AI computing platform of NVIDIA has once again shattered its performance records on the latest MLPerf round. This year, by winning every single test out of all the six areas in the second version for data center and edge computing systems, NVIDIA GPU dominates for the second time in a row.
NVIDIA started off by redefining computer graphics, igniting the PC gaming market in the year 1999, and made revolutionary changes in parallel computing as well. Currently, NVIDIA is powering up Artificial Intelligence with its graphics processing unit (GPU) deep learning to make computers, self-driving cars, and robots with the ability to understand and interact with the world. The exceptional performance of NVIDIA A100 GPU will promote AI application into daily operations.
MLPerf inference benchmark is intended to measure the speed of ML inference performance using a trained model. The MLPerf benchmark test covered the ever-growing areas of AI such as speech recognition, understanding natural language, recommendation system, and medical imaging.
Earlier the usage of GPU for the interface was familiar to only a very few leading high-tech companies. However, now the situation has changed with NVIDIA’s AI platform that is available through different companies. To improvise business operations and additional services companies like those representing various industries, and major cloud and data center service providers use NVIDIA’s AI platform.
NVIDIA takes AI inference to new heights by offering more capacity to the public cloud which is seen for the first time. The total capacity on GPU has grown by 5 times every year. In the new recommender test, NVIDIA A100 GPU outdid the latest available CPUs by around 273x for data center inference. According to the MLPerf Interface, 0.7 benchmark means the performance of a single DGX A100 server is the same as the 950 dual-socket CPU servers. This can offer the customers’ cost efficiency when implementing AI recommender models from research to production.
NVIDIA in 1995 launched its first product and found its way to the top through significant contributions and the invention of GPU. An amazing fact is that 11 partners of NVIDIA other than NVIDIA’s own submission in the data center and edge category, submitted 1029 results in total using NVIDIA GPUs, representing over 85 percent of the total submissions.