The MI300X and MI300A are built using advanced production technologies, including a '3.5D' packaging technique. The MI300X features 304 compute units, 192GB of HBM3 capacity, and 5.3 TB/s of bandwidth, while the MI300A combines 24 threaded CPU cores and 228 CDNA 3 compute units. AMD claims that the MI300X platform has 2.4X more memory capacity and a claimed 1.3X advantage in computational horsepower over Nvidia’s H100 HGX platform. The MI300A, on the other hand, is said to have a 1.8X advantage over the H100 in FP64 Matrix/DGEMM and FP64/FP32 Vector TFLOPS.
Key takeaways:
- AMD has launched its Instinct MI300X AI accelerator and the Instinct MI300A, the world’s first data center APU, using advanced production technologies and '3.5D' packaging.
- The Instinct MI300X GPU reportedly delivers up to 1.6X more performance than the Nvidia H100 in AI inference workloads and offers similar performance in training work, with more than twice the HBM3 memory capacity than Nvidia’s GPUs.
- The AMD Instinct MI300A combines both a CPU and GPU in the same package, and is claimed to deliver up to 4X more performance than Nvidia’s H100 GPUs in some workloads.
- The MI300X and MI300A are designed to work in groups of eight in AMD’s generative AI platform, facilitated by 896 GB/s of throughput between the GPUs across an Infinity Fabric interconnect.