Game Developer Conference 2025 Retrospective
- Lingheng Tony Tao
- Mar 26
- 15 min read
1 Abstract
This report documents the experiences and key takeaways of Lingheng Tony Tao from attending the Game Developers Conference 2025 (GDC). It covers both the technical knowledge and soft skills gained throughout the event, with a particular focus on sessions related to graphics rendering, optimization, AI in games, and game design philosophy. In addition to the professional content, the report also includes reflections on the social activities and networking opportunities Tony participated in during the conference. Through this report, Tony aims to capture the valuable insights, inspirations, and lessons learned from one of the most influential events in the game development industry.
2 Technical
I attended GDC as a student, graphics engineer, and technical artist. At the Entertainment Technology Center, my concentration has been on rendering techniques and tool development, so I chose to focus on lectures, sessions, and talks related to these areas.
The talks were incredibly, astonishingly refreshing and inspiring. However, they also prompted me to reflect on whether I truly understand the industry and to reconsider the direction of my own career path.
2.1 How Have I Learned
For each of the section I attended, I kept taking notes on my MacBook — here’s a screenshot:
As you can see in the screenshot, I did have a syntax for taking my notes.
The talks are organized into sections. In some cases, the speakers provided clear divisions in their slides; in others, I summarized the structure based on what I heard.
Keywords are highlighted in bold.
Topics I’m unfamiliar with are marked with a ○ to-do toggle, indicating that I plan to research them further.
Important points are underlined for emphasis.
The syntax kept the notes highly readable, and clearly help me organize what I learnt from the talks.
2.2 What Have I Learned
This part I will only cover the technical knowledge I gained from GDC.
First of all, here are the key takeaways from the Advanced Graphics lectures. These sessions were presented by representatives from companies like NVIDIA, AMD, Unity, Unreal, and Google, and many others, covering new features enabled by recent technical advancements—such as the release of the 50-series graphics cards.
The talks I attended could be categorized into the following:
Physically Based Rendering topics
Of course, the industry takes a very different perspective on PBR. The game industry’s mission is to deliver compelling experiences to non-technical players - people who neither know nor care about the technical complexity behind the algorithms. What these players do care about are:: - Does this look good?- Does this perform well?Almost every talk in PBR discusses how they achieve different goals. The talks (0317 Advanced Graphics: RTX for Indiana Jones, and 0317 Advanced Graphics: Full Ray Tracing in Black Myth Wukong) covered the following techniques that worth researching and learning:
Opacity MicroMaps (OMM). This is a technique designed to address the challenges of rendering foliage in full ray tracing pipelines. It reduces the reliance on alpha testing, thereby improving performance and efficiency during rendering. It works by precomputing and storing opacity information at a micro-level, allowing the ray tracer to skip unnecessary intersections with fully opaque or fully transparent areas.
Spatial Hashing Radiance Caching (SHaRC). This is a technique aimed at optimizing global illumination in real-time rendering. Although I need to revisit the details, it generally involves using spatial hashing to efficiently store and retrieve radiance information across a scene, reducing the computational cost of lighting calculations in dynamic environments.
Shader Execution Reorder (SER). This is a technique designed to reduce shading divergence in ray tracing and other complex rendering workloads. It works by analyzing and reordering shading workloads - often referred to as a sort-first approach - so that threads with similar execution paths are grouped together. This reduces the performance loss caused by thread divergence on the GPU, leading to more coherent and efficient execution.
The above are the techniques that were mentioned in the lectures as a key feature, or mentioned by both lectures. Of course, it’s impossible - and unnecessary - to cover every single feature discussed. Each company adopts solutions that are specifically tailored to their own projects, and the situation can vary drastically for other projects with different constraints and expectations.
However, one common theme remains clear: the industry is still grappling with the challenge of making real-time ray tracing truly accessible to players. Ray tracing is fundamentally expensive, and performance remains a major concern even for AAA titles. With the aid of technologies like DLSS 4.0 and the topics above, we may finally begin to see a meaningful leap forward.
