On a visit to Seoul, South Korea with the Blognone team for the Samsung AI Fam Tour, we had the opportunity to listen to the presentation by Justin Choi, Vice President and Head of the Security Team at Samsung Electronics’ Mobile eXperience Division, discussing the future of the AI market and autonomous driving from the perspective of memory chip manufacturers where Samsung dominates.
Choi emphasized the increasing need for processing power in the industry, especially from AI, not just on server-side but also on the edge devices. Whether it’s about privacy concerns, fast AI response, or even self-driving cars, he highlighted the development towards Level 4 and 5 autonomy in the future, transcending the current Level 3 where the driver must be ready to take control at all times.
Therefore, the chip industry’s focus should not only be on CPUs or GPUs but also on memory. Samsung already has High-Bandwidth Memory (HBM) catering to industry demands and continues to invest in chip development, including HBM manufacturing.
Another area Samsung is dedicated to advancing is Process-in-Memory (PIM), adding processing units to memory to reduce reliance on CPU and GPU while incorporating Programmable Computing Units (PCU) for AI processing tasks.
In terms of efficiency, HBM-PIM significantly boosts processing performance compared to current memory chips by up to 4 times and reduces energy consumption by around 70%. Samsung aims to integrate PIM into memory chips across all markets, including mobile devices, data sensors, and HPC, hinting towards the future release of HBM-PIM, GDDR-PIM, and LPDDR-PIM into the market soon.
TLDR: Samsung’s Justin Choi discussed the future of AI and autonomous driving, emphasizing the importance of memory development, including HBM and PIM technologies for enhanced processing power and energy efficiency in various industry sectors.
Leave a Comment