According to multiple channel vendors revealed to Dutch technology media Tweakers during the 2026 Taipei International Computer Show, AMD's next-generation Radeon gaming graphics card based on the RDNA 5 architecture is still a long time away from its official debut, and it will not be available until the third quarter of 2027 at the earliest. Some partners even believe that it will not be available until the end of 2027 or even early 2028. This means that players will have to wait at least another two to two and a half years from the current generation of products to see a truly new generation of Radeon gaming GPUs.
AMD's latest generation of products for consumer players is the Radeon RX 9000 series released last year, based on the RDNA 4 graphics architecture. This year, AMD launched the Radeon RX 9070 GRE based on this architecture for 1440p resolution players, continuing to tap the potential of the existing platform to fill the product line. However, for users looking forward to "cross-generation upgrades", the absence of the new generation of RDNA 5 game graphics cards will undoubtedly be further stretched.
According to Tweakers, after communicating with multiple AMD board partners during Computex, most manufacturers set the time window for the first batch of RDNA 5 gaming GPUs in the second to third quarter of 2027. Some people bluntly said that this expectation was "too optimistic" and believed that a more realistic time would be the end of 2027 or even the beginning of 2028. If this rhythm comes true, the new generation of desktop gaming graphics card market will be maintained mainly by iterative products of the existing architecture for some time to come.
According to reports, the core reason for the overall delay in next-generation gaming graphics cards is the extremely tight supply chain environment, especially the continued shortage and rising prices of key components within the graphics card. In the current market, AI-related hardware is given priority by major manufacturers to ensure production capacity. Storage manufacturers are difficult to meet the needs of data centers, AI accelerator cards and traditional PC markets at the same time in the short term, resulting in a significant squeeze on game graphics card materials. Judging from industry chain expectations, this supply and demand imbalance may continue until 2028-2029, which will continue to affect the product rhythm and pricing space of the new generation of consumer-grade GPUs.
In terms of competitors, NVIDIA has also encountered a similar situation: the RTX 50 series has now entered 1.5 years after its launch. The market has spread the so-called "RTX 50 SUPER" mid-term refresh plan, but it is still essentially a product enhancement of the existing architecture. Previously, NVIDIA was widely expected to launch a new Rubin architecture in 2027, and has released relevant signals through the announcement of Rubin CPX GPU. However, in recent times, officials have rarely mentioned the specific progress of this architecture. This also confirms to a certain extent that the entire high-end GPU industry is being constrained by AI hardware demand and upstream production capacity constraints.
Regarding RDNA 5 itself, the current public information mostly comes from early leaks and software code traces, and the overall information is still relatively limited. There have been previous rumors that the RDNA 5 flagship chip will integrate more than 12,000 stream processing cores in a single GPU and adopt a 128-core design per computing unit. At the same time, some early configurations show that its high-end models can provide up to 96 computing units, 384-bit or even 512-bit memory buses, and 24GB-32GB of video memory capacity. Mid-range and entry-level chips correspond to fewer computing units and memory buses ranging from 384-bit to 64-bit. The memory capacity range is roughly between 8GB and 24GB to cover the needs of mainstream players in different price ranges.
Judging from software-side signals, some next-generation RDNA GPUs (internal IP identification is GFX13) have appeared in the early Linux kernel code base, showing that AMD is laying the groundwork for drivers and ecology for the new architecture. Previously, there was information that Radeon gaming GPUs based on the RDNA 5/UDNA architecture were originally planned to enter mass production on TSMC’s N3P process node in the second quarter of 2026, but under current supply chain pressure, this time point may also be postponed to the end of 2026. AMD recently stated publicly that achieving the so-called "perfect" Radeon gaming platform still requires several generations of product evolution, implying that the company will continue to adjust its route on subsequent architectures to improve light tracing performance, image enhancement capabilities and bandwidth utilization efficiency.
Based on the current industry chain news, whether it is AMD or NVIDIA, the traditional PC game graphics card business will have to find a balance under the squeeze of the AI wave and production capacity shortage in the next two to three years. Against this background, the mid-term refresh, specification expansion and price strategy of the existing RDNA 4 and RTX 50 series architectures will play a role in maintaining the activity of the gaming GPU market for a longer period of time. The truly next-generation high-end gaming graphics cards may not be released until the supply of memory and advanced processes has eased.