AMD’s Latest Patent Unveils Innovative “Chip Stacking” Technique for Enhanced Die Utilization

AMD has recently filed a patent that hints at some exciting developments for its future Ryzen SoCs, particularly in the realm of multi-chip stacking, which could pave the way for enhanced die scalability.

In a move that underscores AMD’s innovative streak, Team Red is considering adopting “overlapped” chip stacking, placing smaller chiplets beneath a larger die within a single package. Known for trailblazing features like the “3D V-Cache” in their “X3D” lineup of processors, AMD isn’t resting on its laurels.

Reports of a new patent surfaced via Twitter user @coreteks, revealing that AMD is exploring what is being called a “novel packaging design.” This approach is set to revolutionize chip stacking by minimizing interconnect delays and delivering significant performance improvements.

The patent describes a creative stacking method where smaller chiplets are partially overlapped with a larger die. This strategy allows AMD to maximize the use of the contact area, thereby enabling more chiplets—and consequently more functions—to be packed into a single die. This room for additional chiplets suggests more cores, larger caches, and expanded memory bandwidth can be packed in, all without needing a larger die, which could propel performance to new heights.

A posted image on Twitter illustrates how future Zen SoCs might be structured. The concept revolves around partially overlapping chiplets to foster compactness and efficient interconnections. The dotted lines in the image represent a larger die layered over smaller chiplets, indicating a potential pathway for future designs.

A particularly exciting element of this approach is the potential for reduced interconnect latency. By bringing components closer together through overlapping chiplets, communication speeds up. Moreover, power gating becomes more manageable, as the design allows for individual unit control in the segmented chiplets.

AMD stands out as a frontrunner in implementing multi-chiplet designs, which they have applied not only to CPUs but also to GPUs. Previous reports have covered AMD’s exploration of multi-chiplet GPU configurations, suggesting a broader strategic move away from traditional monolithic designs. Given this trend, it’s not far-fetched to anticipate AMD employing a similar strategy for the mainstream Ryzen SoCs, echoing their “X3D” CPU approach. However, only time will tell.

For AMD to maintain its competitive edge, particularly against a formidable rival like Intel, innovation in the CPU space is crucial. By embracing “multi-chiplet” designs, AMD is well-positioned to lead the way in advancing CPU architecture and performance.

Feel free to share this intriguing news story with others on Facebook or Twitter!