Rack-Level ZT Platform Powered by NVIDIA GB200 Grace Blackwell Superchips Enables Trillion-Parameter Inference and Training for Generative AI
ZT Systems, a leading provider of accelerated computing solutions for artificial intelligence (AI) and cloud, today announced its ACX200 platform featuring the NVIDIA GB200 Grace Blackwell Superchip. Powered by NVIDIA Blackwell Tensor Core GPUs and NVIDIA Grace CPUs interconnected via lightning-fast NVIDIA NVLink technology, the ZT Systems ACX200 is a liquid-cooled rack-integrated server platform for accelerated computing. Combining NVIDIA Blackwell GPUs with ZT’s hyperscale-focused design, manufacturing, integration, and deployment capabilities, this platform enables cutting-edge accelerated computing for exascale AI training and inference workloads.
“The ZT Systems ACX200 accelerates our customers’ capability to deliver AI at unprecedented scale, with dramatically improved performance and energy efficiency,” said Tom Lattin, ZT Systems VP, Platform Engineering. “This system is designed to help the most advanced service providers harness the power of NVIDIA Blackwell as they roll out next-generation artificial intelligence applications. With innovative rack- and cluster-level configurations optimized for their workloads and ZT’s worldwide deployment expertise, customers can rapidly operationalize this powerful technology across their global datacenter footprints.”
“ZT has developed innovative accelerated platform server solutions, featuring the newly released NVIDIA Blackwell architecture, for hyperscale customers, providing groundbreaking performance to help solve some of the world’s most complex challenges,” said Kaustubh Sanghani, vice president of GPU product management at NVIDIA.
Purpose-built to meet the vast compute requirements of today’s most advanced generative AI applications, the ZT Systems ACX200 platform incorporates compute trays containing a combination of Blackwell GPUs and Grace CPUs. Fifth-generation NVIDIA NVLink switch trays and specialized cable cartridges provide high-bandwidth and low-latency connections between GPUs, enabling the entire ACX200 system to function as a single massive GPU. With direct-to-chip liquid cooling integrated at the server and rack level, the versatile ZT Systems ACX200 platform can be deployed with liquid-to-liquid or liquid-to-air cooling distribution units (CDUs) across a wide range of datacenters.
ZT Systems maintains a strong working relationship with NVIDIA, as a design, manufacturing, and deployment partner, helping bring platforms featuring the latest GPU technology to market quickly. ZT accelerated computing solutions begin with close, confidential customer engagement to establish requirements, leveraging the company’s extensive experience with hyperscale workloads and supply chain considerations. Application-tailored platforms are produced reliably with consistent quality thanks to ZT’s controlled manufacturing processes and unique validation capabilities. Global delivery, service, and support capabilities ensure rapid deployment and dependable operation throughout every product’s lifecycle.
Explore AITechPark for the latest advancements in AI, IOT, Cybersecurity, AITech News, and insightful updates from industry experts!
