AMD is combining its gaming (RDNA) and data center (CDNA) GPU architectures into a unified platform aimed at improving flexibility and efficiency across workloads. Announced in late 2024, this design resembles Nvidia’s approach of supporting both gaming and compute tasks within one ecosystem. Insider reports suggest it could double performance in ray tracing and AI tasks like image upscaling, while offering a 20% boost in traditional rendering. Though initially called UDNA, the gaming version may be marketed as RDNA 5. Console integration and SoC development are expected, but no official rollout timeline has been confirmed.
AMD is creating a single GPU architecture that merges its current gaming and data center technologies into one platform. This move marks a significant change in how the company designs GPUs, combining the RDNA architecture used for gaming with the CDNA architecture built for high-performance computing and data centers. The unified design aims to deliver a more flexible GPU that can efficiently handle a variety of workloads.
The idea behind this unified approach is similar to Nvidia’s CUDA ecosystem, which allows GPUs to handle both gaming and computing tasks within the same system. AMD first announced this architectural plan in late 2024, but the company has not yet shared specific implementation timelines or confirmed which products will feature the new technology.
Technical performance projections
According to industry insider Kepler_L2, who has gained credibility through past predictions, the new architecture promises significant performance enhancements across various areas. The expected improvements in ray tracing and AI functions are especially notable, with forecasts of roughly doubling the performance compared to current-generation hardware. This upgrade would address one of AMD’s long-standing weaknesses in comparison to Nvidia’s products.
The speculation is that improvements in ray-tracing performance focus on the computational efficiency of light simulation, rather than directly increasing frame rates. When the insider mentions doubling ray-tracing performance, this refers to cutting the time needed for ray-tracing calculations in each frame, not doubling the overall frame rate in ray-traced games. This distinction is important for understanding the practical effects on gaming.
The architecture also aims at AI-driven features like image upscaling and frame generation. These technologies use machine learning algorithms to improve visual quality and boost frame rates without needing proportional increases in raw processing power. The expected improvements could make AMD’s products more competitive with Nvidia’s well-established AI acceleration capabilities.
Traditional rasterization performance, which converts 3D models into 2D images for display, is expected to improve by about 20% per compute unit. This improvement would lead to more efficient rendering of standard game graphics, potentially resulting in smoother gameplay and better visual quality across many applications.
Console integration and market impact
The timing of this architectural development aligns with AMD’s reported success in supplying APUs for upcoming PlayStation and Xbox consoles. Incorporating the new architecture into these gaming systems could enable better lighting effects, more detailed visuals, and enhanced AI-driven features that help sustain performance in demanding gaming situations.
The console market offers a significant opportunity for AMD to demonstrate the capabilities of its unified architecture. Gaming consoles typically require GPUs that balance performance, power efficiency, and cost, making them an ideal testing ground for architectural innovations. Success in this market segment could lead to broader adoption across AMD’s product portfolio.
Naming convention uncertainty
Recent information suggests potential confusion regarding the final designation of the architecture. While AMD initially indicated the unified design would be called UDNA, more recent reports suggest the gaming-focused implementation may retain the RDNA naming convention, specifically RDNA 5. This discrepancy suggests that AMD may be revisiting its branding strategy or that different market segments may be receiving distinct naming approaches.
The insider sources suggest that the UDNA designation may not materialize as previously expected, with RDNA 5 becoming the primary identifier for the next-generation gaming architecture. This naming evolution reflects the fluid nature of product development and marketing decisions within the semiconductor industry.
Competitive positioning
AMD’s current RDNA 4 architecture has achieved competitive parity with Nvidia’s offerings in several areas, providing strong traditional rendering performance and competitive memory configurations. However, path tracing and advanced ray-tracing applications are still areas where AMD’s hardware typically lags behind Nvidia’s solutions.
The expected advancements in ray tracing and AI performance could help AMD narrow this competitive gap. While Nvidia has traditionally held advantages in these areas through dedicated hardware acceleration and software optimization, AMD’s unified architecture approach might lay the groundwork for more competitive products.
System-on-Chip development
Beyond discrete graphics cards, the new architecture is expected to influence AMD’s System-on-Chip (SOC) designs for gaming consoles. Interestingly, reports suggest that the console-focused SOC implementations will not incorporate 3D V-Cache technology in their CPU portions. This decision likely reflects cost and complexity considerations specific to console applications, where manufacturers must balance performance with affordability for mass-market adoption.
Development timeline and market implications
AMD has not provided official confirmation regarding implementation timelines or specific product availability. The transition to a unified architecture represents a significant undertaking that affects multiple market segments, from gaming graphics cards to data center accelerators and console processors. As of now, AMD will not be disclosing any new information at Hot Chips in August and will present a paper on AMD RDNA 4 and the Radeon RX 9000-series GPU.
This new architectural consolidation could streamline AMD’s development processes by reducing the number of distinct GPU designs the company needs to maintain. A single, unified architecture could allow for more efficient resource allocation and potentially speed up feature development across various market segments.
The success of this architectural transition will likely shape AMD’s competitive edge in both the gaming and data center sectors. As AI workloads grow more vital across different applications, having a unified architecture that can handle both traditional graphics and AI acceleration could offer strategic benefits in a changing market landscape.
WHATCHA THINK—GOOD ENOUGH TO SHARE WITH YOUR FRIENDS AND RELATIVES?