AMD UDNA RDNA 5 GPUs may feature 96 compute units and 384-bit memory bus

AMD´s next-generation UDNA (RDNA 5) GPUs could bring hefty upgrades, including up to 96 compute units and a wide 384-bit memory bus, boosting competition in high-end graphics and Artificial Intelligence workloads.

AMD is set to return to the high-end GPU market with its upcoming UDNA, also referred to as RDNA 5, architecture. Leaked details from trusted source Kepler_L2 suggest that the flagship Navi 5X SKUs will pack up to 96 compute units (CUs) and sport an expansive 384-bit memory bus, marking a significant leap from the previous RDNA 4 generation, which was limited to mid-range offerings.

The leaked roadmap also outlines AMD´s plans for multiple tiers within the UDNA lineup. Mid-tier GPUs are expected to offer configurations such as 64 compute units alongside a 256-bit memory bus, catering to mainstream enthusiasts. Entry-level models could feature 32 compute units paired with a 128-bit memory bus, ensuring a comprehensive product range. The exact memory capacities remain unknown, and there is ongoing speculation that GDDR7 could be featured, but AMD´s final choice is yet to be revealed. These preliminary specifications underscore AMD´s intent to challenge both gaming and professional markets, where high memory bandwidth and compute performance are critical.

Following RDNA 4´s absence in the extreme performance segment, the UDNA (RDNA 5) family signals AMD’s renewed focus on reclaiming ground in premium GPUs, which are increasingly important for gaming as well as data-intensive Artificial Intelligence workloads. Mass production of the UDNA GPUs is scheduled for the second quarter of 2026, with availability targeted for the latter half of the year. As launch approaches, AMD is expected to disclose more specifics on memory types and capacities, and enthusiasts are keenly anticipating microarchitectural updates that will define the next wave of Radeon graphics cards.

52

Impact Score

Pope Leo XIV forms Artificial Intelligence study group

Pope Leo XIV has created a Vatican study group on Artificial Intelligence as he prepares to publish his first encyclical. The effort signals a push for an ethics-based approach centered on human dignity, peace, labor, and truth.

Europe accelerates Artificial Intelligence in defence

European militaries are moving from limited Artificial Intelligence support tools to deeper integration in targeting, decision support and weapons systems. France, Germany and the United Kingdom are leading major programmes, while Ukraine is shaping how the technology is tested and deployed.

New LLM architectures target long-context efficiency

Recent open-weight language models are adding targeted architectural changes to cut the cost of long-context inference. Key ideas include cross-layer KV sharing, per-layer embeddings, compressed attention, and wider residual pathways.

Simple Artificial Intelligence recommendations for small business growth

Research from the University of Warwick and Nanyang Technological University, Singapore, examines how small and medium sized enterprises can use simpler Artificial Intelligence recommendation systems without large datasets or costly infrastructure. Findings from a field experiment suggest low data approaches can still increase customer engagement and spending.

Contact Us

Got questions? Use the form to contact us.

Contact Form

Clicking next sends a verification code to your email. After verifying, you can enter your message.