Abstract:
The release of next-generation gaming graphics cards from Nvidia and AMD may be delayed again. The latest news is that Nvidia’s GeForce RTX 60 series based on Rubin architecture, as well as most of AMD’s new generation Radeon gaming graphics cards based on RDNA 5 architecture, may be postponed until 2028.
The main reason behind this is not simply a matter of R&D progress, but the continued strong demand for AI that is constantly devouring key resources such as chips, packaging, and video memory. Data center products can bring profits to the two companies that are much higher than those of consumer-grade game graphics cards.
This news comes from Kepler_L2, a whistleblower who has been paying attention to the GPU product route for a long time. When he recently responded to questions about the release time of Nvidia and AMD's next-generation gaming graphics cards at the AnandTech forum, he said that almost all of the next-generation product lines of the two companies may be postponed to 2028. In 2027, it is expected that only AMD's GPU codenamed "AT2" is still expected to be launched as planned.
If this statement is ultimately confirmed, it will obviously conflict with previous rumors that the RTX 60 series may be released in the first half of 2027. YouTube channel Moore's Law is Dead previously claimed that Nvidia is already preparing the next generation RTX 60 series, including the flagship RTX 6090, and plans to launch it in the first half of 2027. However, Kepler_L2 directly denied this statement.
In fact, there has been uncertainty about the release time of the next generation of gaming graphics cards. In June this year, Dutch media Tweakers asked multiple AMD graphics card partners during Computex. Some manufacturers expect AMD's next-generation graphics cards to be shipped from mid-2027 to the second half of 2027, but others believe the products may be delayed to early 2028. Since there is often a time lag between graphics card launches and actual large-scale launch, even if AMD announces RDNA 5 at the end of 2027, the actual product may already enter the retail market in 2028.
The outbreak of the AI industry is becoming an important factor affecting the GPU product route. In recent years, large-scale language models, generative AI, and AI data centers have continued to expand, resulting in long-term tension in the production capacity of advanced chips, HBM high-bandwidth memory, DRAM, and advanced packaging used in AI accelerators. For Nvidia and AMD, the demand for data center AI products is not only much higher than that of traditional consumer-grade graphics cards, but also has higher profit margins. Therefore, in the case of limited production capacity, there is an obvious commercial motive for prioritizing resource allocation to AI business.
Video memory supply may also be a factor that affects the release time of gaming GPUs. The GDDR video memory used in game graphics cards shares some manufacturing resources with other DRAM products, and HBM, which is widely used in AI accelerators, takes up more and more advanced production capacity of memory manufacturers. The market has already experienced rising graphics memory prices and tight supply, which has also put continued pressure on the manufacturing costs of consumer-grade graphics cards.
AMD’s next-generation RDNA 5 architecture still lacks official information, but previously leaked news shows that the architecture may undergo major design adjustments, including improving dual-emission execution capabilities to improve the efficiency of shader workloads. This change not only involves GPU hardware design, but also requires the cooperation of drivers and software ecosystem. Therefore, AMD may be more inclined to set aside sufficient time to complete product verification rather than rush to launch it to meet the fixed release time.
The current AMD next-generation product information leaked also includes a GPU code-named AT2. It is rumored that AT2 uses RDNA 5 architecture and has 8 groups of Shader Engines, each group contains 5 computing units, a total of 40 computing units, and may be equipped with 12GB of video memory and 192-bit video memory width.
According to the rumored product positioning, AT2 may be a mid-to-high-end model in AMD's next-generation products, and its performance positioning may be close to Nvidia's existing GeForce RTX 5080. Since AMD's current flagship Radeon RX 9070 XT is mainly in competition with the GeForce RTX 5070 Ti, if the AT2 is launched according to the rumored specifications, its performance positioning will be significantly higher than AMD's current product line.
However, the above AT2 specifications and positioning currently come from third-party information that has not been confirmed by AMD, so there is still great uncertainty.
Nvidia has publicly demonstrated the Vera Rubin platform for AI data centers, but remains silent on consumer-grade Rubin gaming GPUs. According to previous product route rumors, Rubin will become the infrastructure of the RTX 60 series, but there is currently no official information on specific product specifications and release time.
AMD’s flagship RDNA 5 GPU has also seen some revelations. Kepler_L2 previously revealed that AMD's next-generation flagship Navi 5X GPU may have a maximum operating frequency of 3.4GHz, 12288 cores and 32GB GDDR7 video memory. If these specifications are finally realized, AMD's new generation of flagship products will see significant improvements in both hardware scale and video memory capacity.
It is worth noting that there are still clear differences in the current opinions on whether to release the next generation of gaming GPUs in 2027 or 2028. Previously, the time windows given by many AMD board partners were mainly concentrated in the second half of 2027 to early 2028, and the latest revelations further point to 2028, and it is believed that most of the next-generation game graphics cards of Nvidia and AMD may be delayed at the same time.
If this timetable eventually becomes reality, the life cycles of the RTX 50 series and Radeon RX 9000 series will be significantly longer, and consumers will have to wait longer for the next generation of mainstream GPUs than in recent years. For Nvidia and AMD, as the AI data center business continues to grow rapidly and profits are much higher than those in the consumer market, the gaming GPU product line may have to give way to the AI accelerator business with more commercial value.
Comments