NVIDIA RTX
By NVIDIA
NVIDIA RTX is NVIDIA's brand and hardware architecture line for GPUs that include dedicated ray-tracing and AI tensor cores, spanning consumer GeForce RTX gaming cards, professional RTX workstation cards, and RTX-branded mobile and…
Definition
NVIDIA RTX is NVIDIA's brand and hardware architecture line for GPUs that include dedicated ray-tracing and AI tensor cores, spanning consumer GeForce RTX gaming cards, professional RTX workstation cards, and RTX-branded mobile and embedded chips. The defining hardware feature is a set of RT cores that accelerate the geometric calculations behind real-time ray tracing, paired with tensor cores that power AI-based rendering features such as denoising and upscaling. RTX distinguishes newer NVIDIA GPU generations from older ones that lacked this dedicated ray-tracing hardware.
Overview
RTX describes both a GPU architecture feature set and the products built on it, introduced by NVIDIA to bring real-time ray tracing, a rendering technique previously reserved for offline film production, into interactive applications running at tens of frames per second. Before RTX, real-time 3D graphics relied almost entirely on rasterization, a faster but less physically accurate technique for approximating light and shadow. RTX hardware sits alongside traditional shader cores, giving developers the option to blend rasterization with true ray-traced lighting for scenes that need it. Mechanically, an RTX GPU adds two specialized unit types to NVIDIA's standard shader cores. RT cores perform the ray-triangle intersection math needed to trace how light rays bounce through a 3D scene, work that would be prohibitively slow on general shader cores alone. Tensor cores, originally developed for AI training and inference, are repurposed to run the neural networks behind features like DLSS, which upscale a lower-resolution rendered frame and denoise ray-traced images that would otherwise look grainy at real-time frame rates. Software must be written to specifically call these ray-tracing APIs to benefit. Within NVIDIA's lineup, RTX branding spans multiple products: GeForce RTX cards for gaming, RTX-branded professional workstation cards for CAD, rendering, and design software with certified drivers, and RTX-enabled laptop chips for mobile workstations and gaming notebooks. This differs from NVIDIA's data center GPUs like the H100, which include tensor cores for AI but typically omit ray-tracing hardware and display outputs entirely, since their target workload is AI training and inference rather than graphics rendering. In practice, RTX hardware is used by game developers to add realistic reflections, shadows, and global illumination to titles, by architects and product designers to preview photorealistic renders interactively rather than waiting on an offline render, and by visual effects studios for faster look-development iterations before a final offline render. The tensor cores on RTX cards are also used informally by developers and hobbyists to run small AI inference workloads locally, borrowing compute originally intended for upscaling. The main trade-off with ray tracing is performance cost: even with dedicated RT cores, fully ray-traced scenes can still tax a GPU heavily, which is why AI upscaling features are commonly paired with ray tracing to recover frame rate. Ray-traced effects also require developers to specifically implement support, so older or unsupported titles gain nothing from RTX hardware. Buyers whose workloads do not involve ray tracing or AI upscaling, such as basic office or media playback use, generally do not need an RTX-tier card and can use lower-cost GPUs instead.
Key Features
- Dedicated RT cores accelerate real-time ray-triangle intersection calculations
- Tensor cores power AI upscaling and denoising features like DLSS
- Spans consumer GeForce RTX, professional workstation, and mobile chip variants
- Enables real-time global illumination, reflections, and ray-traced shadows in games
- Requires developer-side implementation of ray-tracing APIs to take effect
- Distinguishes newer NVIDIA GPU generations from earlier rasterization-only designs
- Used informally for local AI inference via tensor cores
- Professional RTX cards add certified drivers for CAD and design software