Scalable, bare-metal HPC Cloud for Al and machine learning DATA SHEET Benefits • TrueHPC solution - full HPC processing power and speed • Scalable service that can flex with your HPC and GPU workloads • Non virtualised - compute and storage together on campus • Flat line pricing and up to 50% lower cost than hyperscale cloud • Easy to migrate and set-up, with no lock-in • “HPC Blueprints” - deployments tailored to your applications • Supported by an expert HPC technical team • Powered by 100% sustainable, renewable Icelandic energy. Features • Bare-metal HPC Cloud (reserved or on-demand) • Latest CPU and GPU technologies • Fast InfiniBand interconnects - low latency • Local, dynamic, private and archive storage options • Built using open source technologies • Hosted in a secure, HPC optimised campus • Expert, full life-cycle DevOps support • Dedicated cloud portal for efficient workload management • Free Trial platform – try before you buy. The rapid evolution and deployment of AI, machine learning and deep neural network technologies are placing intensive demands on compute. To support this, hpcDIRECT is designed to specifically cater for the extreme algorithmic workloads characterised by AI. hpcDIRECT provides the perfect mix of latest specification compute on a reserved and flexible, on-demand basis. Bare metal servers (with a selection of GPUs) together with storage (shared or dedicated), fast interconnects and enhanced data transfer are offered via a secure and ultra low carbon cloud. • Reserved – HPC infrastructure designed and delivered to specification, within a minimum one-month SLA. • On-demand – Enabling an existing hpcDIRECT customer to rapidly access additional compute and storage at peak times. Deployment hpcDIRECT can be deployed through a reserved or on-demand model providing an agile, flexible and highly scalable HPC service: