System Scaling Key to Semiconductor Progress
System scaling technologies exist but can't go mainstream until IC designers and manufacturing experts cooperate.
Advanced semiconductor nodes below 10nm promise tens of billions of gates and memory bits, yet the current strategy of transistor scaling has reached its practical limits for most applications. Only the largest companies are likely to invest in SoCs at 7nm or 5nm where high-volume applications like today’s smartphones offer the potential payback to justify the risks and high development costs in the range of hundreds of millions of dollars.
System scaling may be the answer. While it is not a new concept, it is taking center stage as the market looks for integration alternatives to traditional transistor scaling. It offers another path to pursuing Moore’s Law by moving the integration focus from the transistor to the system.
System scaling — often referred to as multi-die IC — encompasses the concept of integrating complex systems at the functional/building block level as opposed to the transistor level. It is based on the integration and packaging of multiple die using a variety of different advanced packaging technologies including, 3D IC, 2.5D IC, wafer level packaging (WLP), and the like. System scaling does not replace the need for the SoC which will remain as a key functional component of the system or, in die-form, a key building block of a multi-die IC.
To read the full article, click here
Related Semiconductor IP
- nQrux® Root of Trust IP
- AXI to UCIe Bridge IP
- UCIe 2.x Controller IP
- SWI3S (SoundWire I3S Interface) Peripheral Controller Core IP
- OpenTitan-based RISC-V Secure Element
Related Blogs
- A Systematic Approach to Robust LVDS Integration in Advanced Node ASICs
- An Steegen's Secrets of Semiconductor Scaling
- How SerDes Became Key IP for Semiconductor Systems
- How Is the Semiconductor Industry Handling Scaling: Is Moore's Law Still Alive?
Latest Blogs
- From Bug Hunting to Engineering Methodology: Lessons from AI-Driven Development
- Manufacturing Intelligence: Turning EDA Data into Trusted Action
- Execute-in-Place: Getting More from Embedded NVM
- Rethinking RTL flows with AI-driven hybrid formal verification
- Arm and NVIDIA: Building the trusted compute foundation for the agentic AI era