Microsoft cloud computing service will be the first to use Instinct MI300X

Microsoft and AMD have announced a pioneering collaboration: Microsoft’s cloud computing services will be the first to implement the Instinct MI300X compute card, synergizing with the fourth-generation EPYC server processors to forge novel Azure cloud virtual machines. These innovative instances, optimized for cloud computing and generative artificial intelligence, promise enhanced performance, revitalized services, and additional computational prowess.

The Azure ND MI300x v5 virtual machines will be the inaugural cloud platform to utilize the Instinct MI300X compute card. This integration heralds a monumental leap in performance, courtesy of the CDNA 3 architecture GPU. Its staggering 192GB of high-capacity memory is poised to accelerate high-performance computing (HPC) and artificial intelligence (AI) workloads. Furthermore, Microsoft and AMD have showcased how AMD’s product suite enables groundbreaking services and computational capabilities in cloud and generative AI, cryptographic computations, cloud computing, and the realm of more intelligent, sophisticated personal computers.

Per AMD’s official presentation, the Instinct MI300X is a pure GPU design, manufactured using a 5nm process. It boasts 8 XCDs (up to 304 CU/CDNA 3 architecture) and 8 HBM3 stacks (totaling 192GB), distinguishing it as a traditional accelerator for data centers.

Microsoft asserts that alongside the Instinct MI300X compute card, the fourth-generation EPYC server processors will also empower a range of virtual machines, from general-purpose to memory-intensive and compute-optimized. These new virtual machines not only offer competitive pricing but also deliver a 20% standard performance boost and double the performance enhancement for compute-oriented virtual machines.