How Nvidia's Turing architecture turned light into math

Technology · Published 20 October 2026 · Updated 21 October 2026 Swipe more like this →

In short: For decades, computer graphics relied on clever tricks to mimic reality. Turing changed the game by introducing specialized hardware that actually calculates how light behaves. By moving from traditional rasterization to real-time ray tracing, this architecture allows for realistic reflections and shadows by simulating physical light behavior in consumer products.

The Turing microarchitecture, introduced by Nvidia in August 2018, marked a pivotal shift in graphics processing. While its predecessor, Volta, was primarily for high-performance computing, Turing brought specialized hardware to the consumer market. It was the first architecture capable of real-time ray tracing—a long-standing goal in the industry that simulates the physical behavior of light to produce realistic reflections, refractions, and shadows.

To achieve this, Turing utilizes a multi-core approach. It combines standard CUDA cores for traditional rasterized shaders and compute tasks with two new specialized types: RT cores and Tensor cores. RT cores are specifically designed to process complex geometric structures like quadtrees and spherical hierarchies, accelerating collision tests with individual triangles. Meanwhile, Tensor cores handle deep learning tasks. These cores can perform de-noising—filling in the blanks of a partially rendered image—and use methods developed on supercomputers to increase image resolution.

The engineering behind Turing is massive. Manufactured using TSMC's 12 nm FinFET process, the high-end TU102 GPU contains 18.6 billion transistors. It also features a memory controller supporting GDDR6 and HBM2, with the RTX 20 series notably switching from Micron to Samsung memory chips in November 2018. While ray tracing provides a significant leap—Nvidia reported an 8x performance increase over the Pascal architecture—it does not replace rasterization. Instead, it augments it, using physical light data to enhance photo-realism for off-camera actions.

This architecture is accessible through several APIs, including Microsoft's DXR, OptiX, and Vulkan. It also includes specialized hardware like the Nvidia Optical Flow Accelerator for video interpolation and PureVideo Feature Set J for hardware video decoding. By combining these specialized cores, Turing allows for a hybrid rendering pipeline where traditional techniques and modern physics-based calculations work in tandem.

General information, not professional advice. Spotted an error? Flag the card or see corrections and complaints.

Intellect Plus is a visual educational discovery platform for exploring fascinating facts and ideas. Brain Food, Not Brain Rot. What is Intellect Plus?

Related

• How a chip designed for video games became the engine of modern artificial intelligence
• How Nvidia transformed the humble graphics card into the engine of modern artificial intelligence
• How NVIDIA achieved real-time ray tracing, a feat once considered computationally impossible.
• How a student's 32-card gaming rig birthed the engine of modern AI
• Why the modern graphics card is no longer just for video games
• From black car luxury to global logistics, how Uber reshaped the modern commute
• How the Transmission Control Protocol keeps the modern internet from falling apart
• Why do near-perfect human replicas trigger a deep, visceral sense of revulsion?
• Why the transformer architecture changed everything about how machines process human language
• Unicode is the invisible language that allows your computer to speak every human tongue.
• Why rational individual choices often lead to the collective destruction of shared resources
• Is the Pentagon rebranding to the Department of War a return to its roots?

More in Technology today · More in Technology · Latest · A–Z · All topics