At GTC 2022, NVIDIA released a new generation of H100
based on the Hopper architecture for the next-generation accelerated computing platform, but it will not be available until the third quarter of this year, which is still a few months away. This means that competitors can adjust the release time of their own products and make targeted deployments.
Moore’s Law is Dead recently said that although Intel has not officially launched Ponte Vecchio, the code name of the follow-up product has been determined, called Rialto Bridge, and it is planned to be launched next year. Compared to AMD’s Instinct MI250X, and Nvidia’s GH100, Intel’s first HPC GPU is already behind in time. Until now, Intel has not set a specific release date for the Ponte Vecchio.
Ponte Vecchio is Intel’s first exascale computing GPU. It uses the most advanced packaging technology in Intel’s history and has more than 100 billion transistors. It is the integrator of Intel’s advanced technology at this stage. According to Intel’s presentation at ISSCC 2022, Ponte Vecchio has 63 modules, including 16 Xe-HPG architecture computing chips, 8 Rambo cache chips, 2 Xe base chips, 11 EMIB connection chips, 2 Xe Link I/O chips, 8 HBM chips, and 16 modules responsible for TDP output, integrated with the Foveros 3D package through EMIB.
AMD and Nvidia will not slow down the development of a new generation of computing GPUs. AMD will have the Instinct MI300 series based on the CDNA 3 architecture and MCM multi-chip packaging, and Nvidia should be a GPU based on the Blackwell architecture. Together with Intel’s yet-to-be-confirmed Rialto Bridge, these compute GPUs are expected to launch within the next two years.