惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

腾讯CDC
The Cloudflare Blog
IT之家
IT之家
V
V2EX
雷峰网
雷峰网
MyScale Blog
MyScale Blog
P
Proofpoint News Feed
Stack Overflow Blog
Stack Overflow Blog
博客园 - Franky
Engineering at Meta
Engineering at Meta
S
SegmentFault 最新的问题
GbyAI
GbyAI
Microsoft Azure Blog
Microsoft Azure Blog
博客园 - 司徒正美
云风的 BLOG
云风的 BLOG
小众软件
小众软件
博客园 - 叶小钗
Blog — PlanetScale
Blog — PlanetScale
C
Check Point Blog
A
About on SuperTechFans
B
Blog
月光博客
月光博客
宝玉的分享
宝玉的分享
Last Week in AI
Last Week in AI

Latest from Tom's Hardware

Our experts review your astonishing PC builds and setups in Rig Rundown — from wall-mounted setups to a system… News outlets are blocking Wayback Machine from archiving their pages — 23 outlets concerned AI companies might abuse fair use and use it to train their models Mark Zuckerberg reportedly working on AI clone of himself — Meta insiders claim 3D photoreal animated Zuck will be able to engage with employees on his behalf Score a massive $700 off this 4K-ready Lenovo gaming PC with an RTX 5070 Ti, now just $1,899 — epic Legion Tower 5i pre-built ships with a 20-core Intel CPU, 32GB DDR5 and a 2TB SSD Pay $1,349.99 for Gigabyte's Aero X16 laptop and save $300 on this 32GB beast with RTX 5070 graphics —… Veteran Windows dev shows off AI running on 47-year-old PDP11 with 6 MHz CPU and 64KB of RAM — 'gloriously absurd' project runs transformer model written in PDP-11 assembly language Half of all US employees now use artificial intelligence at work, crossing landmark threshold for first time — Gallup data shows daily and weekly usage hitting all-time high of 28% in Q1 2026, with 65% feeling positive about its impact on productivity China has spent 3.6 times more than the US on chipmaking subsidies over the past decade — $142 billion and counting, easily outweighs CHIPS Act FAA approves military use of drone-killing laser weapons in US airspace — decision comes after it was decided ‘systems do not present an increased risk to the flying public’ Nvidia says AI cuts 10-month, eight-engineer GPU design task to overnight job — company is still 'a long way' from AI designing chips without human input Small Missouri town ousts half its city council after $6 billion AI data center approval — petition calls for mayor's removal as frustration (and violence) over AI data centers mounts New tech can see a CPU's transistors in action — terahertz radiation can potentially steal data as a chip is… Intel's Nova Lake CPUs gear up to seize AMD’s 3D V-Cache gaming throne — early leak points to up to 52 cores, blazing DDR5-8000 support, and massive 175W TDP Acer Predator GX850 SFX power supply review: Solid electrical performance with good efficiency NZXT to cough up $3.45 million over 'predatory' Flex PC rental scheme in RICO class-action settlement — in-debt customers to get up to $5,000 of relief, eligible renters to be granted ownership Bulbous 15x fan PC case side panel dubbed the ‘Superdome’ lowers temps by 20 degrees —  $600 worth of Noctua fans arrayed in 3D-printed structure Approvals for Nvidia and AMD AI chip exports to China stall under government bottleneck —  20% staff turnover… Espresso Lite 15 Review: An entry-level portable monitor with a splash of color Save a massive $700 on this 4K-ready HP gaming PC with a 9800X3D and RTX 5070 Ti, now just $2,499 — discounted HP Omen 35L pre-built powerhouse ships with 32GB DDR5 RAM and a 1TB SSD 'CopprLink' destroys every eGPU standard in new test, achieves near-native-level performance with an RTX 5090 — setup requires $2,300 worth of additional hardware Website backup crippled by 1.6MB Friends GIF that was replicated 246,173 times, breaking Linux's EXT4 filesystem limit — Jennifer Aniston's 'happy dance' animation ate up 377 gigabytes of data due to security policy Why we spent 50+ hours retesting Intel’s Core Ultra 270K Plus and 250K Plus Just $284.99 for 32GB of Team T-Create Classic DDR5-6000 RAM is the cheapest going right now — this double-dipping… Grab MSI’s RTX 5080 gaming laptop for just over $2,000 — offers fast 240 Hz QHD+ display, dual storage slots, and expandable DDR5 memory Lenovo hikes Legion Go 2 handheld gaming PC to almost $3,000 for 2 TB model — Handheld now costs more than AMD's Strix Halo devices despite relatively weaker Z2 Extreme chip Iran's forced nationwide internet blackout becomes second-longest on record as it passes 1,000 hours offline — possessing Starlink terminals punishable by death, country using 'military-grade jamming' against service Tiny 3-inch cube PCs bring a splash of color to the passive PC market with red, orange, green and blue options — Intel Twin Lake-powered Kubb Mini PCs start at $500 Veteran Microsoft engineer says original Task Manager was only 80KB so it could run smoothly on 90s computers — original utility used a smart technique to determine whether it was the only running instance Tech enthusiast gets Doom to run on a 40-year-old printer controller — ancient Agfa Compugraphic 9000PS came with a Motorola 68020 onboard for fast processing Keychron Q6 Ultra 8K Review: 660 hours of battery life at 8 KHz
Testing Nvidia's RTX Mega Geometry tech — VRAM-redu...
Dan Mateescu · 2026-05-09 · via Latest from Tom's Hardware
Testing RTX Mega Geometry
(Image credit: Nvidia)

