Open Arsenal

Rethink the Path Between Data and Decisions.

Conduit Systems is an open-source system architecture platform designed to fundamentally rethink how data moves between memory, compute, and accelerators in modern computing environments. Traditional architectures have focused heavily on improving processor performance, but as AI and high-performance workloads have scaled, the real bottleneck has shifted to data movement. Conduit Systems addresses this by treating memory pathways and interconnects as first-class components, enabling deterministic, high-bandwidth communication across modular chiplet-based systems.

At its core, Conduit Systems introduces a unified framework that combines advanced packaging concepts, simulation-driven design, and runtime-aware execution. Through its integration with CircuitPath, it enables realistic modeling of chiplet connectivity using technologies like UCIe-style interconnects, 2.5D interposers, and 3D stacking constraints. The platform also includes a topology-aware runtime (ConduitOS) that intelligently schedules workloads based on bandwidth availability, thermal limits, and chiplet locality. This allows systems to optimize not just compute performance, but the efficiency of data delivery itself.

What makes Conduit Systems particularly powerful is its modular and extensible design. Developers and researchers can define system architectures using a unified descriptor format, simulate performance before hardware is built, and extend the platform with new chiplet models, packaging layouts, and optimization strategies. Built under the AGPL-3.0+ license, it fosters a collaborative ecosystem where improvements remain open and accessible, accelerating innovation across the industry.

This specification is released under the GNU Affero General Public License v3.0 or later (AGPL-3.0+) and may be used freely with required attribution under Section 7. A Specification Branding License is available for attribution-free deployments, with fees based on usage, scope, and deployment size.