermal mitigation of high power devices have traditionally relied on
engineering the thermal conductivity of substrates and submounts, such as diamond, to mitigate
unwanted temperature rises and device inefficiencies. Increased operating powers and
temperatures combined with efforts in heterogeneous integration have given rise to device
efficiencies being directly correlated to the thermal boundary resistances (TBR) at these
submount interfaces. In this talk, I will discuss examples in which the power densities at failure of
high power and high frequency devices are directly correlated to the TBR at the substrate
interfaces, and the role that defects, roughness and other asperities at the material contacts
have on the TBR. Through a survey of both computational and experimental literature over the
past decade, I will discuss the role of these interfacial imperfections on TBR, and how these can
lead to both increases in TBR, and, in some cases decreases in TBR through judicious engineer-
ing of the properties of these defects. I will then discuss the role of vibrational energies/phonon
dispersions and crystalline disorder on mode conversion, energy transmission, and the resulting
TBR.

TRACK 2: SERVERS OF THE FUTURE

FRIDAY, SEPTEMBER 01, 2017 9:15 am – 10:45 am
Room: Jackson, Second Floor

8-2-1 - Hyperscale Data Centers

Session Organizer: Prof. Yogendra Joshi, Georgia Institute of Technology

Hyperscale data centers have unique thermal challenges. High power chips are enabling
high-performance servers and networking switch and energy efficient cooling facilities are
required to support them. In this session, industry experts share their views on innovative
thermal management at hyperscale.

Ali Heydari Ali Heydari is Senior Technical Director and Chief Data Center Architect at Baidu, the largest
Baidu USDC search engine and AI company in China. In this role, he is server and data center architect in
charge of hardware and data center design, development and deployment in Baidu. Formerly,
18 he was Senior Hardware Engineer at Twitter where he was responsible for grounds up
development of Twitter’s data center ODM server development. Earlier, he was Senior
Hardware Engineer at Facebook where he helped in developing Facebook’s original OCP
server and data center products. Prior to that he worked at Sun Microsystems and spend about
10 years as Associate Professor of Mechanical Engineering at Sharif University of Technology in
Iran. He received his B.S. in mechanical engineering from University of Illinois, Urbana, M.S.,
Ph.D. in mechanical engineering and M.A. in applied mathematics from University of California,
Berkeley.

Driving Innovations in Hyperscale AI-Enabled Data Centers

Hyperscale data centers are estimated to exceed all high-end, midtier to localized servers
deployed in the next 5 years. Artificial intelligence and IOT world will be the driving force behind
the next generation of data centers with the highest adaptation in the hyperscale data center
market. In this presentation, a roadmap of AI-enabling server and data center hardware
technology and infrastructure with emphasis on TCO, reliability and performance is presented.