FIELD NOTE / 2026.09.124 MIN READ / 5 SOURCES

RTX and the Return of Ray Tracing to Real-Time Graphics

Microsoft's DXR API and NVIDIA's 2018 Turing/RTX hardware made ray queries a practical part of real-time game rendering, reviving a decades-old algorithm inside hybrid rasterization pipelines.

Ray tracing spent decades on the wrong side of the frame-time budget

Turner Whitted’s 1980 recursive ray tracer showed that rays could produce reflections, refractions and shadows with conceptual elegance, but the required intersection work was expensive.[5] For film and offline rendering, minutes or hours per frame could be acceptable; games needed a new frame every few milliseconds.

As a result, real-time graphics evolved primarily around rasterization plus carefully engineered approximations for shadows, reflections and indirect light. The 2018 RTX transition did not erase that history. It made selected ray queries cheap enough to join the real-time pipeline.

The milestone was hybrid rendering, not the death of rasterization

Early RTX games generally rasterized most primary visibility and then used ray tracing for a subset of effects such as reflections, shadows or global-illumination samples.

Microsoft introduced DirectX Raytracing as an API extension in March 2018

Microsoft announced DirectX Raytracing, or DXR, on March 19, 2018 as an extension to Direct3D 12.[1] The API supplied abstractions for acceleration structures, ray dispatch and programmable hit or miss behavior.

That software step mattered because hardware acceleration needs a portable programming model. Developers could express ray-tracing workloads through DirectX rather than writing one entirely proprietary path for each vendor.

Acceleration structures became part of the API contract

Efficient ray tracing depends on spatial data structures that avoid testing every ray against every triangle. DXR made the construction and traversal of those structures an explicit part of the graphics workload.

GDC 2018 framed ray tracing as the next DirectX graphics transition

Microsoft’s GDC material presented DXR as the next major graphics capability after decades in which DirectX had primarily supported raster-based real-time rendering.[2] Demos from multiple engine and hardware partners were important because an API without ecosystem support would have remained a specification rather than a platform shift.

The release therefore coordinated operating-system software, game engines, GPU vendors and content developers around a shared model for real-time rays.

NVIDIA’s Turing architecture added dedicated RT Cores

At SIGGRAPH 2018 NVIDIA announced the Turing architecture with new RT Cores intended to accelerate ray traversal and triangle intersection, plus Tensor Cores for AI inference.[3] NVIDIA explicitly described the design as a hybrid rendering architecture that combined rasterization, ray tracing, simulation and AI.

This specialization attacked the part of ray tracing that general shader hardware handled inefficiently. Dedicated units could perform the repeated bounding-volume and intersection work at much higher throughput.

Hardware acceleration changed which effects fit inside one frame

The key engineering question was not whether a path-traced image could be rendered eventually. It was whether enough useful rays could be traced within a 16-millisecond or similar interactive budget to justify changing the game renderer.

GeForce RTX brought the architecture to the gaming market

On August 20, 2018 NVIDIA unveiled the first GeForce RTX gaming GPUs based on Turing and promoted real-time ray tracing as a defining feature of the generation.[4] Titles including Battlefield V and Shadow of the Tomb Raider were announced with RTX effects.

The launch made ray tracing a consumer feature rather than only a professional visualization capability. It also exposed the limits of first-generation hardware: developers had to choose carefully where rays produced enough visible benefit to justify their cost.

Denoising and temporal reuse became essential companions to rays

A fully converged path-traced image may require many samples per pixel, far more than an interactive renderer can afford. Real-time systems therefore combine a small number of noisy ray samples with spatial and temporal filtering, motion information and other reconstruction techniques.

The result is a broader historical lesson: hardware ray tracing alone does not deliver real-time photorealism. It works as part of a pipeline containing rasterization, sampling strategies, denoisers and increasingly machine-learning-based reconstruction.

RTX reconnected modern games with classic graphics research

Whitted-style reflections and shadow rays became newly practical, while more advanced real-time methods drew from decades of research in Monte Carlo integration, global illumination and acceleration structures.[5]

This is why the 2018 transition feels both new and old. The algorithms had deep academic histories; the innovation was the combination of standardized APIs, specialized hardware and production techniques that moved selected forms of ray tracing into a mass-market frame budget.

Old algorithms can become new products when hardware economics change

Computer graphics repeatedly revisits ideas that were once too expensive. A technique can move from research paper to offline production to interactive use as compute, memory and specialized acceleration cross new thresholds.

Why RTX belongs in graphics history

Microsoft’s DXR announcement established a standardized Direct3D path for ray-tracing workloads.[1] Its GDC material positioned the technology as an ecosystem transition.[2] NVIDIA’s Turing architecture supplied dedicated hardware, and GeForce RTX carried it into consumer gaming.[3][4]

Whitted’s 1980 paper makes the continuity unmistakable.[5] RTX did not invent ray tracing; it marked a hardware-and-software threshold at which a decades-old rendering abstraction could become a regular part of real-time graphics.

RESEARCH / PROVENANCE

Works Cited

5 SOURCES
  1. 01
  2. 02
  3. 03
  4. 04
  5. 05

CodeHistory is a living archive. Citations document the evidence used for this edition; later evidence may refine the account.

Contribute / Corrections

Improve the record.

Use this moderated submission form to suggest a correction, provide a source, challenge a priority claim or identify a missing contributor. Submissions are treated as research leads, not automatically published comments.

Submit a research lead

Please do not submit confidential material or claims you cannot support.