NextSilicon, an innovative pioneer in high-performance computing, today announced the launch of Maverick-2, the industry's first intelligent computing accelerator (ICA), which stands out as a revolutionary breakthrough in computing. Designed to deliver unmatched performance and efficiency for HPC, AI, and vector database applications, Maverick-2 is designed to deliver unmatched performance and efficiency for HPC, AI, and vector database applications with intelligent software-defined hardware acceleration, real-time adaptability, and superior performance. In addition to the launch of Maverick-2, NextSilicon also demonstrated its strong capital strength, successfully raising $303 million in funding from high-profile investors including Aleph, Amiti, Liberty Technology VC, Playground Global, Standard Investments, StepStone, and Third Point Ventures. This combination of funding and technological innovation enables NextSilicon to redefine the future of high-performance computing through its revolutionary intelligent computing accelerator architecture.
James H. Laros III, a technician at Sandia National Laboratories and head of the Pioneer Program, said, "Sandia Labs is already using the Maverick-2 accelerator and plans to deploy a new integrated system, Spectra, later this year, which will be delivered in 2025. "Our partnership with NextSilicon for more than four years is not only a great example of NNSA Labs' collaboration with industry partners to refine emerging technologies, but also a testament to NextSilicon's continued innovation in HPC."
Maverick-2 innovates on the accelerator architecture, breaking through the limitations of traditional fixed GPU designs based on the von Neumann architecture. It combines advanced telemetry technology with NextSilicon's patented algorithms to dynamically adjust the acceleration effect in real-time according to the needs of different applications. This adaptability allows Maverick-2 to deliver more than 4x more performance per watt than traditional GPUs and more than 20x more performance per watt than high-end CPUs, while cutting operating costs by more than half.
Optimized for critical code and workflows that frequently execute in HPC and AI environments, the Maverick-2 ICA's self-optimizing architecture enables immediate and scalable performance gains. This is achieved by creating custom software-defined processor cores for each application, and while the application is running, telemetry data is fed into NextSilicon's intelligent algorithms to optimize performance, power, and utilization in real time. As a result, Maverick-2 efficiently scales performance for HPC applications that require extreme performance and reduces power consumption by 50 to 80 percent compared to traditional GPUs.
Figure：NextSilicon unveils Maverick-2：industry first intelligent computer accelerator
Steve Conway, Senior Analyst, noted, "Traditional CPU and GPU architectures are often constrained by high-latency pipelines and limited scalability. Clearly, we need to reduce energy waste and unnecessary computation in our HPC and AI infrastructure. NextSilicon solves these critical problems with Maverick-2, an entirely new architecture specifically designed to meet the specific needs of HPC and AI. Maverick-2 is expected to break down the barriers to migration of traditional architectures and accelerate innovation in this area.”
While other alternative accelerators have been challenging the dominance of traditional GPUs and CPUs for years, they often face the complexity of application porting and vendor-specific framework lock-in. NextSilicon's Maverick-2 ICA is the first to break through these barriers with a robust, user-friendly software suite. Maverick-2 natively supports popular programming languages such as C/C++, FORTRAN, OpenMP, and Kokkos, and many applications can be run on Maverick-2 without modification. This software compatibility dramatically reduces development time and shortens the cycle time for research and insight. NextSilicon also plans to integrate CUDA, HIP/ROCm, and mainstream AI frameworks to fully support customers' application needs.
Maverick-2 not only solves today's HPC commercial and research computing challenges, but also incorporates NextSilicon's unique intellectual property designed to address tomorrow's evolving workloads. Its intelligent and flexible architecture can easily address new use cases such as converged HPC-AI workloads, vector database applications, AI model breakthroughs, and emerging generative AI applications.
Phil Pokorny, CTO of Penguin Solutions, said: "We are very pleased to support the launch of the Maverick-2 Intelligent Compute Accelerator and see its great potential to improve performance and efficiency in the HPC market. Our company excels at turning cutting-edge ideas into reality, driving faster, larger outcomes to help our customers tackle their most demanding workloads. Maverick-2's intelligent software-defined approach complements our existing technology, giving organizations a powerful tool to optimize their infrastructure and make breakthroughs in HPC and AI. As application complexity continues to increase, innovations like Maverick-2 will open up new possibilities for customers and accelerate scientific and commercial advancements.”
