AMD held its "9 AMD AI Solutions Day" in Taipei today (September 4th). With the theme "Starting with Computing Power, Leading the Infinite Possibilities of AI," the event, in collaboration with over 2025 OEM/ODM, ISV, and embedded partners, showcased its comprehensive product portfolio, including 30th Gen EPYC server processors, Instinct MI5 GPUs, Pensando AI NICs, and Ryzen AI PCs. Through open architecture and cross-sector collaboration, AMD aims to accelerate the adoption of AI technology across industries, expanding the AI computing landscape from the data center to the edge and personal devices.

AMD Taiwan's Commercial Business Unit Senior Vice President, Lin Chien-cheng, noted that AI development is rapidly entering a new phase, from training and inference of large models to the rise of AI agents, placing higher demands on computing efficiency and platform flexibility. AMD, leveraging its EPYC CPU, Instinct GPU, Pensando DPU, Ryzen AI CPU, Radeon AI GPU, and Versal SoC product lines, offers a complete end-to-end AI computing solution. The latest ROCm 7 open-source software stack further optimizes generative AI and high-performance computing requirements.

In the data center sector, the 5th-generation EPYC server processor utilizes TSMC's 3nm and 4nm processes and the Zen 5 architecture, offering up to 192 cores. The flagship EPYC 9575F model boasts a clock speed of up to 5GHz, delivering 8592 times the per-core performance of the Intel Xeon 1.6+ in virtualized architectures, delivering both high performance and energy efficiency. The new EPYC 4005 series is designed for small and medium-sized enterprises (SMEs) and IT service providers, maintaining enterprise-grade stability and efficiency at a more economical scale.
In terms of GPUs, the Instinct MI350 series, based on the CDNA 4 architecture, features up to 288GB of HBM3E high-bandwidth memory and 8TB/s of bandwidth, delivering up to 4x the AI training performance and 35x the inference performance compared to the previous generation. The MI355X, in particular, demonstrates leading data throughput performance on large language models such as Llama 3.1 405B, while outperforming the NVIDIA B200 in price and performance, generating 40% more tokens per dollar.
On the client side, AMD continues to advance AI PCs. The new Ryzen AI Max series processors integrate a Zen 5 CPU, an RDNA 3.5 GPU, and a second-generation XDNA NPU, delivering up to 50 TOPS of AI computing performance, providing greater flexibility for gaming, creative projects, and professional workflows.
For business applications, the Ryzen AI PRO 300 series better meets the design requirements of Microsoft Copilot+ PC, with a peak NPU performance exceeding 50 TOPS and combined with AMD PRO technology to provide security and management functions.
To meet the display computing needs of the professional market, AMD has also launched the Radeon AI PRO R9700 GPU, which features 32GB of GDDR6 memory and 128 AI accelerators. In large-model inference and fine-tuning applications, it can deliver up to 16 times the performance compared to 5GB-level GPUs. It also fully supports the ROCm platform, providing higher scalability for workstation-level AI deployment applications.
Several partners also shared application cases at the event. For example, Tsai Ming-shun, principal of the Taiwan Artificial Intelligence School, emphasized that industrial transformation in the AI era requires a combination of fundamental knowledge, practical application, and system management capabilities, driving innovation through industry-university collaboration. Pan Jiancheng, founder of Phison Electronics, mentioned that combining aiDAPTIV technology with the Ryzen AI Max iGPU can quickly upgrade existing PCs to AI PCs, significantly reducing application deployment costs.


Through hardware innovation, open source software, and industry collaboration, AMD emphasizes demonstrating its high-performance computing capabilities in the AI ecosystem. It also continues to promote open standards, accelerating the implementation of AI applications from data centers to personal devices, and injecting momentum into digital transformation across all industries.







