Dell and AMD Join Forces to Power the Future of AI Servers
Dell has announced an expansion of its collaboration with AMD in the realm of AI servers, integrating Instinct MI300X compute cards into the PowerEdge series servers and new generative AI solutions. Dell emphasized that as artificial intelligence rapidly evolves, the fusion of open-source technology with cutting-edge hardware acceleration is driving industry innovation. Together, the two companies aim to offer cost-effective on-premises infrastructure solutions to support generative AI applications in enterprises.
Dell will install Instinct MI300X compute cards in the PowerEdge XE9680, providing high-performance capabilities tailored for enterprises leveraging generative AI. The server is equipped with eight Instinct MI300X compute cards, totaling 1.5TB of HBM3, and delivering 42 petaFLOPS of FP8 computational power. Users can streamline deployment through Dell OpenManage Enterprise, utilize APEX AIOps software for intelligent automation, and enhance security with integrated network recovery and zero-trust methodologies. This allows enterprises to swiftly implement and scale their GenAI solutions while maintaining robust security postures.
The Instinct MI300X employs a chiplet design, utilizing both 5nm and 6nm processes, and contains 153 billion transistors. It features the fourth-generation Infinity Fabric solution, comprising 28 module chips, including eight HBM and four compute chips. Each compute chip houses two CDNA 3 architecture-based GCDs, amounting to 80 compute units—320 compute units and 20,480 stream processors in total. For yield optimization, AMD has reduced the number of compute units, resulting in 304 compute units and 19,456 stream processors in practical use. Additionally, the HBM3 capacity reaches 192GB, offering 5.3TB/s memory bandwidth and 896GB/s Infinity Fabric bandwidth.
