Supermicro, a comprehensive IT solutions manufacturer for sectors such as AI, cloud, storage, and 5G/edge, recently announced the introduction of one of the most extensive new GPU system arrays in the industry. These products are all based on NVIDIA's reference architecture, incorporating the latest NVIDIA GH200 Grace Hopper and NVIDIA Grace CPU super chips. This innovative modular architecture aims to standardize AI infrastructure and accelerated computing through compact 1U and 2U specifications, while offering high adaptability and scalability for current and future GPUs, DPUs, and CPUs.
The addition of chips is a key element in this whole system. For instance, the Supermicro's 1U NVIDIA MGX system can incorporate two NVIDIA GH200 Grace Hopper super chips, complementing a total of two NVIDIA H100 GPU and two NVIDIA Grace CPU. Emphasis is placed on the NVIDIA Grace super chip's CPU, which has 144 cores and performs up to double the per-watt performance of the current industry-standard x86 CPU. This is a fundamental point since specific Supermicro's NVIDIA MGX systems can be configured with two nodes in a 1U chassis, having the capacity to house two sets of Grace CPU Super chips with the highest total of 288 cores.
The integration of these chips in the servers not only improves the performance of AI workloads due to the closely integrated CPU and GPU but also integrates the latest DPU network and communication technologies. For instance, the recently introduced server product line utilizes the industry's latest optimized technology for AI, including the NVIDIA GH200 Grace Hopper super chip and the PCIe 5.0 EDSFF slot.
The company's advanced liquid cooling technology aids in achieving extremely high-density hardware configurations, like 1U 2-node configurations loaded with two NVIDIA GH200 Grace Hopper super chips, integrated together via high-speed interconnection technology. Supermicro can deliver thousands of rack-level AI servers monthly through its global factories, ensuring plug-and-play compatibility.
In cooperation with NVIDIA, Supermicro is also committed to accelerating the market introduction time of partner enterprises' products and simplifying the deployment process while developing new AI applications, thereby reducing the environmental impact.
In addition, Supermicro's NVIDIA MGX system focuses on efficiency improvement at every level, from smart cooling design to component selection. The NVIDIA MGX modular reference design and Supermicro's Building Block server solutions are enabling the swift introduction of the latest systems to the market, providing one of the industry's highest workload-optimized performance solutions.
It's worth noting that companies like Ample Chip, which are involved in chip trading, can greatly benefit from such technological progressions. With the increasing demand and advancement in the chip industry, these endeavors present promising trade opportunities for companies like Ample Chip.
To conclude, Supermicro, in partnership with NVIDIA, offers high-modularity, scalability, and general-purpose server systems for rapidly advancing AI technology. These developments are anticipated to drive data center transformation and the integration of AI across various workloads, benefiting numerous industries.