We took Nvidia's RTX Mega Geometry technology through a series of tests in Alan Wake 2 and the RTX Bonsai Diorama Demo to see how this tech reduces VRAM consumption and eliminates visual artifacts, thus helping pave the way to photorealistic real-time graphics.

In 2018, NVIDIA announced its GeForce RTX line of graphics cards based on the Turing architecture, which would allow for hardware-accelerated real-time ray tracing. In November of that year, Battlefield V became the first title to support real-time ray tracing using the Microsoft DirectX Raytracing API (DXR). The game only supported one ray-traced effect – ray-traced reflections.

In 2019, Control launched with support for multiple ray-traced effects – RT reflections, RT transparent reflections, indirect diffuse lighting, RT contact shadows, and RT debris.

Later, we would see full ray tracing – or path tracing – in games like Quake II RTX, Cyberpunk 2077, and more. In contrast to the hybrid ray tracing solution used in some games, path tracing accurately simulates light in a scene. It does this by sampling a wide range of potential light paths a single ray can follow. This method is also used by the film industry to generate photorealistic visuals in movies.

Path tracing is present in a number of modern games, where it can significantly enhance lighting realism depending on the implementation. In addition to this enhanced lighting, there have also been technological advancements in geometric complexity and density for games. Unreal Engine 5’s Nanite is a virtualized geometry system that can render high-quality and high-density assets in real time with pixel-scale detail and high object counts. Other game engines have similar technologies, such as Micropolygon Geometry in the Anvil Engine.

These advancements in geometric detail require new techniques to ray trace full-quality geometry. That’s where Nvidia's RTX Mega Geometry technology comes in.

What is RTX Mega Geometry?

The Microsoft DXR API was designed years ago, before the level of geometric complexity that can be found in modern games. In DXR, geometry is represented using a Bounding Volume Hierarchy (BVH), which can become expensive to rebuild in dynamic scenes with dense geometry and countless complex animated objects. Fully ray tracing such geometry may require frequent BVH updates, placing significant strains on performance. As a result, developers often rely on lower-quality proxy meshes for ray-traced effects. These simplified representations contain only a fraction of the original detail, leading to artifacts such as incorrect self-occlusion and low-fidelity shadows and reflections.

Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.

