BRING YOUR
OWN CODE
Our technology is designed for seamless integration into HPC systems, with minimal code modification and support for many application languages and frameworks, without vendor lock-in.
INTELLIGENT
ARCHITECTURE.
UNPRECEDENTED
PERFORMANCE
At runtime, the NextSilicon software suite intelligently adapts Maverick-2 ICA to the specific needs of your workload. A completely new architecture featuring distributed HBM memory access accelerates workloads that others can't.
WE DON'T
FOLLOW
INSTRUCTIONS
NextSilicon’s smart algorithms profile an entire application, identify the hotspots and the "most likely" flows, and optimize the hardware autonomously, on the fly, and without manual intervention.
ACCELERATE
WHAT MATTERS
Iterative rounds of telemetry and automated self-optimization deliver acceleration tailored to each workload. Experience increased developer productivity with up to 30% less tuning overhead and faster time-to-science.
BETTER
UTILIZATION,
MAXIMUM
OPTIMIZATION
Likely flows are tailored for a particular application by the runtime optimization process. Likely flows can then be replicated hundreds of times for massively parallel processing.
FUTURE-PROOF AND
ADAPTABLE TO NEW
WORKLOADS
Maverick-2 ICA uses built-in intelligence and adaptability to support both today's workloads and wherever tomorrow's HPC and AI innovations may take you.
Intelligent Compute
Accelerator
Close-Up
NextSilicon Maverick-2 is a versatile, software-defined hardware accelerator card for high-performance computing, available in two form factors: a compact, scalable single-die PCIe format, and a dual-die OCP Accelerator Module (OAM). The Maverick-2 ICA combines cutting-edge 5nm process technology, HBM3E memory, and high-bandwidth interfaces to deliver scalable, efficient performance. Whether in a single-die or dual-die configuration, Maverick-2 accelerates the most demanding HPC and AI workloads while reducing power consumption and maximizing efficiency.
NextSilicon software utilizes real-time telemetry and intelligent algorithms to dynamically optimize performance, power efficiency, and scalability for your most demanding HPC and AI workloads. This intelligent, software-defined adaptability not only maximizes immediate performance but future-proofs your infrastructure, allowing it to evolve alongside new applications and workloads. By removing the limitations of fixed-function accelerators, NextSilicon’s approach provides a long-term solution that evolves with your organization to deliver sustainable innovation.
Intelligent Acceleration, Meets Intuitive Developer Tools
Maverick-2 is paired with an intuitive suite of developer tools designed to provide real-time insights into performance and utilization. Our Profiler tool enables you to gain deep insights into how your code behaves on the Maverick-2 platform, offering real-time telemetry data that gives you a clear understanding of performance patterns and bottlenecks. For even more detailed analysis, our Chip Viewer provides live, hardware-level information, including in-depth projections and real-time results. The Projection Viewer goes a step further by visualizing what's happening inside each mill core in real time, ensuring you have a full understanding of how your code is accelerated. With Maverick-2, you are equipped with unprecedented transparency, ease of use, and the performance insights at your fingertips to enable you to drive your next breakthrough.
