AMD Launches Instinct MI300-Series Accelerators to Boost AI and HPC Performance

The new MI300X and MI300A chips reportedly outperform Nvidia's H100, with early adopters including Microsoft, Oracle, and the DoE's Lawrence Livermore National Labs.

AMD has launched its Instinct MI300-series accelerators, including the MI300X and MI300A, aimed at bolstering its standing in the AI and high performance computing (HPC) community.
The MI300X GPUs reportedly exceed the speed of Nvidia's H100, with FP8 performance of 2.6 petaFLOPS, making it 32 percent faster than Nvidia's H100.
The MI300A, a datacenter chip, combines Zen 4 CPU cores and CDNA 3 graphics cores, and is rated for 122.6 teraFLOPS of double precision performance, 1.8x that of the H100.
AMD's MI300-series parts are already being deployed by Microsoft, Oracle, and the DoE's Lawrence Livermore National Labs.
AMD also introduced its AMD Ryzen 8040 Series processors for AI-based laptops, with the performance of the 8040 expected to be 1.4 times better than current Ryzen chips.