NVIDIA DGX-2 AI Supercomputer Sets Records for Training and Inference
by Rich Brueckner from High-Performance Computing News Analysis | insideHPC on (#3ZZPY)
NVIDIA has one-upped performance breakthroughs announced earlier this year, thanks to two new technologies. Related to data augmentation and tensor parameter interleaving, these technologies enabled a single Tesla V100 GPU to do a ResNet-50 training run in just under 24 hours (1,350 images/second). "We also discussed how these same technologies enabled a single DGX-1 server node to train ResNet in just over four hours (7,850 images/second). Since then, we put our 16-GPU DGX-2 server to the test, and using these same enabling technologies, DGX-2 finished the training run in 108 minutes (just under two hours)!"
The post NVIDIA DGX-2 AI Supercomputer Sets Records for Training and Inference appeared first on insideHPC.