Extreme Performance
NVIDIA RTX products deliver better graphics, compute, and AI performance than prior generation NVIDIA products. Unprecedented CUDA Core counts with parallel integer and floating point pipelines, RT Cores for real-time ray tracing, advanced shading, and enhanced VR capabilities are all available. Advanced multi-precision Tensor Cores dramatically accelerate the latest AI, DL, MO, MV, or big data analytics applications like Graphistry or OmniSci. NVIDIA Ada Generation Tensor Cores include Transformer Engine functionality, making them ideal for the generative AI revolution reshaping entire academic disciplines, institution affiliated technology incubators, and industries across the globe
Faster Data
NVIDIA RTX supports the latest GDDR6 memory with optional ECC (Error Correcting Code), providing significantly higher bandwidth. With anywhere from 20GB to 48 GB of GPU memory (40GB or 96GB with NVLink, Ampere architecture products only, application support required) capacities are unprecedented as well. GPU memory is particularly important for RT Core accelerated real-time ray tracing which requires the model or scene to fit into GPU memory for maximum performance. AI training is another example of when having greater GPU memory significantly reduces training time while delivering extremely low-latency inferencing.
Advanced GPU Features
NVIDIA RTX Ada Generation delivers the most sophisticated graphics and compute capabilities ever offered. Gen 3 RT cores deliver even higher photorealistic real-time ray tracing in application viewports and new features like motion blur. Path-traced rendering is also supported. VR benefits from Variable Rate Shading and Foveated Rendering, when used with an eye-tracking VR HMD. Gen4 Tensor Cores support TF32 (Tensor Float 32), which can deliver up to 10x the performance gains without code changes. A new FP8 precision increases application performance by 2x and reduces memory requirements by 2x and works with Transformer Engine hardware and software development frameworks, pre-trained models, and libraries. Ada Lovelace and Ampere also implement structural sparsity to deliver 2x faster inferencing from smaller fully trained and deployed models. From a generative AI. data science or big data analytics perspective, NVIDIA RTX delivers end-to-end acceleration of incredibly complex AI, making NVIDIA RTX the best choice for generative AI, machine vision, or data science teaching and research (among other possibilities).
Rich Visual Workspace
NVIDIA RTX DisplayPort 1.4a outputs can drive up to 4 directly attached 8K displays with full HDR (High Dynamic Range) support, enabling stunning visual quality and an ideal platform on which to build advanced simulation and CAVE environments or to drive the latest VR HMDs. NVIDIA RTX Desktop Manager software optimizes the use of screen real estate across single or multiple high-resolution displays. The NVIDIA RTX Ada Generation or Ampere architecture boards, except the RTX A2000 12GB, support the NVIDIA Quadro Sync II board that provides analog (Genlock) or digital (Frame Lock) HW synchronization for immersive CAVE environments, rendering clusters, interactive digital signage and displays, or live events and broadcasting – area of academic interest ranging from FP32 HPC scientific and technical computing to the fine arts and humanities.