The industry has devoted significant effort to improving visual fidelity - enhancing water caustics, global illumination, shadows, and occlusion. Yet, I found myself facing an uncomfortable realization: these improvements are often only noticeable in before-and-after comparisons. Presenters repeatedly went back and forth between two slides to highlight visual differences, and while the enhancements were certainly there, I couldn’t help but wonder - will players actually notice the difference during gameplay?
Some aggressive solutions are even pushing ray tracing onto mobile platforms. In the session from Samsung and Google (0319 Google: Unlock The Full Potential of Your Game With Vulkan and ADPF) they showcased advanced features that enable ray tracing on Android devices, with titles like Diablo Immortal being highlighted as examples. These techniques were truly remarkable, bringing significantly enhanced visual fidelity to mobile gaming - more convincing reflections, improved soft shadows, and overall better realism in virtual worlds.
However, these advancements don’t address one of the most fundamental limitations of mobile platforms: bandwidth. Even with advanced lighting techniques, models with aggressive LOD settings or inherently low vertex counts simply don’t look good. No matter how impressive the global illumination may be, the visual quality is still constrained by geometry and memory limitations.
This is a slide page from the presentation. I deliberately chose to put what I saw from the camera here, instead of a screenshot of the page - by looking at this page, you certainly can find some visual differences, such as the reflections on the ice and water. However, you might really want to ask in your heart: does it look much better?
With all that being said, I have never doubted the incredible effort and necessity behind the development of these techniques. The ingenuity behind each feature is truly remarkable, jaw-dropping, even. They reminded me of what first drew me into the world of computer graphics. The elegance with which these challenges are tackled is nothing short of brilliant - these graphics algorithms beautifully embody the art of engineering.
Many of the concerns I’ve expressed stem from the nature of Physically Based Rendering itself. Its ultimate goal is to replicate reality within a game world - to build an engine that renders the world as it truly works. But PBR, by its very nature, is difficult to advance without breakthroughs in physics and simulation.
And yet, I find myself deeply in love with PBR. As I reflect on my career path, I realize that I need to seriously consider whether I am ready to commit to this journey - where making even small improvements often requires enormous effort, and progress can be slow for a long time. But perhaps that’s also where the beauty lies.
Graphics profiling and optimization
Another topic that particularly drew my interest was profiling and optimization in Unity. I attended all of the Unity-sponsored talks, which provided a wealth of insights and practical tips on performance tuning.
Across the sessions, several key strategies were repeatedly emphasized:
Become familiar with Unity’s internal profiling tools. These include the Unity Frame Debugger, Profiler, and a range of new features introduced in Unity 6, such as the Render Graph, Overdraw Viewer, and Graphics State Collections. Mastering these tools is essential for understanding performance bottlenecks directly within the Unity ecosystem.
Use external profiling tools effectively. Tools like Xcode Frame Capture, NVIDIA Nsight, and RenderDoc provide detailed insights into frame-level performance. They make rendering costs visible and straightforward, allowing developers to pinpoint exactly what’s happening during each frame.
Understand the root causes of performance issues. Whether it’s shader complexity, compilation time, memory usage, or asset loading inefficiencies, identifying the underlying reasons is critical. Optimization should be guided by a clear understanding of these contributing factors.
However, most of these tips and strategies can already be found in Unity’s official documentation or in various tutorials available online. Optimization and profiling are ultimately a form of craftsmanship - while understanding the theory is important, nearly anyone can follow optimization steps once they grasp the basics. The real challenge lies in developing the intuition to identify bottlenecks and the sensitivity to spot outliers in performance data.
No matter how detailed the sessions may be, becoming truly proficient in profiling and optimization requires extensive hands-on experience. It’s only through working on real projects, encountering real constraints, and repeatedly solving real problems that one can develop a deep expertise in this area.
