Numascale

Numascale AS operates as a fabless semiconductor company, providing interconnect hardware solutions for scale-up and scale-out datacenter servers. Founded in 2008 as a spin-off from Dolphin Interconnect Solutions in Oslo, Norway, the company's technological lineage traces back to the Norwegian firm Norsk Data. The founding team included Kare Lochsen. The current leadership consists of Morten Toverud as CEO and Einar Rustad as CTO. Rustad's background includes roles as a CPU architect at Norsk Data and Director of Processor Architecture at ACRI in France, in addition to being a founder of Scali.

The core of Numascale's business is its proprietary NumaConnect™ technology, a cache-coherent interconnect that allows multiple standard servers to function as a single, large shared-memory system under one operating system instance. This approach addresses the underutilization of resources in traditional datacenters, where dual-socket servers often lead to over-provisioned memory. By creating scale-up systems, Numascale's technology enables better performance for in-memory applications and can reduce the total cost of ownership by minimizing expenses related to IT, power, and maintenance. The company's products are processor-architecture agnostic and have been implemented for both AMD and Intel platforms.

Numascale's primary offering is its line of Node Controllers, such as the UNC5, which connects to Intel's Ultra Path Interconnect™ (UPI) to enable servers to scale from 8 to 32 CPUs. This hardware allows any processor in the system to access any data location through direct load and store operations, which simplifies programming and allows compilers to automate parallelism more effectively. The company follows an OEM (Original Equipment Manufacturer) business model, collaborating with large system vendors like Atos, Eviden, Dell, and Hitachi, who then market and support the systems built with Numascale's technology. These systems are targeted at clients with demanding workloads, such as business-critical applications, transactional and analytical databases, graph databases, and Big Data applications that require large-scale shared memory.

Keywords: interconnect technology, cache coherent, shared memory systems, scale-up servers, high-performance computing, datacenter solutions, fabless semiconductor, node controller, NumaConnect, big data analytics, in-memory computing, OEM hardware, server architecture, UPI interconnect, processor agnostic, large-scale computing, memory virtualization, system interconnects, HPC hardware, data analytics infrastructure