AWS said Monday its Amazon Elastic Compute Cloud or EC2 P4d instances bring additional graphics processing unit performance suitable for complex ML models and accelerate ML training by three times.
OmniSci used the P4d technology to deploy a data analytics platform with reduced costs.
The P4d instances are equipped with eight Nvidia-made GPUs with the capacity to produce a maximum performance of 2.5 petaflops. Each instance also features a high-bandwidth GPU memory of 320 gigabytes.
P4d technology has the potential to support large workloads for multiple applications such as materials science, financial risk modeling and drug discovery.
"Through the use of P4d instances, we were able to reduce the cost to deploy our platform significantly compared to previous generation GPU instances thus enabling us to cost-effectively scale massive data sets," said Ray Falcione, vice president for U.S. public sector at OmniSci.