AMD Launches DGFS SuperCompression for 22% Smaller Games

AMD has unveiled a new compression technology called DGF SuperCompression (DGFS) that can shrink game geometry data by up to 22%, potentially saving gigabytes of storage on increasingly bloated game installs. The technology, introduced as part of the DGF SDK 1.2 update on May 7, works as a software-layer compression system that a single CPU core can rapidly decode in real time.

The move addresses one of gaming's most persistent consumer complaints: ballooning install sizes. With modern AAA titles regularly exceeding 100 GB, any meaningful reduction in asset storage is welcome news for both players and developers.

Key Takeaways

• 22% file size reduction on top of existing DGF compression for geometry data
• Decodable on a single CPU core, minimizing performance overhead
• Maintains full compatibility with non-DGF hardware through unified resource formats
• Released as part of the DGF SDK 1.2 update, freely available to developers
• Uses cluster-granularity compression for efficient, targeted decompression
• Enables a single asset pipeline for both DGF-capable and legacy GPUs

What Is DGF and Why Does It Need SuperCompression?

Dense Geometry Format (DGF) is AMD's hardware-oriented geometry compression format designed to make 3D mesh data as efficient as possible for real-time rendering. At its core, DGF ensures that every triangle's required information can be accessed through a single 128-byte aligned memory read — a design choice that dramatically speeds up GPU processing.

However, this hardware-friendly alignment comes at a storage cost. Vertex positions and compression parameters must be duplicated across multiple data blocks to maintain that clean 128-byte access pattern. Padding bits are also inserted to preserve data alignment, further inflating file sizes.

In other words, DGF is excellent for runtime performance but wasteful for on-disk storage. That is precisely the gap DGFS is designed to fill.

How DGFS Achieves Its 22% Compression Gains

DGFS operates as a software-layer compression system that sits on top of the existing DGF format. Rather than replacing DGF, it wraps DGF data in an additional compression pass that eliminates the redundancies inherent in the storage format.

The technology works by:

• Removing duplicate vertex positions that are repeated across adjacent DGF blocks
• Stripping padding bits that are necessary for hardware alignment but redundant in storage
• Consolidating compression parameters that would otherwise be stored multiple times
• Using cluster-granularity encoding to allow selective decompression of only the geometry clusters needed at any given moment

The result is a compressed data stream that cannot be directly consumed by GPU hardware — it must first be decoded back into proper DGF blocks. AMD emphasizes that this decoding process is lightweight enough to run on a single CPU core, meaning it adds negligible overhead to game loading times.

Importantly, DGFS-encoded data can also be decoded into traditional vertex and index buffers, the standard geometry format used by all modern GPUs. This dual-path decoding is what enables cross-hardware compatibility.

Unified Asset Pipeline: One Format for All Hardware

Perhaps the most strategically significant aspect of DGFS is its ability to serve as a universal geometry storage format. Game developers currently face a fragmentation challenge: DGF-capable hardware (primarily AMD's latest GPUs) can consume DGF data directly for maximum performance, while older or competing hardware requires traditional vertex and index buffers.

Without DGFS, developers would need to ship two separate sets of geometry assets — one in DGF format and one in traditional format — or choose one and sacrifice either performance or compatibility. DGFS eliminates this dilemma by providing a single compressed format that can be decoded into either representation at load time.

This has practical implications for game distribution:

• Smaller download sizes for all players, regardless of hardware
• Reduced development complexity with a single asset pipeline
• Lower CDN and distribution costs for publishers
• Faster patching since geometry assets take up less space in updates
• Broader hardware support without storage penalties

For studios already working with DGF, adopting DGFS is described as a straightforward SDK integration rather than a fundamental pipeline overhaul.