Balancing Technical with Aesthetic
Of course, several other graphics sessions focused more on design decisions around interactions and advanced techniques empowered by AI. One particularly memorable talk was from PlayStation Studio Team ASOBI (0321 Flex and Fun: Graphics Magic in ASTRO BOT) - it emphasized that great graphics don’t rely solely on cutting-edge techniques, but on a harmonious integration of all elements that contribute to the gameplay experience.
Another example is a session that talks about how to make game more accessible as a technical artist (0318 Technical Artist Summit: How To Make Game More Accessible). This is truly a remarkable example of showing that the design decisions help shape a better game by using techniques that realize this goal. Technical artist can definitely use their shaders, art assets and special VFX to help the disabled understand the game better.
I had the opportunity to briefly speak with Mr. Tai Yamaguchi, the Principal Graphics Engineer of Team ASOBI. We discussed the team’s decision not to use FFT-based water rendering. Instead, they opted for a Gaussian approach, which offers better interactivity. In the end, it wasn’t a technical limitation - it was a deliberate design choice.
As Mr. Yamaguchi stated during the talk:
“Game graphics are important to immersive experience.”
True. Graphics is a service, not just a show-off. This perspective really stuck with me. It’s a reminder that technical brilliance should ultimately serve the player’s experience, not just exist for visual spectacle.
2.3 What Have I Learned Other Than Graphics
I also attended several sessions on AI in Games and Game Design.
AI, particularly models powered by large language models (LLMs), has become one of the most important trends in game development. Developers are actively exploring ways to integrate LLMs into gameplay experiences. Several talks covered fascinating topics such as how to effectively prompt LLMs, as well as strategies to regulate player behavior, especially in scenarios involving jailbreaking attempts and other edge-case interactions.
One standout example was the session 0317 Games AI: Storytelling in 1001 by Uchan Sun. It was a well-structured, clear, and engaging tech-sharing session. She humorously referred to herself as a Prompt Engineer, highlighting the emerging role of prompt design as both a technical and creative discipline when using AI in game mechanics and narrative design.
Design a structured JSON output. When interacting with LLMs, it’s best to communicate through structured JSON content. The prompt should specify a desired data structure, asking the language model to return specific field, such as booleans, strings, or any game-related values relevant to the current context. This makes it easier to parse and integrate the AI’s response into gameplay systems.
Apply few-shot strategy. Rather than simply describing what you want (i.e. one-shot), provide concrete examples within the prompt. This “few-shot” approach helps guide the LLM’s behavior by showing it how to respond. Giving examples is often more effective than explanation alone.
I am currently co-oping at AlterStaff. Inc, which is another studio (and one of the first several studios) that incorporates AI into its game. These strategies from the GDC talks are extremely helpful.
3 Soft Skills
Conferences like GDC are invaluable spaces for technical exchange - we come together to share ideas, learn from one another, and help drive the industry forward. But beyond the technical content, what truly enables this exchange is something less tangible: soft skills.
The ability to communicate complex ideas clearly, to collaborate across disciplines, and to effectively share knowledge is an essential, though often overlooked, part of being a good developer. Throughout the GDC sessions, I came to deeply appreciate just how vital these soft skills are to our growth—not only as individuals, but as a community.
3.1 Slides!
Each session is, at its core, a presentation - a showcase of the speaker’s knowledge and the ideas they want to share. In this context, the visuals of the presentation - namely, the slide deck - play a critical role as a visual communicator.
One of my biggest takeaways from GDC is this: regardless of your major or industry, everyone should learn how to create a good slide deck. Slide design isn’t just aesthetic; it’s functional. A poorly designed slide can completely ruin an otherwise great talk, while a well-designed slide can elevate even modest content by helping structure the message clearly and memorably.
I won’t name any specific bad examples here, but I must say: some GDC slide decks were as bad as toilet paper. A common issue was excessive, unnecessary text crammed into slides that were only shown for about 10 seconds. It becomes nearly impossible for the audience to follow the speaker - no one can process dense, technically complex information while simultaneously listening and reading at such a pace.
