AMD achieves a significant success by shipping thousands of Instinct MI300x to one of the world's leading private cloud computing platforms.
Vultr aims to establish a world-class artificial intelligence development environment.
AMD has introduced its powerful AI accelerator, the Instinct MI300X, which is based on the company's third-generation CDNA architecture and TSMC's advanced 5nm and 6nm manufacturing processes. This device features 19,456 stream processors, 192 GB of HBM3 memory, 304 compute units, and 1,216 matrix cores, along with a thermal design with a power consumption of 750 watts. In initial tests, the MI300X decisively outperformed Nvidia's RTX 4090.
Vultr, a privately-owned cloud computing platform, has announced its intention to integrate AMD's high-performance hardware into its scalable cloud infrastructure. This integration aims to enhance the management of GPU-accelerated workloads, ensuring optimal performance in both data centers and edge computing environments.
Vultr's CEO, J.J. Kardwell, commented that "innovation thrives in an open ecosystem," highlighting that the future of enterprise AI workloads lies in environments that allow for flexibility, scalability, and security. Kardwell added that AMD's accelerators provide customers with an unmatched cost-performance ratio, emphasizing the importance of combining high memory capacity with low power requirements, which also contributes to sustainability efforts.
The collaboration between Vultr and AMD aims to provide companies with a first-class AI development environment. AMD's architecture, combined with Vultr's platform, will enable businesses to work with pre-trained open-source models, thus facilitating the development of AI projects.
Negin Oliver, Corporate Vice President of Business Development of the Data Center GPU Business Unit at AMD, expressed satisfaction with this close collaboration, emphasizing that Vultr's platform is designed to handle high-performance AI training and inference tasks, improving overall efficiency.
The alliance also optimizes Vultr’s infrastructure for GPU-accelerated Kubernetes clusters, capable of managing complex workloads on a global scale. Although Vultr has not specified the exact number of Instinct MI300X units it has ordered, it has been mentioned that the initial cluster will consist of "thousands."