Course Description

Modern video games employ a variety of sophisticated algorithms to produce groundbreaking 3D rendering pushing the visual boundaries and interactive experience of rich environments. This course brings state-of-the-art and production-proven rendering techniques for fast, interactive rendering of complex and engaging virtual worlds of video games.

This year the course includes speakers from the makers of several innovative games and game engines, such as Activision, Electronic Arts, Sony, IO Interactive, Roblox, Meta, and more. The presenters will cover a wide range of topics, from innovations in real-time ray tracing and path tracing techniques for multiplatform AAA games, to demystification of development of machine-learning based algorithms for upscaling, adaptive tessellation improvements on GPU, real-time volumetric effects, as well as efficient approaches for rendering kit-based UGC representations in dynamic worlds, and more.

This is the course to attend if you are in the game development industry or want to learn the latest and greatest techniques in the real-time rendering domain!

Previous years’ Advances course slides: go here

Syllabus

Advances in Real-Time Rendering in Games: Part I

Tuesday, 21 July 2026 9:00am - 12:15pm PDT

Location: Petree D

Advances in Real-Time Rendering in Games - Game and Rendering Trends, a 2026 Take.
Natalya Tatarchuk (Activision)

Speeding up Path Tracing via ORCA (Online Radiance Cache Acceleration)
Jon Greenberg (EA SEED)

Upgrading PSSR on PlayStation5 Pro
Daniel Craig (Sony Interactive Entertainment)

Variable Rate Ray Tracing in Call of Duty: Modern Warfare 4
Michał Olejnik (Activision Infinity Ward)

Closing Notes for Part I
Natalya Tatarchuk (Activision)

Advances in Real-Time Rendering in Games: Part II

Tuesday, 21 July 2026 2:00pm - 5:15pm PDT
Location: Petree D

Welcome and Introduction to Part II
Natalya Tatarchuk (Activision)

Smolder — Real-Time Volumetric Effect Rendering in Glacier and 007 First Light
Alexander Mueller (IO Interactive)

SLIM: Scaling User-Generated 3D Worlds on Roblox
Sergey Makeev (Roblox)

Adaptive Tessellation and Subdivision
John Hable (Meta)

Closing Notes for Advances in Real-Time Rendering in Games
Natalya Tatarchuk (Activision)

Prerequisites

Working knowledge of modern real-time graphics APIs like DirectX or Vulkan or Metal and a solid basis in commonly used graphics algorithms. Familiarity with the concepts of programmable shading and shading languages. Familiarity with shipping gaming consoles hardware and software capabilities is a plus but not required.

Intended Audience

Technical practitioners and developers of graphics engines for visualization, games, or effects rendering who are interested in interactive rendering.

Course Organizer

Natalya Tatarchuk is a graphics engineer and rendering enthusiast at heart, currently serving as Chief Technology Officer at Activision Publishing. In this role, she leads the technology strategy and execution across major Activision franchises - including Call of Duty - driving innovation at the intersection of cutting-edge tech and game development at scale.

Previously, Natalya was Distinguished Technical Fellow and Chief Architect, VP of Wētā Tools at Unity, where she advanced state-of-the-art rendering, graphics performance, and character creation tools for film and games. Before that, as VP of Graphics for the Unity Editor and Engine, she led Unity’s graphics initiatives across the real-time rendering stack.

Natalya’s roots in AAA game development include nearly a decade at Bungie, where she contributed to the groundbreaking visuals and engine architecture for Destiny and the Halo franchise—including Halo 3: ODST and Halo: Reach. She led the graphics group and contributed to engine development, the visual innovation and cross-platform rendering of Destiny franchise.

Earlier in her career, she worked at AMD’s Graphics Products Group, where she pushed the boundaries of parallel computing and explored advanced real-time graphics techniques, graphics hardware design, and next-generation API development.

One of Natalya’s passions is fostering knowledge-sharing in the real-time graphics community, as she strongly believes that advancing the state of the art is always more powerful when done together. For over two decades, she has organized and curated some of the industry's most influential technical forums, including the Advances in Real-Time Rendering, Open Problems in Real-Time Rendering, and Rendering Engine Architecture courses. Most recently, she has co-organized Rendering Engine Architecture conferences with a few gaming industry colleagues.

