ASRock Rack launched new servers supporting NVIDIA Blackwell architecture at COMPUTEX 2024.
ASRock Rack Inc. launched several scale-up servers leveraging the NVIDIA Blackwell architecture at COMPUTEX 2024, displayed at booth L0818 in Hall 1. The highlights include the ORV3 NVIDIA GB200 NVL72, a liquid-cooled rack, and the 6U8X-GNR2/DLC NVIDIA HGX B200, a direct-to-chip liquid-cooled GPU server. ASRock Rack is also showcasing GPU servers based on the NVIDIA MGX™ modular reference architecture for accelerated computing.
The ORV3-NVIDIA GB200 NVL72 is based on NVIDIA's GB200 NVL72 scale-up system, which connects 36 NVIDIA Grace™ CPUs and 72 NVIDIA Blackwell Tensor Core GPUs with NVIDIA NVLink™ technology, creating a single, massive GPU for faster real-time large language model (LLM) inference at lower total cost of ownership (TCO). Additionally, ASRock Rack launched the 6U8X-EGS2 NVIDIA HGX B100 and the 6U8X-GNR2/DLC NVIDIA HGX B200, equipped with the NVIDIA HGX™ B100 8-GPU system or the NVIDIA HGX B200 8-GPU system respectively. The NVIDIA HGX B100 serves as a drop-in replacement for existing NVIDIA HGX H100 infrastructures. The 6U8X-GNR2/DLC features direct-to-chip liquid cooling technology to push thermal design power (TDP) limits within a 6U rackmount design, while for customers without liquid-cooling infrastructure, an air-cooled 8U8X-GNR2 NVIDIA HGX B200 variant provides faster real-time inference on trillion-parameter models.
All ASRock Rack NVIDIA HGX servers support up to eight NVIDIA BlueField®-3 SuperNICs to leverage NVIDIA Spectrum™-X, an end-to-end AI-optimized Ethernet platform, and will be certified for the full-stack NVIDIA AI and accelerated computing platform, including NVIDIA NIM inference microservices as part of the NVIDIA AI Enterprise software platform for generative AI. ASRock Rack is developing a new product based on the NVIDIA GB200 NVL2 to provide scale-out configurations, facilitating seamless integration with existing data center infrastructure.
"We are showcasing data center solutions, powered by the NVIDIA Blackwell architecture, for the most demanding workloads in LLM training and generative AI inference," said Weishi Sa, President at ASRock Rack. "We will continue expanding our portfolio to further bring the advantages of the NVIDIA Blackwell architecture to mainstream LLM inference and data processing." Kaustubh Sanghani, Vice President of GPU Product Management at NVIDIA, stated, "Working with ASRock Rack, we are offering enterprises a powerful AI ecosystem that seamlessly integrates hardware and software to deliver groundbreaking performance and scalability, helping push the boundaries of AI and high-performance computing."
ASRock Rack also presents the 4UMGX-GNR2, a dual-socket GPU server that is NVIDIA MGX 4U-compliant and supports up to 8 FHFL dual-slot GPUs, such as the NVIDIA H200 NVL Tensor Core GPU for mainstream enterprise servers. It offers five FHHL PCIe 5.0 x16 slots and one HHHL PCIe 5.0 x16 slot supporting NVIDIA BlueField-3 DPUs and NVIDIA ConnectX®-7 NICs for multiple 200Gb/s or 400Gb/s network connections, and is equipped with 16 hot-swap drive bays for E1.S (PCIe 5.0 x4) SSDs. Visitors at the ASRock Rack booth L0818 can also view the MECAI-NVIDIA GH200 Grace Hopper Superchip server, launched earlier this year.