SGI NUMAflex™ Architecture

Global Shared Memory for HPC

SGI’ s NUMAflex architecture creates a system that can share up to 24TB of globally addressable memory, versus the 32GB most clustered architectures are limited to.

By making memory accessible globally, CPUs can access all data in the system’s memory without going through I/O and other bottlenecks. Perfect for HPC applications where latency must be minimized, and reduces the need for the system to pass along “copies” of data, slowing performance further and complicating application design.

With SGI’s NUMAflex architecture, systems administrators need to spend less time tuning high performance computing clusters.

North American Systems serves as SGI’s primary geographic reseller in the Midwest area. We are the expert local resource for SGI’s high-performance computing & data storage needs.

---

•  Print This Page