QuestionQ105Verified/Understand fundamental AI conceptsCopy linkReport a problemAsk AstroTutorSave questionWhat is one way NVIDIA Hopper GPUs speed up AI training compared with NVIDIA Ada Lovelace GPUs?AThey provide multiple 4th generation Tensor Cores per GPU.BThey support Multi-Instance GPU (MIG) to accelerate distributed training.CThey support NVLink to accelerate communications between GPUs.DThey support a RAS engine to dynamically correct issues during training.Show AnswerAnswerExplanationHopper H100 GPUs implement fourth-generation NVLink, a high-speed GPU interconnect that increases multi-GPU scalability by accelerating communication among GPUs during distributed AI training. Ada Lovelace GPUs also use fourth-generation Tensor Cores, so Tensor Core generation is not the distinguishing capability. NVIDIA documents Hopper NVLink in the Hopper Tuning Guide, while its Ada Lovelace architecture reference identifies fourth-generation Tensor Cores for Ada.Learn moreNVIDIA Hopper Tuning GuideNVIDIA Ada Lovelace Architecture vGPU TypesShow comments0Show commentsCommunity DiscussionNo comments yet. Be the first to start the discussion!
Community Discussion