Based on its innovative interconnect Numascale offers scalable and expandable systems for high performance applications. Adding up standard servers will scale your GPU system within a single image operating system and environment with scalable and shared memory and open for better utilization of the GPUs’ computing power.
Numascale offers a price breaker for shared memory systems by offering integration of a simple add-on card to commodity servers. The hardware is now deployed in system with up to more than 1700 cores and the memory addressing capability is virtually unlimited. The technology has a set of interesting advantages that will catch the interest of innovative developers.
Numascale’s Numaconnect is now applied in servers form IBM and Supermicro to create versatile and scaleable systems with large headroom for cores and memory. The systems run standard Linux operating systems and do not rely upon any complicated virtualization layers. The ccNuma approach is based on well established experience in computer architecture.
The Pittsburgh Supercomputer Center recently announced a new step in its path to discover new routes to solving memory and I/O problems for data-intensive applications. While they already have a stable of systems that take new approaches to shared memory and have looked to software-based solutions as alternatives, they have now turned to Numascale to dip their toe in the hardware-based…