Building products using chiplets involves more than treating them as hard IP cores. Many open questions surround the field, explored by panelists in a MEPTEC conference online.
The added complexity of managing reliability as chiplet-based designs become more common will need to be managed using digital-twin techniques, says a professor working in the field.
In a panel session at VLSI Symposia, AMD described how the economics have come down strongly in favor of multichip integration for multicore server processors.
Achronix is introducing an FPGA architecture that pulls a full network-on-chip into the programmable-logic fabric combined with hardened matrix-math processors for AI.
Large-scale MCMs and novel device architectures bookend the papers on machine learning at VLSI Symposia in an event that will also cover chiplet integration and other topics.
An Open Compute Project group working on multichip integration sees a combination of parallel and serial interfaces being important for interchip communication.
The ODSA Workgroup formed by Netronome and others is looking to adopt the PIPE standard for interconnecting chiplets as it starts work on a proof-of-concept module.
Data-center networking specialist Netronome has recruited a number of silicon makers and IP suppliers to a standard for chiplet designs that can be used in SIPs for edge computers and servers.
View All Sponsors