Talks

SPEEDING UP PATH TRACING VIA ORCA (ONLINE RADIANCE CACHE ACCELERATION)

Abstract: Path tracing is unfortunately quite slow on modern hardware and quite challenging to implement in real time game rendering, as a result. We present a straightforward approach to dramatically speed up path tracing via a custom radiance cache designed specifically for real-time rendering use. This technique, ORCA, does not depend on temporal history. All data structures are instantaneous and don't need to survive the frame, making it ideal for fully dynamic scenes. We'll discuss both how it's implemented and how one can hook it into an existing real-time path tracer with minimal changes.

Speaker Bio:

Jon Greenberg has been working in the game industry for over 26 years, spanning 4 generations of console gaming hardware. He cut his teeth at Midway Games Chicago, then continued on to NetherRealm, working on the high-performance demands of fighting games. Running the rendering team on the Mortal Kombat franchise led him to spending... perhaps too much time thinking about evolving violent gore tech in games, an entirely different way of interpreting the term "cutting edge graphics". Following work on "Injustice 2" and "Mortal Kombat 11" he left NetherRealm to join the SEED team at Electronic Arts where he's busy researching new rendering techniques.

UPGRADING PSSR ON PLAYSTATION5 PRO

Abstract: Machine-learning-based upscaling has become a core technology in real-time rendering, yet how these systems work is often opaque. This talk covers PlayStation Spectral Super Resolution, or PSSR, and specifically the recent redesign of Upgraded PSSR on PlayStation 5 Pro. The launch version of PSSR framed almost the entire reconstruction pipeline as a single Color-Predicting Network. This gave the model a wide surface area for optimization, but it also implicitly asked the model to solve a number of disparate problems that already had efficient closed-form solutions. The talk will show how that generality became a cost, and how the upgrade walked much of it back: restoring explicit closed-form solutions where possible and focusing the model on the pattern recognition it is best suited for. The result improved visual quality and temporal stability while reducing runtime, memory, and training cost by giving the model less to do, not more.

Speaker Bio:

Daniel Craig is a Principal Software Engineer at SIE, where he focuses on real-time rendering. Most recently, he was the lead researcher for Sony’s Upgraded PSSR ML upscaling library. He began his career at Adaptec, working on SCSI RAID controllers, before moving into the games industry about 22 years ago. His background in systems engineering shifted toward rendering around the launch of the PS3, and he has worked in rendering ever since.
Outside of work, Daniel is an avid mountain climber — a passion that grew from the realization that climbing is, at its core, practical real-time optimization.

VARIABLE RATE RAY TRACING IN CALL OF DUTY: MODERN WARFARE 4

Abstract: Real-time ray tracing in a fast-paced multiplayer shooter presents a unique challenge: image quality must remain stable under rapid camera motion and dynamic object movement - while frame time budgets remain unforgiving. In Call of Duty: Modern Warfare 4, traditional fixed-rate ray tracing with screen-space denoising proved insufficient: disocclusion artifacts and temporal lag are far too distracting during competitive play.

We present a production-proven Variable Rate Ray Tracing (VRRT) system designed to deliver consistent quality without sacrificing deterministic performance.

Our approach dynamically allocates ray budgets per pixel, supporting both sub-1 sample-per-pixel tracing (spatial undersampling with temporal reconstruction) and selective supersampling (>1 spp) where stability is critical. We leverage a temporal gradient pre-pass and explicit disocclusion detection to proactively identify unstable pixels and selectively increase sampling before artifacts appear. On top of that, we deploy a fully GPU-driven frame-level scheduler that redistributes rays spatially while maintaining a constant total ray count per frame. This guarantees stable GPU cost and eliminates performance spikes, even under heavy camera motion or large dynamic events.

Speaker Bio:

Michał Olejnik is a Senior Expert Rendering Engineer at Infinity Ward Poland, where he leads a team focused on real-time ray tracing and path tracing. He began his graphics career at Plastic, a demoscene-born independent studio, working on the PlayStation 4 exclusive experimental title Bound. In 2017, he joined the newly established Infinity Ward Poland, eventually focusing on ray tracing research and development. Since then, he has led the development of multiple ray tracing systems across the Call of Duty franchise, including Modern Warfare (2019) and Black Ops 7 (2025), as well as a prototype real-time path tracer in Modern Warfare III (2023). His current work focuses on advancing next-generation real-time ray tracing techniques.

