NVIDIA HGX B300
Search documents
Compal Showcases Comprehensive Data Center Solutions at 2025 OCP Global Summit; Unveils SGX30-2 High-Performance AI Server Built with NVIDIA HGX™ B300
Prnewswire· 2025-10-16 16:00
Core Insights - Global data centers are facing challenges due to increased AI compute demand, larger working sets, and stricter energy efficiency requirements [1] - Compal Electronics presented a vision for future data centers, offering end-to-end solutions encompassing compute, memory, and cooling at the OCP Global Summit [1][2] Compute Solutions - Compal showcased its latest AI server, SGX30-2 / 10U, built on the NVIDIA HGX B300 platform, featuring eight NVIDIA Blackwell Ultra GPUs and dual Intel® Xeon® 6 processors [2] - The NVIDIA Blackwell Ultra GPUs provide up to 2.1 TB of HBM3e memory and 1.8 TB/s GPU-to-GPU NVLink bandwidth, achieving a total interconnect bandwidth of 14.4 TB/s [3] - The system delivers 144 PFLOPS of FP4 inference performance and approximately 72 PFLOPS of FP8 performance, marking a 7x increase in compute performance compared to the previous generation [3] Memory Expansion Technologies - Compal introduced CXL and RDMA technologies for AI Memory Expansion, addressing memory bottlenecks in large language model training and HPC [4] - The CXL.mem protocol allows servers to access pooled SCM-based Memory Expanders, enabling efficient processing of larger datasets beyond HBM limits [4] - RDMA facilitates direct data movement across servers and data centers, supporting large-scale memory disaggregation and resource pooling [6] Infrastructure Evolution - The integration of intelligent DMA and low-latency data-path design transforms NVMe storage into a memory-like extension, enhancing data access speed without increasing CPU overhead [5] - Compal's technologies aim to create a unified, energy-efficient infrastructure that evolves from current PCIe-based systems to next-generation GPU direct storage and CXL-enabled architectures [6] - Compal positions itself as a system integrator and strategic partner for enterprises in their digital transformation journey [7] Company Background - Founded in 1984, Compal is a leading manufacturer in the notebook and smart device industry, recognized for its innovative product designs [8] - In 2025, Compal was listed among Taiwan's top 7 manufacturers and consistently ranks among the Forbes Global 2000 companies [8] - The company is actively developing emerging businesses, including cloud servers and smart medical solutions, leveraging its integrated R&D and manufacturing capabilities [8]
Is Supermicro Stock The AI Winner Everyone's Overlooking?
Benzinga· 2025-09-12 16:44
Core Insights - Super Micro Computer Inc (SMCI) is experiencing significant growth due to its AI hardware offerings, with a stock increase of 5.3% following the announcement of global shipments of its Nvidia Blackwell Ultra-powered systems [1][2]. Group 1: Product and Performance - The Blackwell Ultra systems, including NVIDIA HGX B300 and GB300 NVL72 racks, offer up to 7.5 times the performance of Nvidia's Hopper platform, 50% more HBM3e memory, and up to 1.1 exaFLOPS of FP4 compute [3]. - Liquid cooling technology in these systems can reduce power consumption by up to 40%, addressing critical challenges in AI data center growth [3]. Group 2: Market Positioning - Supermicro is establishing itself as a comprehensive solution provider for AI infrastructure, with over 10 new SKUs featuring Blackwell and Blackwell Ultra, simplifying deployment for enterprise and hyperscale customers [4]. - The company is transitioning from a niche hardware supplier to a key player in the global AI factory ecosystem, benefiting from the convergence of Nvidia's chip dominance and OpenAI's data center ambitions [5]. Group 3: Financial Performance - As of the latest publication, SMCI shares were trading at $45.48, reflecting a 3.49% increase, and the stock is within a 52-week range of $17.25 to $66.43 [6].