The Toronto Wireline Workshop

Abstract:

As industry is starting to deploy systems based on 224Gbps/lane and growing pains are becoming more apparent, AI companies are clamoring already for more bandwidth. A first look at 448Gbps is already highlighting the enormous challenges of running even very short links on copper. However, industry doesn’t seem yet ready to throw the towel on pluggable modules as gateways to optical interconnect given the risks and limitations of <a href="http://CPO.

Some” target=”_blank” title=”CPO.

Some”>CPO.

Some of the questions that will be covered:

– Can we clearly articulate what are the priorities in developing the next generation of interconnect for AI?
– What are the main obstacles to the adoption of short reach optics to replace the last inches of electrical connectivity?
– What can we learn from advances in C2C and in particular by the success of UCIe?
– Does it make sense to still have pluggable modules and what is the best way to do that?
– Was LPO a success or a bust and can we move forward with it?
– What are the promises and possible pitfalls of CPO?

Topics touched upon:

– System level (Computing, AI) requirements & considerations (architectural trends, efficiency, bw, latency, cost <a href="http://etc.)
-” target=”_blank” title=”etc.)
-“>etc.)
– Optical and electrical energy efficiency considerations
– HW technology developments and limitations (interposer, packaging, connectors, cables)
– System trade offs analysis (retiming vs. linear, vs. CPO)
– Chiplets and co-packaging technology
– Analysis of latest OIF development on next generation interconnect

Room: GB202, Bldg: Galbraith Building, University of Toronto, 35 St George Street, Toronto, Ontario, Canada