BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Photonics for Computing: from Optical Interconnects to Neuromorphi
 c Architectures - Nikos Pleros\, Aristotle University of Thessaloniki\, Gr
 eece
DTSTART:20190613T090000Z
DTEND:20190613T100000Z
UID:TALK126223@talks.cam.ac.uk
CONTACT:Microsoft Research Cambridge Talks Admins
DESCRIPTION:How should someone exploit photonics in computing? Simply repl
 acing the electrical with optical wires and increasing the datarate is the
  first and obvious answer\, but the idiosyncrasy of photons can lead to im
 proved architectures that can offer additional functionality in Datacom an
 d computercom environments. This talk will concentrate on how photonics ca
 n bring significant functional benefits in computing architectures\, spann
 ing from disaggregated rack-scale through disintegrated board-scale and ev
 en to the emerging neuromorphic platforms\, presenting how innovative opti
 cal switching\, photonic Network-on-Chip and optical neuromorphic function
 s can shape a radically new computing environment with increased granulari
 ty\, modularity\, performance and energy efficiency.\nWe will discuss the 
 main performance and energy challenges currently faced by the computing in
 dustry and we will present our recent research on photonic technologies to
 wards realizing resource disaggregation at all hierarchy levels spanning f
 rom rack- down to board-chip-scale computing. We will demonstrate high-por
 t count optical switch layouts based on the Hipoλaos Optical Packet Switc
 h architecture with up to 1024x1024 input/output ports that allow for low-
 latency values well below the 1μsec target of disaggregated DCs. We will 
 also demonstrate how silicon photonics and electro-optic PCBs can be utili
 zed towards realizing on-board resource disaggregation for multi-socket co
 mpute node applications\, holding the credentials to replace the dominant 
 QPI interconnect by offering >60% improvement in energy consumption with a
  single-hop flat any-to-any interconnect topology for >8-socket connectivi
 ty\, i.e. >100% improvement over QPI’s single-hop connectivity capabilit
 ies. Finally\, we will demonstrate how optics can be incorporated in the e
 merging paradigm of neuromorphic computing towards offering high-performan
 ce at a significantly lower power envelope\, presenting experimental reali
 zations of an optical sigmoid function\, a WDM-accelerated Convolutional N
 eural Network Architecture and an optical Recurrent and Gated Recurrent Un
 it (GRU).\nThis work has been carried out within the frame of the European
  H2020 projects ICT-STREAMS\, L3MATRIX and MASSTART.\n
LOCATION:Auditorium\, Microsoft Research Ltd\, 21 Station Road\, Cambridge
 \, CB1 2FB
END:VEVENT
END:VCALENDAR