<p>➀ NextSilicon's Maverick-2 accelerator chip adopts a dataflow architecture, emphasizing FP64 compute performance for HPC applications, differentiating itself from NVIDIA's shift away from FP64.</p><p>➁ The chip utilizes HBM3e memory and optimizes ALU density by reducing traditional CPU overheads (e.g., branch prediction), while its software identifies computational hotspots for efficient offloading to the accelerator.</p><p>➂ Maverick-2 employs adaptive "Mill Cores" to reconfigure dataflow paths dynamically, enabling high parallelism and flexibility akin to FPGAs, with early adoption by Sandia National Lab for real-world deployment.</p>
Related Articles
- AMD's Instinct MI325X smiles for the camera: 256 GB of HBM3E10 months ago
- This is the Massive NVIDIA 800G OSFP to 2x 400G QSFP112 Passive Splitter DAC Cableabout 6 hours ago
- Intel posts return to growth and profitability in Q3 2025, but significant challenges remain — achieves $13.7 billion revenue with $4.1 billion operating profit2 days ago
- Flying Assistants for Atmospheric Research: EAH Jena Receives Funding for Pioneering MAVAS Project3 days ago
- French data centre AI inference chip ‘in production’3 days ago
- Quantum Chip Solves Calculations 13000x Than Fastest Supercomputer3 days ago
- Reference-Class Linear Actuators Available for Fast Delivery4 days ago
- Formnext 2025: Fraunhofer IAPT Expands the Automation of Additive Manufacturing5 days ago
- Trump says Intel has made a 'fortune' and America has made $40 billion after the US invested in ailing chipmaker — Intel has gained $73 billion in market cap since investment5 days ago
- Empa Entrepreneur Fellowship Awarded - Three Researchers Receive Start-up Support5 days ago