Microsoft has announced the general availability of a new virtual machine series, ND MI300X v5, designed specifically for demanding AI tasks. These VMs are powered by AMD's latest MI300X GPUs and boast industry-leading high-bandwidth memory (HBM) capacity and bandwidth.
Key features of ND MI300X v5 VMs
- Unmatched HBM capacity and performance: Each VM packs 8x AMD MI300X GPUs, each with a staggering 1.5 TB of HBM, allowing you to train and run larger AI models faster and with fewer GPUs.
- Scalability for massive workloads: ND MI300X v5 VMs integrate seamlessly with Azure Virtual Machine Scale Sets, enabling you to efficiently scale your AI workloads to thousands of GPUs. High-throughput, low-latency InfiniBand connectivity ensures smooth communication between VMs.
- Optimized for OpenAI models: These VMs were specifically designed to run Microsoft's own Azure AI services, including the powerful GPT-4 Turbo natural language model. They offer leading cost-performance for popular OpenAI and open-source models.
- Flexibility and ease of use: Whether you want to use your own models, leverage Azure's pre-built models, or access open-source models from the Azure AI catalog or Hugging Face, ND MI300X v5 VMs provide the best performance at the most competitive price. Plus, the scalability of the Azure cloud allows you to easily adjust resources as needed.
Benefits for developers and businesses
- Faster development and deployment: The VMs' open-source ROCm software platform simplifies porting models and solutions, reducing development time and costs.
- Access to cutting-edge technology: Be among the first to leverage the power of AMD's latest MI300X GPUs for your AI projects.
- Scalability for large-scale AI applications: Efficiently train and run massive AI models on a highly scalable infrastructure.
Availability
ND MI300X v5 VMs are now available in Canada Central. For more information and to get started, visit the Microsoft documentation page.
Read more here>>
Join the conversation! Your thoughts help the community grow.