Park Yong-in, who oversees Samsung Electronics’ system semiconductor design operations as President of the System LSI Division, expressed confidence in the success of the company’s latest mobile application processor, the Exynos 2400.
On Oct. 26, following his keynote address at Semiconductor Expo 2023 held at COEX in Seoul, Park met with reporters and stated, “The Exynos 2400 boasts superior graphics processing capabilities compared to North American competitors, and I believe it will perform well.” He further added, “I hope the upcoming Samsung Galaxy S24 series, which will feature the Exynos processor, will also be successful.”
Earlier this month, Samsung Electronics unveiled the Exynos 2400, a next-generation mobile processor suited for the age of generative AI. This processor is slated to be incorporated into Samsung’s flagship Galaxy S24 series. The Exynos AP faced challenges earlier this year when not a single chip was included in the Galaxy S23 due to performance issues, denting the brand’s pride. However, the company is aiming for a resurgence by significantly improving various performances, leveraging a 4-nanometer (nm) foundry, neural processing unit (NPU), central processing unit (CPU), and more. Compared to its predecessor, the Exynos 2200, the CPU performance has improved by 1.7 times and AI capabilities by 14.7 times, marking an ambitious aim to close the gap with global leaders like Qualcomm and Apple.
Park also hinted at the potential use of the gate-all-around (GAA) 3-nm process, which Samsung Electronics’ foundry began mass-producing as a world first. Regarding this, he noted, “In the realm beyond 100 million pixels, we (Samsung Electronics) dominate.” When pressed further, he responded succinctly, “We are preparing, and it will go well.”
In his keynote, Park emphasized the significance of growth in AI semiconductors. “Prices of GPUs used as AI accelerators are on the rise, and they have significant energy consumption issues. Consequently, there’s a growing shift towards alternative processes optimized for AI workloads, such as NPUs and RISC architectures,” he explained.
Highlighting the environmental impact of technological advancements, Park stated, “Due to generative AI, the total annual power consumption of data centers now exceeds that of countries like Spain, the U.K., or Italy. By 2030, the power consumption will likely surpass this year’s 470 terawatts, reaching over 1,000 terawatts.”
Describing the features of the GPU embedded in the Exynos 2400, Park particularly mentioned its ray tracing technology, which has been highly praised by GPU experts. Ray tracing is a graphics technology that realistically represents objects by tracing light as it refracts, reflects, and penetrates them. This GPU was co-designed with global semiconductor company AMD and was also implemented in Samsung’s flagship smartphone, the S23, released earlier this year. “This technology will also be incorporated into phones released next year. Smartphones are now performing functions that supercomputers used to do,” he noted.
Park predicts a shift towards “on-device AI” in the future. On-device AI refers to the majority of AI operations being executed on edge devices, connecting to the cloud selectively. “As AI applications increase, cloud operational costs remain an issue, and challenges like data latency and security persist. The term ‘on-device AI’ will become prominent, and Samsung’s goal is to place a powerful device, akin to a supercomputer, in everyone’s hands, enabling them to use generative AI anytime,” he emphasized.
Regarding Samsung Electronics’ upcoming challenges, Park stated, “In the visual domain, we aim to develop cameras that can capture what’s unseen by 2027. In the olfactory and auditory domains, many sensors and algorithms will be necessary. The main challenge for the semiconductor division will be designing mobile devices while minimizing power consumption and heat generation.”