RTX Mega Geometry is a rendering technology that can significantly increase geometric detail in ray-traced games, allowing full-fidelity geometry to be traced without traditional trade-offs. RTX Mega Geometry adds an optional Cluster Acceleration Structure (CLAS), which generates batches of up to 256 triangles and is designed to be GPU-driven. This not only dramatically speeds up the process of rebuilding the BVH, but it also reduces the CPU overhead associated with BVH management.

Testing RTX Mega Geometry

(Image credit: Nvidia)

While Mega Geometry is supported on GPUs as far back as the RTX 20-series, the RTX 50-series has purpose-built hardware for the technology. The fourth-generation RT Cores in Blackwell GPUs add two new engines: a triangle cluster intersection engine and a triangle cluster compression engine. Together, they double the ray-triangle intersection rate compared to third-generation RT Cores, while also reducing VRAM usage by several hundred megabytes when ray tracing dense geometry.

The benefits of this technology vary depending on how it’s used. When applied to existing assets, it can improve performance and reduce VRAM consumption. Alternatively, it can eliminate the need for the compromises mentioned earlier – such as proxy meshes – enabling significantly higher-fidelity results without the associated visual artifacts. The latter use case comes at the cost of performance.

RTX Mega Geometry in Alan Wake 2

Testing RTX Mega Geometry

(Image credit: Nvidia)

Alan Wake 2, released in October 2023, showcases a range of modern rendering technologies. It uses mesh shading to handle highly detailed, high-density geometry and supports path tracing on PC, making it an ideal candidate for Mega Geometry.

Mega Geometry was added to the game in title update 1.2.8 in early 2025, and Remedy opted to use it on existing assets to optimize performance and VRAM usage, as opposed to further increasing geometric complexity. Testing on an RTX 4090 back in early 2025, I compared build 1.2.8 to the previous 1.2.7 release, which did not use Mega Geometry, and observed VRAM savings of about 1 GB along with a 13% performance increase at both native 4K and 4K with DLSS Quality, with max settings and path tracing enabled. These numbers are roughly in line with the figures shared by NVIDIA at GDC 2026.

RTX Bonsai Diorama Demo

The RTX Bonsai Diorama Demo released by NVIDIA was developed in the NVIDIA RTX Branch of Unreal Engine (NvRTX) 5.6, and showcases RTX Mega Geometry, path tracing, RTX Dynamic Illumination (RTXDI), DLSS Super Resolution, DLSS Ray Reconstruction, and DLSS Frame Generation. The main focus of the demo is the fully path-traced Nanite meshes in real time.

The great thing about the demo is that you can turn Mega Geometry on and off to compare image quality and performance. So that’s exactly what we will do.

The debug visualization in the screenshots below illustrates the precision of Mega Geometry when path tracing Nanite meshes and the substantial increase in triangle count when it is enabled.

Image

1

of

2

Testing RTX Mega Geometry
(Image credit: Tom's Hardware)

Because Mega Geometry adds full quality Nanite geometry to the BVH, ray-traced shadows no longer exhibit the visual artifacts that are present in Lumen, such as missing or incorrect shadows. In the comparison below, you can see that every object in view has accurate, pixel-perfect shadows with Mega Geometry enabled.

Image

1

of

2

Testing RTX Mega Geometry
(Image credit: Tom's Hardware)

Similarly, reflections are no longer plagued by the issues that result from using low-quality proxy meshes. In the comparison below, you can see missing leaves from the reflection within the center mirror when Mega Geometry is disabled. With Mega Geometry enabled, you get the full Nanite mesh.

Additionally, pay attention to the tree on the far left. With Mega Geometry off, you get incorrect self-occlusion. This is entirely resolved with Mega Geometry.

Image

1

of

2

Testing RTX Mega Geometry
(Image credit: Tom's Hardware)

The technology clearly improves image quality, but how does it perform?

