A
AT HEART

Designed for HPC, Maverick-2 delivers efficient performance at any scale. Move beyond the status quo and the pain of porting. Experience what intelligent acceleration can do for you.

SCROLL
meet us
at our booth
#1419

INTRODUCING THE
INTELLIGENT COMPUTE
ACCELERATOR

Maverick-2, the first in a new subcategory of acceleration technology: the Intelligent Compute Accelerator (ICA), introduces a fundamentally new approach that combines intelligent software-defined hardware with real-time optimization, designed to deliver the best performance, efficiency, and flexibility across any workload.

TECHNOLOGY THAT ADAPTS TO YOU

Compute built for HPC

Maverick-2 surpasses the limits of traditional CPUs & GPUs by offering unparalleled computational performance, and a unique adaptable architecture that reduces application bottlenecks, freeing scientists to pursue breakthroughs without being hindered by technical hiccups. Maverick-2 seamlessly adapts to today's HPC and AI/ML workloads as it offers the opposite of the rigid, high-latency designs, and high power consumption of its competitors.

Embrace Intelligent Acceleration

NextSilicon's accelerators provide out-of-the-box performance, using a revolutionary software-learning algorithm that dynamically optimizes hardware configurations at runtime, reducing the time to science. Maverick-2 optimizes computational performance by combining this flexible hardware with intelligent software algorithms. These algorithms utilize runtime information to analyze an application's behavior and efficiently accelerate the most computationally intensive code.

Enjoy effortless porting

Maverick-2’s intelligent developer toolchain for HPC & AI operates with no code changes or software stack reinvestment, freeing you from vendor lock-in. It supports popular programming languages and frameworks natively, like C/C++, FORTRAN, OpenMP, and Kokkos, with upcoming integrations planned for CUDA, HIP/ROCm, and leading AI frameworks. This enables many applications to run without modification, simplifying porting, and eliminating the need for a proprietary software stack while reducing time-to-science by weeks, months, or even years.

Make the future greener

Going green has never been so easy. Maverick-2 sets a new standard for efficient computing with more than 4x performance-per-watt advantage over traditional GPUs more than 20x that of high-end CPUs, while reducing total cost-of-ownership (TCO) by more than half. By intelligently optimizing for power efficiency in real-time, Maverick-2 enables sustainable innovation, making it the ideal choice for data centers and researchers striving to meet environmental and operational goals without compromising on performance.

INNOVATIVE BY NATURE.
POWERFUL BY DESIGN.

HOST INTERFACE

PCIe Gen 5 (x16) card

PROCESS TECH

TSMC 5nm

FREQUENCY

1.5 GHz

PACKAGING TECH

2.5 D

MEMORY CAPACITY

96 GB

MEMORY TECH

HBM3E

MAX POWER

300 W

CACHE COHERENCE

128 MB

PROGRAMMABLE MODEL

Any program

HOST INTERFACE

PCIe Gen 5 (x16) OAM

PROCESS TECH

TSMC 5nm

FREQUENCY

1.5 GHz

PACKAGING TECH

2.5 D

MEMORY CAPACITY

192 GB

MEMORY TECH

HBM3E

MAX POWER

600 W

CACHE COHERENCE

256 MB

PROGRAMMABLE MODEL

Any program

INTELLIGENTLY
ACCELERATE

Built on principles of efficient performance, adaptability, and future-proof design. Maverick-2 ICA is designed to seamlessly adapt to future demands due to intelligent software-defined hardware with real-time optimization, designed to deliver the best output across any workload. Depart from fixed CPU and GPU architectures, and embrace purpose-built intelligent acceleration to power your legacy, new, and future applications.

YOUR NEXT
INNOVATIONS

Adaptable to support your evolving future of HPC, AI, and more.

DIVE DEEPER

STAY UP
TO SPEED

Sign up for our NextLetter and be the first to receive news and updates on our groundbreaking innovation, white paper, performance benchmarks, event reports, clients' success stories, and events. Let's connect!