The Bigger Picture: Gaming's Storage Crisis

AMD's DGFS arrives at a time when the gaming industry is grappling with an escalating storage problem. Titles like Call of Duty: Modern Warfare III have pushed past 200 GB, and even mid-tier games routinely demand 50-80 GB of disk space. The shift to high-fidelity assets — driven by technologies like Nanite in Unreal Engine 5 and the broader push toward film-quality geometry — is only accelerating this trend.

Geometry data, while not the largest contributor to game file sizes (textures typically hold that distinction), represents a meaningful and growing share. As developers adopt denser meshes with millions of micro-polygons per scene, geometry compression becomes increasingly critical.

AMD is not alone in this space. Nvidia has invested in its own mesh compression technologies, and Epic Games' Nanite system includes built-in geometry streaming and compression. However, AMD's approach with DGF and DGFS is notable for being an open SDK rather than a proprietary engine feature, potentially enabling broader industry adoption.

Microsoft's DirectStorage API, which accelerates asset decompression from NVMe SSDs to GPU memory, represents a complementary technology. DGFS-compressed geometry could theoretically benefit from DirectStorage's fast I/O pipeline, though AMD has not yet confirmed specific integration details.

Technical Deep Dive: Cluster-Granularity Compression

One of DGFS's key architectural decisions is its use of cluster-granularity compression. Rather than compressing entire meshes or scenes as monolithic blocks, DGFS operates on individual geometry clusters — small, self-contained groups of triangles that can be independently compressed and decompressed.

This design choice offers several technical advantages. First, it enables streaming-friendly decompression: a game engine can decompress only the clusters visible in the current frame, reducing memory bandwidth requirements. Second, it supports parallel decoding across multiple CPU cores when needed, though AMD notes that even a single core provides sufficient throughput.

Cluster-granularity also aligns naturally with modern rendering architectures. Both AMD's and Nvidia's latest GPUs process geometry in cluster-sized batches through mesh shaders, meaning the decompression granularity matches the consumption granularity. This architectural alignment minimizes buffering overhead and memory waste.

The 22% compression figure AMD cites represents the reduction compared to storing raw DGF blocks. When measured against uncompressed traditional geometry formats (vertex + index buffers), the total compression ratio is substantially higher, since DGF itself already provides significant compression over traditional formats.

What This Means for Developers and Players

For game developers, DGFS offers a low-friction path to smaller builds. Studios already using the DGF SDK can upgrade to version 1.2 and begin encoding their geometry assets with SuperCompression. The unified format approach also simplifies multi-platform shipping, particularly for titles targeting both AMD and Nvidia hardware.

For players, the benefits are straightforward: smaller downloads, faster installs, and more games fitting on limited SSD space. A 22% reduction in geometry data might translate to several gigabytes saved on a large open-world title, though the exact savings depend on how much of a game's total footprint is geometry versus textures, audio, and video.

For the broader industry, DGFS represents another step in the ongoing arms race against storage bloat. Combined with texture compression standards like BCn and ASTC, audio codecs like Opus, and video compression like Bink 2, geometry compression is becoming a first-class concern in the game development toolkit.

Looking Ahead: Adoption and Integration

The success of DGFS will ultimately depend on developer adoption. AMD's decision to release it as an open SDK lowers the barrier significantly, but integration still requires developer effort. Key factors to watch include:

• Whether major game engines like Unreal Engine and Unity integrate DGFS natively
• How DGFS interacts with DirectStorage and other I/O acceleration technologies
• Whether Nvidia releases competing or complementary geometry compression standards
• Adoption timelines from major AAA studios and their shipping schedules
• Performance benchmarks comparing DGFS decode times across different CPU architectures

AMD has been steadily building out its geometry compression ecosystem, with DGF initially launched alongside its RDNA 4 GPU architecture. The addition of DGFS signals that AMD views geometry compression not as a one-time feature but as an evolving platform that will continue to improve.

As game fidelity continues its relentless march upward, technologies like DGFS will become essential infrastructure rather than optional optimizations. AMD's early investment in this space could pay significant dividends as the industry's storage problem intensifies.