Using power-aware SI analysis and thermal-aware PI analysis as part of a system design and signoff methodology for GDDR6 designs.
Graphics processing units (GPUs) and graphics double data rate (GDDR) memory interfaces are essential to graphics cards, game consoles, high-performance computing (HPC), and machine learning applications. These interfaces enable data transfer speeds of over 665GB per second today and will continue to support well over a terabyte per second (TBps) in next-generation GDDR interfaces. Signal integrity (SI) and power integrity (PI) are becoming intertwined with the thermal issues caused by ultra-fast data transfer rates, ultra-low voltage swings, and high-density GDDR6 designs often implemented on silicon interposers. This paper describes using power-aware SI analysis and thermal-aware PI analysis as part of a system design and signoff methodology for GDDR6 designs.
Click here to read more.
100% inspection, more data, and traceability will reduce assembly defects plaguing automotive customer returns.
Engineers are finding ways to effectively thermally dissipate heat from complex modules.
Increased transistor density and utilization are creating memory performance issues.
Lots of unknowns will persist for decades across multiple market segments.
FPGAs, CPUs, and equipment receive funding in China; 98 startups raise over $2 billion.
Suppliers are investing new 300mm capacity, but it’s probably not enough. And despite burgeoning 200mm demand, only Okmetic and new players in China are adding capacity.
100% inspection, more data, and traceability will reduce assembly defects plaguing automotive customer returns.
From low resistance vias to buried power rails, it takes multiple strategies to usher in 2nm chips.
Some of the less common considerations for assessing the suitability of a system for high-performance workloads.
Manufacturing 3D structures will require atomic-level control of what’s removed and what stays on a wafer.
Different interconnect standards and packaging options being readied for mass chiplet adoption.
Engineers are finding ways to effectively thermally dissipate heat from complex modules.
Disaggregation and the wind-down of Moore’s Law have changed everything.
Leave a Reply