The A100 Tensor Core GPU demonstrated the fastest performance per accelerator on all eight MLPerf benchmarks. The DGX SuperPOD system, a massive cluster of DGX A100 systems connected with HDR InfiniBand, also set eight new performance milestones.
Today’s announcement is the third consecutive showing for NVIDIA in MLPerf training tests, an industry benchmark formed in 2018.
The latest benchmarks featured two new tests and one substantially revised test: One ranked performance in recommendation systems, another tested conversational AI using BERT. The last, the reinforcement learning test involved diverse operations from gameplay to training.
Much of the same software used for the latest MLPerf benchmarks is available to developers today on NVIDIA’s software hub NGC.
To learn more about today’s announcements, we’ve published two in-depth technical blogs. Optimizing NVIDIA AI Performance for MLPerf v0.7 Training, and, Accelerate your AI Training with MLPerf Containers and Models from NVIDIA NGC.