NVIDIA posted machine learning performance figures of its Ampere A100 GPU on the NVIDIA blog. The GPU giant says its A100 is up to 4.2x faster than what its predecessor, the NVIDIA V100, delivered at launch time. Compared with current performance of the V100, which has been boosted significantly thanks to software optimizations, the A100 is up to about 2x faster in some tests.
For example, the tests show at equivalent throughput rates today’s DGX A100 system delivers up to 4x the performance of the system that used V100 GPUs in the first round of MLPerf training tests. Meanwhile, the original DGX-1 system based on NVIDIA V100 can now deliver up to 2x higher performance thanks to the latest software optimizations.
These gains came in less than two years from innovations across the AI platform. Today’s NVIDIA A100 GPUs — coupled with software updates for CUDA-X libraries — power expanding clusters built with Mellanox HDR 200Gb/s InfiniBand networking.
EE Times analyzes the results over here and compares them with Google's TPU.