LumaCore™ 192-Core ARM
High-core-count ARM compute for dense, parallel workloads in an efficient node architecture.
THE NEXT GENERATION OF COMPUTE
Next-generation data center infrastructure built on LumaCore™ 192-core ARM compute nodes with terabytes of locally shared memory, interconnected by LumaPlane™, our high-bandwidth optical backplane.
High-core-count ARM compute for dense, parallel workloads in an efficient node architecture.
Large DDR5 ECC memory capacity, shared coherently among all 192 cores within each node.
High-bandwidth optical interconnects linking independent compute nodes.
Designed to improve compute per watt and reduce cooling and operating costs.
OPTICAL BACKPLANE
LumaPlane is our purpose-built optical interconnect architecture, designed to link independent compute nodes with scalable bandwidth and low-latency communication. Each node retains its own locally shared memory.
HARDWARE PLATFORM
Concept architecture. Memory capacity, bandwidth, mechanical configuration and thermal performance remain subject to hardware validation.
DESIGNED AS A SYSTEM
192 ARM cores per node for workloads that benefit from many general-purpose processors.
Terabyte-scale local memory accessible to the cores of each independent node.
A purpose-built optical backplane for connecting nodes without cross-node memory coherence.
An integrated approach to power, density, airflow and total cost of ownership.
ABOUT LUMALINX
LumaLinx is developing an integrated hardware platform combining LumaCore™ high-density ARM compute, large local memory, and LumaPlane™ optical connectivity. Our goal is to simplify scale-out infrastructure and improve useful compute per watt.
Get in touch ↗