Major Keynotes to Highlight Computex 2024 in Taipei

At Computex 2024, Nvidia announced new GeForce RTX AI laptops from ASUS and MSI, equipped with advanced AI hardware and Copilot features. These laptops are optimized for AI inferencing, training, gaming, 3D rendering, and video processing, offering significant performance boosts over Macs.

In gaming, AI enhances environments with DLSS and improves non-player characters (NPCs). Project G-Assist, another announcement at Computex, aims to help developers create AI assistants for games, offering context-aware responses, performance tracking, and system optimization during gameplay. Jesse Clayton, director of product for AI PC at Nvidia, emphasized the transformative impact of AI on PCs and its integration into major applications. Originally an April Fools' joke in 2017, Nvidia's G-Assist has returned as a real tech demo. The AI chatbot is designed to guide players through games and optimize PC settings. While currently a demonstration, G-Assist hints at future possibilities for game developers and RTX GPU owners, offering in-game assistance and configuration based on chat inputs.

At Computex 2024, Nvidia introduced its latest GPU lineup, featuring the Rubin, Rubin Ultra, Blackwell Ultra, and Supercharged Vera. These GPUs are part of the new Rubin architecture, named after pioneering astronomer Vera Florence Cooper Rubin. Set to release in 2026, the Rubin series will succeed the Blackwell series and aims to enhance data processing and computing power.

Nvidia is preparing for the 400GbE era with its Spectrum-X initiative, showcasing the 51.2T Spectrum-X800 Ultra for the ConnectX-8 era. The company has a roadmap extending to the ConnectX-9 NIC and a 102.4T switch era for 1.6Tbps networking. Nvidia plans to achieve 800Gbps networking by 2025, requiring PCIe Gen6, and aims for 1.6Tbps networking by 2026, indicating plans for a PCIe Gen7 NIC. This development highlights the accelerating pace of PCIe speed advancements.

AMD is aggressively updating its Instinct AI Accelerator lineup to rival NVIDIA in the AI segment. They plan to release a new or refreshed AI accelerator annually:

- In Q4 2024, the MI325X "CDNA 3" is set to launch with 288 GB HBM3E memory and impressive compute performance, doubling down on certain specifications compared to NVIDIA's H200.

- In 2025, the MI350 "CDNA 4" will be introduced, featuring a 3nm process and up to 288 GB HBM3E memory, with new data type support and OAM compatibility.

- For 2026, AMD has announced the MI400 "CDNA Next," continuing its yearly product refresh strategy.

AMD's roadmap positions them as a strong competitor in the AI accelerator market with a swift product development cycle, beginning with the MI300 series, which has seen rapid adoption.

At Computex 2024, AMD unveiled its new Ryzen AI 300 Series, the next generation of Ryzen laptop processors designed for generative AI workloads. This rebranding of the top-tier Ryzen 9 chips, now labeled with the HX suffix, denotes the best and fastest in the series without indicating power consumption. The new chips leverage AMD's latest architectures: XDNA2 for neural processing, RDNA 3.5 for integrated graphics, and Zen 5 for general processing. The series kicks off with the Ryzen AI 9 HX 370 and Ryzen AI 9 365, both featuring a 50 TOPS NPU, with the HX 370 positioned as the higher-end model.

Additionally, AMD announced its first Zen 5 desktop processors, set to launch in July. Leading the new Ryzen 9000 series is the flagship Ryzen 9 9950X, touted as "the world’s most powerful desktop consumer processor." This 16-core, 32-thread CPU boasts 80MB of L2+L3 cache and a 5.7GHz boost clock, promising a 16% uplift in instructions per cycle (IPC) performance over previous generations, with significant gains in both productivity and gaming.