Nvidia, known for its high-quality AI chips, has announced the release of its latest innovation: the H200 Tensor Core GPU. This new chip, boasting an impressive 141 gigabytes of memory, delivers a significant performance boost of 60% to 90% compared to its predecessor, the H100 model, when used for inference tasks in AI models.
The H200-powered systems will be made available in the second quarter of 2024 through Nvidia’s hardware partners and major cloud service providers, including Amazon Web Services (AWS), Google Cloud, Microsoft Azure, and Super Micro.
Ian Buck, Nvidia’s vice president of hyperscale and high-performance computing, expressed his excitement about the new product, stating, “With Nvidia H200, the industry’s leading end-to-end AI supercomputing platform just got faster to solve some of the world’s most important challenges.”
Nvidia’s GPUs have gained popularity due to their ability to handle parallel computations necessary for training AI models and serving customers. The current high-end model, H100, was released earlier this year and comes with a price tag of around $25,000 per GPU. These GPUs quickly became a hot commodity in the technology industry as the demand for generative artificial intelligence surged.
While AMD’s upcoming MI300 AI product has also garnered attention in the inference applications space, it remains to be seen how it will withstand the competition from Nvidia’s upgraded H200.
Nvidia’s recent investor presentation indicates a shift in its product cycle strategy, moving from a two-year cycle to a one-year cadence specifically for AI chips. The company plans to launch additional high-end AI products in 2024 and 2025, as revealed in the presentation.
This announcement solidifies Nvidia’s commitment to more frequent and higher performance product releases, putting its rivals at a disadvantage.