Sure! Here’s the translation of your text into American English:
—
Amazon has taken a significant step in the field of artificial intelligence with the launch of its SageMaker HyperPod system, designed to enable organizations to maximize the performance of NVIDIA Blackwell GPUs within a single system. This advancement promises to be a milestone in AI innovation, offering impressive processing power that reaches 360 petaflops in dense floating-point (FP8) calculations and 1.4 exaflops in sparse floating-point (FP4) calculations.
The new P6e-GB200 UltraServers, based on the NVIDIA GB200 NVL72 architecture, provide exceptional performance for developing and deploying AI models that reach trillions of parameters. This integration of UltraServers with the distributed training environment of SageMaker HyperPod will allow companies to rapidly scale model development, minimize downtime, and simplify the process from training to large-scale deployment.
The architecture of the P6e-GB200 UltraServers connects 36 NVIDIA Grace CPUs with 72 Blackwell GPUs within a single NVLink domain, providing unparalleled access to computing resources. The architecture allows access to up to 72 Blackwell GPUs in a single NVLink domain, significantly enhancing performance and efficiency in AI model training.
In addition, this solution stands out for its high-performance network, which offers up to 130 TBps of low-latency NVLink bandwidth between the GPUs, improving communication in large-scale AI tasks. The UltraServers can also support up to 405 TB of local NVMe SSD storage, optimizing model training through fast data access.
The combination of SageMaker HyperPod with the P6e-GB200 UltraServers represents a critical advancement toward more agile and efficient training and deployment, especially for the most advanced generative AI models. This is particularly beneficial for organizations managing multiple teams and projects, ensuring maximum utilization of infrastructure and reducing operational costs.
Interested parties can access the capabilities of UltraServer through flexible training plans available in the AWS local zone in Dallas. Organizations looking to maximize their artificial intelligence efforts can register in the SageMaker console to start exploring this new technology.
—
Let me know if you need any further assistance!
Source: MiMub in Spanish