Some might argue, “Well, GDC sessions are recorded - just watch the video later.” But I don’t buy that. Not everyone has access to GDC Vault—and even though I do, why should I spend extra time trying to digest poorly delivered content? More importantly, a badly presented session is a disservice to the live audience who took the time to be there, expecting clarity and engagement.
Meanwhile, I want to express my appreciation for the sessions that truly got it right. The slide visuals from nearly every Adobe and Unity talk stood out as excellent examples. One that particularly impressed me was the 0318 Technical Artist Summit: How to Make Games More Accessible. It made effective use of minimal text, allowing the speaker to take the lead in delivering the message. It was a genuine pleasure to hear the speaker describe the images featured on the slides, building her narrative around the two bold, central keywords on the page. This approach was clear, engaging, and easy to follow. As an audience, we’re there to hear the speaker’s insights—not to read through a wall of text as if we’re listening to an audiobook.
3.2 How To Tell Your Story
I can’t help but continue praising the session 0318 Technical Artist Summit: How to Make Games More Accessible - this is, in fact, the third time I’ve mentioned it in this report! But for good reason: this talk taught me not only what story to tell, but more importantly, how to tell it.
The presenter, Camille Kay from Epic Games, could have simply listed features and technical implementations one by one. Instead, she chose to tell stories - stories about how colorblind players experience the world, and how hearing impairments can lead to players missing crucial content in a game. Her talk made it clear: we want our audience to feel the emotional impact that drives our design choices. These reasons are often deeply human - sentimental and meaningful, not merely technical. The techniques behind accessibility features - like custom shaders or adaptive UI menus - may not be technically complex. But the intention and empathy behind them are what set her apart from other technical artists. That’s what made her message so powerful.
This talk was a powerful reminder of how we should sell our ideas. For an audience — especially a non-technical one — it’s rarely the low-level code or hardcore math/physics that resonates. What people truly want to hear is why a team chose a specific approach, and how it will make a real impact for players, studios, or even the broader industry. Framing a topic with that emotional and practical context helps the audience understand why it matters. Without that, we may find ourselves quietly questioning: Are we really building — or listening to — the right things?
4 Social
I made an effort to participate in several social activities during GDC — handing out my business card, striking up conversations with people in queues, and talking with attendees sitting next to me during technical lectures. I can’t say for sure what these interactions might lead to — perhaps future opportunities, perhaps not — but I believe taking the initiative to connect with others is already a meaningful step.
A particularly memorable moment happened during the session 0321 Flex and Fun: Graphics Magic in ASTRO BOT, presented by Tai Yamaguchi, whom I mentioned earlier. He discussed water simulation and physically-based particle systems — topics that closely overlapped with a project I had worked on the previous semester for my elective course, 15-673 Visual Computing Systems. Not only did I immediately recognize the techniques he described, but I also deeply empathized with the technical constraints he mentioned: targeting 60 FPS on the PS5 (where anything below is considered a bug), supporting interactivity with water, and preserving a unique visual style. His talk felt incredibly familiar, almost as if he were describing the same development concerns I had recently faced.
After the session, my friend Messi Tu encouraged (or rather, pushed) me to introduce myself to Mr. Yamaguchi. I was hesitant at first — I wasn’t sure what I could offer in a conversation, and I didn’t want to interrupt him after his pressured talk. But I did have two advantages: I speak Japanese (not fluently, but well enough to clearly express myself), and I’m a new graduate with relevant experience in 3D graphics — particularly in areas he had just discussed.
Thanks to Messi’s encouragement, I finally approached Mr. Yamaguchi, and to my surprise, he showed genuine interest in me and the project I had worked on. We spoke in Japanese (and to my own surprise, I was able to introduce both myself and my work fluently!), and he asked for my portfolio website to learn more about my project. Our follow-up conversation later confirmed that we had established a strong and meaningful connection.
