NVIDIA CEO Asserts GAA-Based Technologies Will Deliver 20% Performance Increase, Emphasizing Ongoing Focus on Architectural Innovations

NVIDIA’s CEO Jensen Huang expresses a positive outlook regarding the potential of Gate-All-Around (GAA) technologies to significantly enhance performance in the coming years.

The Integration of GAA Technologies in NVIDIA’s Future: Insights on TSMC’s N2 Process

Jensen Huang’s perspective diverges from traditional beliefs surrounding Moore’s Law; he does not let it dictate performance improvements associated with new architectures. NVIDIA consistently aims to redefine performance expectations across their product line-ups, particularly in artificial intelligence (AI) and beyond. Over the years, they’ve established a crucial trajectory that challenges conventional scaling principles. When queried about GAA technology, Huang acknowledged its potential to enhance performance significantly but emphasized that it may not be a revolutionary shift.

Specifically, Huang estimates that GAA transistors might provide a performance increase of around 20%.However, he did not elaborate on the specific domains where these improvements would manifest, leading to some ambiguity. In the context of large-scale AI infrastructures, the synergy and interconnection of individual components will be vital. NVIDIA has achieved remarkable advancements by prioritizing these interconnections, often surpassing the benchmarks set by Moore’s Law.

TSMC 2nm yields are now well above 60 percent claims analyst

The journey towards GAA technology adoption in the semiconductor industry has been marked by challenges. Samsung, as the first company to offer a fully developed 3nm GAA process, has struggled with yield rates, having reportedly dropped to a concerning 20%.In contrast, TSMC appears to be on a more promising track with the introduction of the N2 (2nm) line of processes.

NVIDIA is anticipated to leverage GAA technology in its forthcoming Feynman AI architecture, slated for release in 2028, potentially utilizing TSMC’s N2 process. It’s worth noting that NVIDIA typically does not adopt cutting-edge processes immediately but rather allows them time to mature within the market. For Team Green, the “Huang Law”serves as the guiding principle—prioritizing performance enhancements through innovative architectural developments. Over the past decade, this approach has enabled NVIDIA to boost chip efficiency by more than 1, 000 times.

For more insights, visit Tom’s Hardware.

Source & Images: Wccftech