AMD is slated to make a flurry of new product announcements and launches, harnessing the capabilities of its forthcoming Zen 5 core architecture. The Zen 5 architecture promises to cater to a wide array of market segments, from data centers and servers to laptops and more, showcasing AMD’s ongoing commitment to innovation and performance.
In a significant move, AMD is targeting the 10th of October for the release of its next-generation server and data center lineup. This lineup is anticipated to be headlined by products such as the Ryzen AI PRO 300 series, the 5th Gen EPYC “Turin,” the MI325X accelerator, and new advancements related to Xilinx, which are expected to encapsulate new security features and technologies that utilize AI capabilities. The Ryzen AI PRO 300 series is poised to offer PRO technologies while maintaining configurations similar to the standard Ryzen AI 300 “Strix” APUs.
The 5th Gen EPYC family will be available in two variants: a 128-core Zen 5 (4nm) design and a more advanced 192-core Zen 5C (3nm) design, both compatible with existing sockets that accommodate up to 96 and 128 cores in the “Classic” and “Dense” flavors respectively of the previous Zen 4 generation.
The Instinct MI325X is expected to be an upgrade over the current MI300X accelerator, boasting up to 288 GB of HBM3E VRAM and reaching bandwidths of up to 6 TB/s. This accelerator is designed to deliver enhanced FP16/FP8 compute capabilities and support significantly larger AI models, potentially up to a 1 trillion parameter model. It is clear that AMD’s product launch on 10.10 predominantly targets server, data center, and professional market segments, but we may also get glimpses of upcoming consumer-oriented products.
Looking ahead to CES 2025, AMD plans to introduce the Strix Halo and Krackan Point APUs. The Strix Halo APUs are engineered for the enthusiast segment, boasting up to 16 CPU cores, 40 GPU cores, a chiplet design providing substantial processing power for demanding applications. On the other hand, the Krackan Point APUs will cater to mainstream platforms, featuring a monolithic design with up to 8 CPU cores and 8 GPU cores.
The anticipated features for the AMD Ryzen AI HX Strix Halo include a Zen 5 Chiplet Design, up to 16 cores, 64 MB of shared L3 cache, and 40 RDNA 3+ Compute Units paired with a 32 MB MALL Cache for the integrated GPU. Additionally, a 256-bit LPDDR5X-8000 Memory Controller, an integrated XDNA 2 Engine, and the capacity to deliver up to 60 AI TOPS illustrate the expected performance capabilities. A launch is expected in the first half of 2025, and these APUs are likely to be part of the FP11 Platform, which ranges from 55W to 130W.
The AMD Ryzen AI HX Krackan Point APUs are anticipated to debut with a Zen 5 Monolithic Design, incorporating up to 8 cores combining Zen 5 and Zen 5C cores, along with 16 MB of shared L3 cache and 8 RDNA 3.5 Compute Units. These APUs will support LPDDR5X and DDR5 memory, feature an integrated XDNA 2 Engine, and offer up to 50 AI TOPS. They are also expected to launch in the first half of 2025 and be a part of the FP8 Platform, suitable for 15W to 45W configurations.
AMD might also introduce “PRO” and “MAX” naming for its higher-end Ryzen AI 300 family with the debut of the Strix APUs, a strategy that will be intriguing to monitor. With AMD’s extensive roll-out of the Zen 5 architecture across diverse market segments, the tech community awaits a promising array of new and powerful computing solutions.
