Turing (microarchitecture)

Turing (microarchitecture)
Fritzchens Fritz · CC0 · source
Name	Turing
Developer	NVIDIA Corporation
Produced start	2018
Produced end	2020
Process	12 nm
Cores	CUDA, RT, Tensor
Architecture	Volta successor
Predecessor	Pascal (microarchitecture)
Successor	Ampere (microarchitecture)

Turing (microarchitecture) is a GPU microarchitecture developed by NVIDIA Corporation and introduced in 2018 as a successor to Pascal (microarchitecture), targeting markets including gaming, workstation, data center, and professional visualization. It debuted alongside products announced at events such as Gamescom and NVIDIA GPU Technology Conference, with design goals influenced by trends in real-time rendering, machine learning, ray tracing, and high-performance computing.

Overview

Turing was announced by NVIDIA Corporation executives at presentations featuring collaborations with companies like Microsoft, Electronic Arts, Epic Games, Autodesk, and Adobe Systems, positioning the microarchitecture for integration into platforms from Asus, Gigabyte Technology, MSI, Dell Technologies, and HP Inc.. The architecture introduced dedicated hardware blocks for ray tracing and deep learning inspired by research from institutions such as Stanford University, Massachusetts Institute of Technology, University of California, Berkeley, and industrial labs including DeepMind and OpenAI. Its release intersected with industry standards and APIs developed by Khronos Group, SIGGRAPH, Microsoft DirectX, and Vulkan (API), influencing adoption in engines like Unreal Engine, Unity (game engine), id Software, and middleware from NVIDIA GameWorks.

Architecture and Design

Turing integrated traditional CUDA cores with new tensor cores and RT cores, reflecting influences from prior work at NVIDIA Research and comparisons to designs from AMD, Intel Corporation, and custom accelerators from Google and Amazon Web Services. The microarchitecture emphasized concurrent execution of integer and floating‑point operations, asynchronous compute for workloads seen in Blizzard Entertainment titles and Bethesda Softworks engines, and updated memory hierarchies using GDDR6 memory from suppliers like Micron Technology, Samsung Electronics, and SK Hynix. On-chip innovations referenced techniques found in publications from IEEE conferences and standards from ISO bodies, while implementation details mirrored design practices at fabs such as TSMC and GlobalFoundries.

Performance and Features

Turing's performance claims centered on improvements for real‑time ray tracing via dedicated RT cores and acceleration for deep learning via tensor cores, with benchmarks discussed by outlets including AnandTech, Tom's Hardware, TechPowerUp, PC Gamer, and Digital Foundry. Features such as variable rate shading, mesh shading, and concurrent multi‑issue execution targeted rendering techniques used in Battlefield V, Control (video game), Metro Exodus, and creative applications from Blackmagic Design and Foundry (software). Power and thermals were debated in reviews referencing thermal solutions from Cooler Master, NZXT, and OEM designs from Lenovo and Acer Inc., while comparisons invoked compute measurements from LINPACK, SPEC, 3DMark, and machine learning workloads using TensorFlow, PyTorch, and libraries maintained by NVIDIA CUDA Toolkit.

Product Implementations

NVIDIA shipped Turing across product lines including consumer GeForce RTX 20 Series, professional Quadro RTX branded cards, and OEM modules for NVIDIA DGX systems and virtualization platforms used by companies like VMware and Citrix Systems. Laptop implementations appeared in thin‑and‑light models from Razer, Alienware, MSI, and ASUS Republic of Gamers, while workstation integrations targeted customers through HP Z Workstation and Lenovo ThinkStation lines. Third‑party board partners such as EVGA, ZOTAC, Palit, and PNY Technologies produced variants with custom coolers, BIOS options, and factory overclocks, and enterprise deployments were adopted by cloud providers like Google Cloud Platform, Microsoft Azure, and Amazon Web Services.

Software and Driver Support

Driver support for Turing was delivered via NVIDIA GeForce Game Ready Driver, NVIDIA Studio Driver, and enterprise releases coordinated with Microsoft Windows, Linux kernel maintainers, and distributions such as Ubuntu (operating system), Red Hat Enterprise Linux, and SUSE. Software ecosystems included integrations with DirectX Raytracing (DXR), Vulkan Ray Tracing, and developer tools from NVIDIA Nsight, CUDA Toolkit, cuDNN, and frameworks promoted by Khronos Group, SIGGRAPH, and standards bodies. Game developers from Activision, Ubisoft, Square Enix, and Capcom enabled Turing features through engines like Frostbite, RE Engine, and Snowdrop, while professional applications from Autodesk, Adobe Systems, SideFX, and Adobe Premiere Pro received optimized updates.

Reception and Legacy

Turing received mixed critical reception in coverage by The Verge, Wired, Bloomberg, Reuters, and TechCrunch, praised for ray tracing capabilities yet critiqued for price and power consumption relative to competing offerings from AMD Radeon Technologies Group and upcoming products from Intel Arc. Its introduction accelerated ecosystem support for hardware ray tracing, influencing subsequent architectures such as Ampere (microarchitecture) and industry adoption in consoles like PlayStation 5 and Xbox Series X/S through partnerships with Sony Interactive Entertainment and Microsoft Xbox. Turing's legacy includes stimulating research in real‑time rendering, contributions to professional visualization workflows at organizations such as NASA, CERN, and leading visual effects studios like Industrial Light & Magic and Weta Digital, and shaping standards work in Khronos Group and developer practices at Epic Games.

Category:Graphics processing units