It is important to note that the level of geometric detail and density observed above with Mega Geometry enabled was simply not feasible in Unreal Engine 5 prior to Mega Geometry, even on high end GPUs like the 5090. This level of fidelity could not be attained at an acceptable level of performance. Having said, while Mega Geometry makes this possible at playable framerates on high-end GPUs, path tracing such high-fidelity meshes is still extremely demanding.

Test system

Testing RTX Mega Geometry

(Image credit: Tom's Hardware)

On the 5090, the performance cost of enabling RTX Mega Geometry is 23% at 1080p, 24% at 1440p, and 21% at 4K. At a rendering resolution of 1440p or below, performance is above 60 FPS. This gives you the option of using frame generation to further increase motion clarity without significantly impacting perceived latency. Typically, you want to use frame generation with a base framerate of at least 60 FPS – ideally higher – so that the game still feels responsive.

Testing RTX Mega Geometry

(Image credit: Tom's Hardware)

On the RTX 5070, the performance cost is 27% at 1080p, and 26% at 1440p. Unfortunately, 60 FPS with Mega Geometry is not possible in this demo on the 5070 unless you use frame generation or DLSS with an internal rendering resolution below 1080p. This result is not entirely surprising, as standard path tracing is already extremely taxing on this GPU.

Testing RTX Mega Geometry

(Image credit: Tom's Hardware)

We included the 5060 here because the RTX Bonsai Diorama Demo developer guide listed this as the “recommended GPU.” Clearly, that must include the use of DLSS Super Resolution and Frame Generation, because without them, the 5060 averages below 30 FPS in this demo with Mega Geometry enabled. Indeed, with DLSS Super Resolution and Frame Generation at 1080p, the 5060 can exceed 100 FPS, but the resulting experience suffers in both image quality and input latency.

The Future of RTX Mega Geometry

At GDC 2026, NVIDIA announced that Mega Geometry would be implemented in the upcoming 2026 titles Control Resonant and The Witcher 4.

Testing RTX Mega Geometry

(Image credit: Nvidia)

During their Future of Path Tracing presentation at GDC 2026, NVIDIA highlighted the RTX Mega Geometry foliage system coming to The Witcher 4.

The new level-of-detail (LOD) system for foliage selectively updates only the relevant parts of the scene based on camera movement, and represents level-of-detail in a manner that is efficient to ray trace. This is done without exhibiting traditional LOD pop-in.

The system leverages Opacity Micromaps (OMMs) for distant LODs, which are used as a simplified representation of the geometry that is lightweight in memory. On the RTX 40-series and 50-series, OMMs are hardware-accelerated.

Image

1

of

4

Testing RTX Mega Geometry
(Image credit: Nvidia)

Later in the presentation, NVIDIA showcased the system in a demo featuring a vast 5x5 km forest composed of roughly 60 million plants spanning over 200 species, including around 1 million trees. Notably, the entire scene runs without streaming, with all assets resident in memory. Every element is represented as full geometry – down to individual pine needles – with some of the largest trees reaching approximately 10 million polygons each. The scene is rendered with fully dynamic, path-traced lighting, and the tree assets were provided by CD PROJEKT RED.

Testing RTX Mega Geometry

(Image credit: Nvidia)

What was surprising was the level of performance achieved in this demo. On the RTX 5090 at 4K using DLSS Quality mode (internal rendering resolution of 1440p), the demo ran at 80 FPS. This is nearly 18% faster than the Bonsai Diorama demo ran on the 5090 at 1440p.

On the RTX 4070 at 1440p using DLSS Quality mode (internal rendering resolution of 960p), it ran at about 58 FPS.

RTX Mega Geometry represents a meaningful step toward truly photorealistic real-time graphics. With advancements like the new LOD system for foliage and ongoing performance optimizations, it has the potential to enable scenes of unprecedented geometric complexity that can be rendered with fully dynamic, path-traced lighting at playable framerates on modern hardware.

Dan Mateescu is a PC enthusiast with many years of experience benchmarking PC hardware. In 2021, he started his own YouTube channel called 'Compusemble' where he benchmarks hardware in video games and the latest tech demos.