AMD at Computex 2025: Between RDNA 4, AI PCs and workstation updates – expectations for the presentation

Taipei – AMD will unveil its new product strategy at Computex 2025 on May 21 at 11:00 a.m. local time (UTC 8). The event will be chaired by Jack Huynh, SVP and GM of the Computing and Graphics Group, flanked by industry partners and OEMs who are also expected to have their say. The keynote will focus on three strategic pillars: Gaming, AI PCs and Enterprise Solutions.

RDNA 4 – mainstream graphics cards just around the corner

While the Radeon RX 9070 series has already arrived on the market and has maintained its position between enthusiast and upper-class hardware in relevant benchmarks, there has been no official expansion of the series into the lower price range to date. The RX 9060 series has already been announced by name, but without concrete technical key data. An expansion of the RDNA-4 portfolio is expected with a GPU that will presumably once again rely on TSMC’s 4 nm process, but with reduced compute density and a potentially slimmed-down Infinity Cache. The only realistic candidate for the chip base is the Navi 44, about which little reliable information has been circulating to date. AMD is likely to use this opportunity to consolidate its position in the highly competitive mid-range segment – not least because Nvidia already offers corresponding models with the RTX 4060/4060 Ti. It remains to be seen whether AMD will opt for GDDR6 with 16 Gbps or a memory interface below 128 bits, which in turn would have an impact on memory bandwidth and ultimately performance.

AI PCs – more than just buzzwords?

The second category of the presentation concerns so-called “AI PCs”, a term that has recently been used in an inflationary manner – often as a fraudulent label. AMD had already indicated that future mobile and desktop CPUs would be equipped with dedicated NPU units (Neural Processing Units) to accelerate locally running AI applications such as language models, image processing or office automation. It remains to be seen whether the next generation of Ryzen processors, presumably based on “Strix Point” or a refresh of the Phoenix APUs, will actually offer a significant improvement on the hardware side. The integration of XDNA architecture components from the Xilinx acquisition should at least find its way onto the data sheet. In practical terms, however, the decisive factor remains whether the software ecosystem – above all Microsoft with Windows 11 – is prepared to use these NPU units in real applications beyond demos.

Enterprise: EPYC and Instinct on the sidelines

While the focus of the presentation is clearly on the consumer and prosumer segment, AMD is also announcing an update to its enterprise division. However, specific details about the EPYC series – such as a refresh of the Turin processors or initial indications of a Zen 5-based server platform – are likely to be mentioned in passing. The same applies to the Instinct accelerators, where AMD has caught up technically with the MI300 series, but still suffers from Nvidia’s market dominance. As AMD’s infrastructure products are strongly tailored to hyperscalers and large customers, a detailed presentation at Computex is not to be expected. However, there may be initial roadmap updates or collaborations with partners such as Lenovo or Dell, which could point to new AI workstations.

AMD’s Computex keynote should deliver tangible results, especially in the GPU and AI PC segment. A launch or at least a deeper look at the Radeon RX 9060 series seems likely, as well as new Ryzen CPUs with extended NPU integration. Only a brief sign of life is expected for the enterprise segment. To what extent AMD’s “vision for AI PCs” is more than a rebranding of existing technologies remains to be seen. The event begins on May 21 at 11:00 a.m. local time (Taipei), or on May 20 at 11:00 p.m. CEST. A livestream will be provided via AMD’s official channels.

Source: AMD