SMOLDER — REAL-TIME VOLUMETRIC EFFECT RENDERING IN GLACIER AND 007 FIRST LIGHT

Abstract: In recent years, volumetric rendering techniques have become more popular for real-time rendering in video games. However, they are usually limited to specific use cases like clouds or volumetric fog, or they require a lot of custom setup work to achieve a believable result. With 007 First Light, we introduced Smolder, a fully integrated, scalable, real-time volumetric effect rendering framework, into our Glacier engine. Smolder enables us to use volumetric effects as if they were any other type of VFX. It is fully integrated with all lighting systems in Glacier, as well as with the various layers in the scene — opaque, transparent, and volumetric fog. This enables our artists to work freely with Smolder when building the environments, cinematics, and in-game moments, and it enables us to add player-controlled volumetric effects to Bond’s various gadgets. This presentation provides an in-depth look at how Smolder works, as well as the path we took to get there. It evaluates the benefits and drawbacks of the various parts of the Smolder algorithm and concludes with a look at where Smolder can go from here.

Speaker Bio:

Alexander Mueller is a Senior Rendering Engineer at IO Interactive, working on the custom Glacier engine. He specializes in lighting and performance work, with a focus on volumetrics and global illumination. Before joining IOI, he worked as a render programmer at Ubisoft Mainz.

SLIM: SCALING USER-GENERATED 3D WORLDS ON ROBLOX

Abstract: Roblox games are built from arbitrary user-generated content, kitbashed by millions of creators without traditional art-direction constraints, fixed content budgets, manually authored LODs, or predictable runtime behavior. These games must run across a broad range of devices with vastly different compute, memory, graphics, and network capabilities.

In this talk, we present SLIM - a cloud-based system that automatically generates lightweight, device-adaptive runtime representations of Roblox worlds while preserving authored appearance, behavior, and semantics. We cover the motivation, architecture, integration, and production challenges of building and deploying SLIM at scale across petabytes of existing user-generated content. We discuss the tradeoffs in representation generation, client selection, fidelity preservation, and fallback to original content when full runtime semantics are required. We close with future directions for device-adaptive representations and server-assisted rendering.

Speaker Bio:

Sergey Makeev is a veteran graphics engineer and technical leader, currently serving as a Senior Technical Director at Roblox. With over twenty years of experience in AAA game development and massive-scale platforms, he specializes in architecting high-performance rendering stacks and core engine technologies. At Roblox, Sergey architected SLIM, a platform that automatically optimizes the performance and visual fidelity of Roblox worlds across millions of devices. He also pioneered Roblox's Layered Clothing system, allowing multi-layered garments to automatically conform to any avatar body type. Additionally, he headed the "Future Is Bright" initiative, modernizing the Roblox rendering engine and moving it to physically-based rendering and advanced lighting.

Earlier in his career, Sergey served as a Technical Director, building several multiplatform engines from scratch and providing technical leadership for major titles like Skyforge, Warface, and Armored Warfare. He is an active contributor to the real-time graphics community and has published his technical work in both GPU Pro 5 and GPU Zen 2.

ADAPTIVE TESSELLATION AND SUBDIVISION

Abstract: This talk addresses the problem of tessellation and subdivision using compute shaders. We start with the tessellation pattern itself, introducing clamped parallelogram tessellation and discussing tradeoffs with existing tessellation patterns. We then scale this up into a full compute-based pipeline to tessellate meshes with screen-adaptive tessellation while welding seams. And finally, we extend this framework to Catmull-Clark subdivision with semi-sharp creases.

Speaker Bio:

John Hable is a rendering engineer at Meta. He has 21 years of experience working for Electronic Arts, Naughty Dog, Unity, and Epic. Recently, he has focused on visibility buffer rendering and scalable geometry.

Direct contact (note: responses may be delayed):

ceruleite.bsky.social
Mastodon @mirror2mask