I have to say that staying connected with people in the industry has been incredibly valuable to me. It’s one of the most effective ways to get myself noticed and involved in the professional community. Those already working within the industry are often best positioned to evaluate the relevance and impact of what I’m doing.
Moving forward, I will continue to actively build and maintain connections with professionals in the field.
5 Takeaway
All in all, what I gained from GDC goes far beyond technical knowledge. While this is a concluding section, I still want to take the time to revisit some of the most meaningful moments I experienced at the conference.
At the ETC alumni gathering, I had a long and heartfelt conversation with my good friend Jiayu He, a Technical Artist at Meta and an ETC Class of 2024 graduate. Although the current job market is bleak, and it has been difficult for me as a new graduate to secure a full-time position, Jiayu encouraged me not to give up. I was feeling especially discouraged after several rejections—or worse, having ongoing interviews suddenly canceled due to the position being filled. But Jiayu asked me a powerful, soul-searching question: “Suppose you land your dream job right now. What would your next five years look like?”
I replied, perhaps I would throw myself into the work if I liked it, or do the bare minimum if I didn’t, saving my time and energy for things I truly care about. He disagreed. He told me that I should never settle for a job I don’t believe in. Even in a tough market, I should remember what work really means to me. I should walk into any interview with confidence and say, “I want to join you — not just because of the paycheck — but because what I want to create can be realized here, and you can help me become the person I aspire to be.”
As someone who calls himself a Technical Artist, I had overlooked the “artist” side of that title. Jiayu’s words reminded me that my skills should not just help me earn a salary, but help me express who I am. To walk the path of an artist means leading with identity and intentionality.
At GDC, we saw through countless successful case studies the power of self-awareness—knowing what you’re doing and why. Whether it’s in technology or design, every speaker who left a strong impression showed us what it means to find your own path and grow through the trials along the way. On the other hand, the less compelling talks often reflected a lack of that same self-recognition, sometimes only realized too late. As an engineer, perhaps I only need textbooks, online tutorials, and technical practice to get by. But as an artist, I still have a long way to go. In response to this reflection, I even removed the title Technical Artist from my LinkedIn headline and temporarily labeled myself only as a Graphics Engineer. I believe I am qualified as an engineer — curious, analytical, persistent. But to be an artist? That journey has just begun. Still, I’m grateful to have realized this now. In that moment, it felt like I had refilled my purpose, and suddenly the road ahead became visible again.
On a long drive from Union City to Palo Alto, I shared with my friend Messi Tu a theory I once heard in CMU’s course 15-251. Humanity has existed for only a few thousand years, and technological progress in the last century alone has been staggering. With the pace of advancement, especially in fields like VR/AR, it is not unimaginable that one day we may be able to create a virtual world indistinguishable from reality. If that’s true, how can we even be sure the world we’re in now is real? And if such worlds can be created, then in time, perhaps infinite virtual worlds could be made. Statistically speaking, the chance that this is the real world would be close to zero.
This isn’t a call to nihilism. It’s a reminder that reality might be more tolerant of experimentation than we think. We may never be able to prove the permanence of anything we build, but we should still bravely try what we’ve never done before. Ironically, it’s a philosophy we may best understand through games.
GDC taught me something profound: that building a convincing virtual world is a collective endeavor. Experts from every discipline contribute in their own way to solving impossibly complex problems. Even if I — and the industry as a whole — remain unsatisfied with the current state of ray tracing, it is through the relentless efforts of these individuals that we inch closer to a world with infinite tolerance for iteration and creativity.
For me personally, I’m deeply grateful to have found a meaningful connection with this industry, and I’m eager to keep growing within it. The knowledge and inspiration I’ve gained from GDC will continue to shape how I see myself—not just as an engineer, but as an aspiring artist. And when the day comes that I can confidently call myself a Technical Artist again, I hope to bring real value back to this field. That, to me, is what GDC was truly about.
Comentarios