NVIDIA Tesla V100 Benchmark Results on Rescale

Rescale has incorporated NVIDIA's advanced Tesla V100 GPU, based on the Volta architecture, into its ScaleX platform, allowing users to utilize these GPUs starting November 13, 2017, as part of their standard batch workflow. The V100 systems, equipped with 1, 4, or 8 GPUs connected via a 300GB/s NVLink interconnect, are optimized for deep learning tasks with Tensor Cores that enhance model training speed, outperforming previous P100 systems by significant margins in benchmarks using frameworks like Caffe2 and TensorFlow. Rescale offers users the opportunity to start using these powerful GPUs immediately with sample jobs available for testing, provided they have a Rescale account, and is promoting a Deep Learning Kickstart Program in collaboration with Skyscale, offering hardware credits to selected applicants for GPU usage. This initiative aims to leverage the V100's capabilities in high-performance and cloud computing, as highlighted by Mark Whitney, a director of engineering at Rescale with a background in high-performance computing and cloud technologies.