AMD DGF Tech Offers Massive Increase In Geometry In Ray Traced Games With Future RDNA GPUs, Achieves Up To 30% Compression With Current GPUs Hassan Mujtaba • at EDT Add on Google AMD has further detailed its DGF or Dense Geometry Format technology, which rivals NVIDIA's RTX Mega Geometry, delivering vastly higher geometry counts in ray tracing games while offering new compression methods designed with future RDNA GPUs in mind. As Geometry Scale Increases in Games, Tech Such as AMD's DGF & NVIDIA's RTX Mega Geometry becomes vital in powering advanced ray-traced games Last year, AMD announced its Dense Geometry Format technology, or DGF in short. This tech aims to handle massive polygon counts by streaming geometry clusters rather than full scenes. The principle of AMD's DGF is simple. It is designed to store compressed triangle meshes by taking a standard triangle mesh and storing it into small meshes or meshlets. A single DGF-meshlet consists of 64 vertices and 64 triangles, and is stored within a 128-byte DGF-block, which also includes the meta information. A set of DGF-blocks represents a mesh. Related Story NVIDIA’s AI Chips Reached China’s Alibaba Through Thailand, As US Indicts Supermicro Execs In $2.5 Billion Smuggling Ring Now, AMD is unveiling more details for DGF, stating that it has the potential to deliver a massive increase in geometry detail for applications such as ray-traced rendering in games, content creation, virtual production, and several other real-time 3D workloads. With the emergence of Nanite in Unreal Engine, the bar has been raised for complex models. These technologies also introduce small triangle formats that are rasterized & use software rasterization in compute shaders in some cases, which creates challenges with ray tracing rendering. The use of compression formats to make these triangles viable in ray tracing APIs requires decoding, and this affects memory latency, causing instability, stuttering, and lower performance. Even if hardware acceleration is used, current structures are just too big to support upcoming content. As per AMD, the current "black box" API structure for ray tracing acceleration has some limitations, which include: The allocation for “pre-build” memory must be sized for the worst-case compression rate, increasing the minimal footprint and adding complexity and inefficiency to the build process. The implementation must store enough information to exactly reproduce the input triangle order for index references, which will increase memory consumption. The implementation must convert the input into a hardware format, ultimately impacting performance, silicon area, and/or power. A mandatory runtime transcode discourages implementations from parts of the design space that are denser but harder to encode. The result is an indirect increase in memory consumption. To address this, AMD has worked with Samsung and various other software developers to create a standard and efficient geometry compression format, which is part of DGF, and called DGF "SuperCompression". DGF SuperCompression further compresses the DGF data to reduce storage costs by up to 30%. The technology also works on non-DGF hardware, so while older RDNA GPUs such as RDNA 4 and prior generations support DGF supercompression with up to 30% reduction in storage sizes, future GPUs such as RDNA 5 will offer fuller support, leading to even bigger gains. We address this by introducing DGF SuperCompression (DGFS), a software system which further compresses DGF data to reduce its storage cost. Geometry encoded with DGFS is no longer directly consumable by hardware, but can be made considerably smaller. DGFS is able to exactly reconstruct a given set of input DGF blocks, and also supports efficient decode to a conventional vertex and index buffer, which enables DGFS content to run on non-DGF hardware. AMD AMD has published some raw storage footprints and the resultant savings across a variety of models for both DGF blocks and uncompressed DGFS streams, running on its current-generation Radeon RX 9070 XT (RDNA 4) graphics card: CrabDragonStatuetteBuddhaBikeTriangles(Millions)2.147.2210.001.091.68DGF Size (MB)10.2229.2540.994.946.96DGFS Size (MB)8.4820.1529.313.955.54Savings17.06%31.09%28.48%20.03%20.47% CrabDragonStatuetteBuddhaBikeTriangles(Millions)2.147.2210.001.091.68DGF Size (MB)7.1920.1528.653.354.56DGFS Size (MB)5.7315.6723.312.633.69Savings20.29%22.22%18.61%21.34%19.04% CrabDragonStatuetteBuddhaBikeTriangles(Millions)2.147.2210.001.091.68Meshlet Decode Time (sec)0.030.090.150.020.02DGF Decode Time (sec)0.050.150.220.030.04 Note - Decode time results generated by a system with an AMD Ryzen 9 7950X 16-core processor, 64GB DDR5 6000 RAM, AMD Radeon RX 9070 XT graphics card, MSI PRO PRO X670-P WIFI motherboard, and Microsoft Windows 11 2025 Update. There's no doubt that geometry is going to see a huge leap in next-generation games. Witcher IV is one example that looked absolutely gorgeous even in the demo that was showcased a few months back. In addition to that, NVIDIA already leverages RTX Mega Geometry in Alan Wake 2 and is also coming to Control Resonant. AMD's DGF (Dense Geometry Format) paves the way for its future neural rendering architectures, such as the RDNA 5 lineup, which is expected to land on PCs and next-gen consoles. Though there's no timeline on when we will see AMD's next-gen architectures in action, AMD has been discussing some wonderful features, such as FSR Diamond, and more as part of its Project Amethyst with Sony. FeatureNVIDIA RTX Mega GeometryAMD Dense Geometry Format (DGF)Primary GoalAccelerate RT for ultra-dense, dynamic meshes (Nanite-style).Efficiently pack and render micro-meshes for raster & RT.Core ArchitectureBlackwell / RTX 50-series (4th Gen RT Cores).UDNA / Future Architectures (Post-RDNA 4).MechanismSpecialized triangle cluster intersection & compression engines.128-byte compressed blocks (up to 64 vertices/triangles).BVH ImpactUp to 100x faster BVH updates for dynamic scenes.Focuses on low-overhead scene representation & storage.Rendering ScopePrimarily focused on Ray Tracing/Path Tracing.Broad optimization for both Rasterization & Ray Tracing.VRAM StrategyHeavy hardware-level geometry compression.Lossy block-based compression for bandwidth savings.StatusAvailable/Launching in 2026 hardware.In development for future GPU generations. About the : A Software Engineer by training and a PC enthusiast by passion, Hassan Mujtaba serves as 's for hardware section. With years of experience in the industry, he specializes in deep-dive technical analysis of next-generation CPU and GPU architectures, motherboards, and cooling solutions. His work involves not only breaking news on upcoming technologies but also extensive hands-on reviews and benchmarking. Follow on Google to get more of our news coverage in your feeds. Further Reading NVIDIA Rubin & Rubin Ultra Platforms Facing Design/Spec Issues As Per Rumors While AMD MI500 Positioned For 2H 2027 Launch AMD Finally Overtakes Intel in Q1 Data Center Revenue as Agentic AI Forces Hyperscalers to Hoard CPUs Over GPUs AMD Radeon RX 9070 XT and RX 7650 GRE Crash Below MSRP In China As Inflated Pricing Triggers Demand Collapse AMD Eyes Samsung 2nm Tech As Alternative In Tackling Wafer Supply Constraints With Talks In “Advanced Stages” Read all on AMD DGF Tech Offers Massive Increase In Geometry In Ray Traced Games With Future RDNA GPUs, Achieves Up To 30% Compression With Current GPUs
