AMD on the "2025 AMD AI SOLUTIONS DAY" held in Taipei, AMD shares the breakthrough processes, packaging technology, performance and innovative applications of the latest AMD EPYC processors, Instinct series GPUs, AMD embedded products and Ryzen AI PCs with more than 30 OEM/ODM, ISV and embedded partners.
Lin Jian-chan, deputy general manager of the commercial business department of AMD Taiwan, said in a speech that AMD is committed to promoting AI processes in all aspects by EPYC CPU, Instinct GPU, Pensando DPU, Ryzen AI CPU, Radeon AI GPU, and Versal self-adjusting system single chip (SoC). At the same time, in order to continuously promote the open AI industry system, AMD has launched the latest version of the open AI software stacking ROCm 7, aiming to meet the continuous growth of generative AI and high-performance computing, and comprehensively optimize the experience of developers.
With AMD EPYC processor and AMD Instinct GPU, AMD provides support for the AI computing power of data centers on the cloud. AMD emphasizes that EPYC processors continue to show strong electric power, bringing top-notch performance to the wide range of data center work such as enterprises, AI and cloud. Its 5th generation AMD EPYC processor uses telco 3-nanometer and 4-nanometer process technology, and is equipped with the Zen 5 core architecture, compatible with the widely deployed SP5 platform, and provides a wide range of cores from 8-core to 192-core, providing broken performance and energy efficiency.
The newly launched AMD EPYC 4005 series processors are designed for specific purposes and provide suitable size and enterprise-level capabilities and leading performance solutions for small and medium-sized enterprises and IT custodian service providers. With up to 16 Zen 5 cores, the AMD EPYC 4005 series processors provide a perfect balance of performance, reliability and efficiency on an economically affordable and easy-to-deploy platform.
In addition, AMD continues to deliver on its Instinct GPU's annual blueprint promise, the latest Instinct MI350 series GPUs are a brand new rod for generative AI and high-performance computing stacking, efficiency and scalability. Based on the AMD CDNA 4 architecture, the Instinct MI350X and MI355X GPUs provide the leading industry's 288GB HBM3E memory capacity and up to 8TB/s bandwidth, bringing up to 4 times AI computing and up to 35 times recommended performance than the previous generation of products, bringing excellent throughput to AI training and recommendations.
As for the fact that MI355X provides the highest volume of recommendation performance for large models in DeepSeek R1, Llama 3.1 405B, etc., helping various industries realize revolutionary AI solutions. At the same time, the MI355X has also seen a significant improvement in the performance-price ratio, which can generate up to 40% of the tokens per dollar compared to the NVIDIA B200 solution.
At the same time, AMD continues to promote the open AI industry system, meeting the growing demand for generative AI and high-performance computing through the latest version of AMD's open AI software stacking ROCm 7, while comprehensively improving developer experience. ROCm 7 has more complete industry standard framework support, expanded hardware compatibility, and new development tools, drivers, APIs and libraries to accelerate AI development and deployment.
In terms of client AI solutions, the new Ryzen AI Max series processors provide workstation-level performance for gamers, creators and daily users, with up to 16 Zen 5 CPU cores, up to 40 AMD RDNA 3.5 picture computing units, and AMD XDNA 2 neural processing unit (NPU) with up to 50 TOPS AI processing capabilities, achieving the best power in a super-convenient size. In addition, the Ryzen AI Max PRO series processors equipped with AMD PRO technology allow users to handle large-scale engineering and construction models and handle complex AI-accelerated workloads.
Also, the Ryzen AI 300 series processors are powered by up to 8 Zen 5 CPU cores and RDNA 3.5 display architecture. With AMD XDNA 2 technology, this series of processors provides up to 5 times higher performance than the first-generation NPU, bringing greater AI computing power. In addition, to meet daily business production needs, the Ryzen AI PRO 300 series processors support a new generation of Microsoft Copilot+ experience. Its leading peak 50+ NPU TOPS AI efficiency exceeds the requirements of Microsoft's Copilot+ PC, creates a new generation of AI experience for Windows 11, and provides enterprises with computing capabilities that support AI transformation..
Finally, the Radeon AI PRO R9700 is AMD's most advanced professional drawing solution designed for local AI recommendations, model microscopy and complex creative work, and supports multi-GPU system deployment for scalability. With 64 computing units, 128 2nd Generation AI accelerators, combined with 32GB high-speed GDDR6 memory, advanced RDNA 4 architecture, and support for FP8, FP16 and INT8 accuracy, the Radeon AI PRO R9700 not only provides sufficient TOPS performance, but also provides the flexibility required to handle today's most stringent AI workloads.