The revolutionary NVIDIA Pascal™ architecture is purpose-built to be the engine of computers that learn, see, and simulate our world—a world with an infinite appetite for computing. From silicon to software, Pascal is crafted with innovation at every level.
The NVIDIA Pascal architecture is built on five technological breakthroughs, enabling a new computing platform that’s disrupting conventional thinking from the desk-side to the data center.
With 150 billion transistors built on bleeding-edge 16 nanometer FinFET fabrication technology, the Pascal GPU is the world's largest FinFET chip ever built. It’s engineered to deliver the fastest performance and best energy efficiency for workloads with near-infinite computing needs.
Pascal is the most powerful compute architecture ever built inside a GPU. It transforms a computer into a supercomputer that delivers unprecedented performance, including over 5 TeraFLOPS of double precision performance for HPC workloads. For deep learning, a Pascal-powered system offers over 12X leap in neural network training—reducing training time from weeks to hours—and a 7X increase in deep learning inference throughput compared to current-generation GPU architectures.
Pascal is the first architecture to integrate the revolutionary NVIDIA NVLink™ high-speed bidirectional interconnect. This technology is designed to scale applications across multiple GPUs, delivering a 5X acceleration in interconnect bandwidth compared to today's best-in-class solution.
The Pascal architecture unifies processor and data into a single package to deliver unprecedented compute efficiency. Using an innovative approach to memory design, CoWoS® (Chip-on-Wafer-on-Substrate) with HBM2 gives you a 3X boost in memory bandwidth performance over the NVIDIA Maxwell™ architecture.
New half-precision, 16-bit floating point instructions deliver over 21 TeraFLOPS for unprecedented training performance. With 47 TOPS (tera-operations per second) of performance, new 8-bit integer instructions in Pascal allow AI algorithms to deliver real-time responsiveness for deep learning inference.
NVIDIA launched the ultra-efficient NVIDIA Maxwell architecture in 2014 to deliver innovative design that dramatically improves energy efficiency. This architecture provides substantial application performance improvements over prior architectures by featuring large dedicated shared memory, shared memory atomics, and more active thread blocks per SM.
The NVIDIA Kepler™ architecture powers tens of thousands of data centers worldwide, from the world's largest supercomputers to university computing systems. The NVIDIA® Tesla® K80, the latest accelerator based on the Kepler architecture, dramatically lowers data center cost while delivering tremendous performance for high-performance computing customers.
Get Pascal now in the NVIDIA® DGX-1™ system, purpose-built for deep learning.
 | NVIDIA Blog |
 | |
 | |
 | YouTube |
