Chiplet Architecture for Next Gen Infrastructure
What's driving the deep interest in chiplet architecture? How does it scale, and is there a role for open interfaces?
Bapi Vinnakota, ODSA Project Lead from Open Compute Project explains:
- Chiplets are currently the fastest and most efficient way to build heterogenous systems, integrating different types of compute into one package. This is essential for accommodating the ever-changing AI workloads.
- The IO bandwidth, both into memory and in terms of raw network, needs to increase dramatically to keep up with the growing demands of data centers. Companies associated with the ODSA are working on solutions to this challenge.
- Open interfaces are believed to be the best way to integrate systems, creating composable systems that integrate chiplet devices from multiple vendors into one usable whole. This is a fundamental belief at the OCP and is key to meeting the requirements of future data centers, especially those with AI-intensive workloads.
Related Chiplet
- Interconnect Chiplet
- 12nm EURYTION RFK1 - UCIe SP based Ka-Ku Band Chiplet Transceiver
- Bridglets
- Automotive AI Accelerator
- Direct Chiplet Interface
Related Videos
- Architecting Next Generation Compute and I/O Chiplet for AI/ML, Cloud and Edge Platforms
- Rethinking chip(let) design for next generation ADAS applications
- TSMC 3Dblox™: Unleashing 3D IC Innovations for the Next Generation of AI, HPC, Mobile and IoT
- Chiplets for Automotive: Scalable Compute with Next-Gen Packaging
Latest Videos
- Accelerate 3D IC designs with Innovator3D IC
- On Package Memory with Universal Chiplet Interconnect Express (UCIe): A Low-Power, High-Bandwith, Low-Latency and Low-Cost Approach
- Breaking down 50 million pins: A smarter way to design 3D IC packages
- Optimizing Data Movement in SoCs and Advanced Packages
- Cache Coherency in Heterogeneous Systems