Synopsys and Alchip Collaborate to Streamline the Path to Multi-die Success with Soft Chiplets

By Manmeet Walia (Synopsys), Manuel Mota (Synopsys), Erez Shaizaf (Alchip)

The shift from monolithic to multi-die design is inevitable — but that's not to say it's straightforward. Traditional monolithic design has clear limitations. Diminishing yield and high cost-per-die become issues when a design nears reticle limits. At this point, multi-die integration can break the system into a series of smaller dies and to overcome scale and cost issues of scale.

Hyperscalers and other high-performance computing companies have noted that chiplets enable collaborative design within a multi-die context that delivers cost advantages and "mix-and-match integration” across heterogeneous IP blocks. Unfortunately, the chiplet ecosystem has yet to be completely standardized. Chiplet-based design incurs challenges around packaging, power delivery, verification, timing, floor planning, security, testability, and thermal management.

Together with Alchip Technologies, a high-performance computing and AI ASIC company, Synopsys addresses these issues to deliver the ROI and physical benefits of a multi-die design.

Chiplet Flexibility for Multi-Die Chip Design

This Synopsys-Alchip collaboration combines decades of IP and EDA development and expertise. Synopsys provides silicon-proven and complete IP such as 112G and 224G Ethernet, PCIe 6.0, PCIe 7.0, and UCIe, as well as industry-leading design and verification EDA tools. Alchip contributes its high-performance computing (HPC)-optimized physical design methodology, large-scale high-speed interface IP integration, and 2.5D advanced package design capabilities.

The growing demand for IO and memory chiplets—small integrated circuits (IC) that combine with other chiplets to make up a bigger system-on-chip (SoC)—drives the collaboration. The process calls for carving out the functionality contained within a monolithic die and transferring it to a smaller companion unit.

Click here to read more ...