What AMD’s Computex CPU Roadmap Is About
AMD’s Computex CPU roadmap refers to the set of near-term Zen 5 refreshes, long-term Zen 6 Medusa processors, and new X3D gaming CPUs that define how AMD plans to update desktop and mobile platforms for gamers, creators, and mainstream PC users over the next several product cycles. At Computex, AMD is focusing on a mix of current-generation refinements and early disclosures about its next architecture. Zen 5-based Ryzen mobile and desktop parts continue to roll out, while the company hints at Zen 6 Medusa Point and Medusa Halo as future high-core-count mobile chips. Alongside these, AMD is extending its platform support for AM5 motherboards, refreshing its classic Ryzen 7 5800X3D, and preparing a new Ryzen 7700X3D. The result is a roadmap that tries to keep older systems relevant while preparing for higher core counts, faster memory, and better gaming performance.
Zen 5 Refreshes for Desktop and Mobile
AMD is not finished with its AMD Zen 5 CPU lineup and plans more SKUs for both laptops and desktops. On mobile, the company has already introduced the Gorgon Halo family, with the Ryzen AI Max+ 495 positioned as the flagship and offering higher boost clocks than its predecessor. Additional Gorgon Halo models are expected to fill out the stack. On desktop, the highlight is a new AMD X3D gaming CPU: the Ryzen 7 7700X3D. This Zen 4-based chip keeps the familiar 8-core/16-thread layout and 96 MB of L3 cache, effectively acting as a slightly lower-clocked sibling to the 7800X3D. According to The FPS Review, the Ryzen 7700X3D is set to arrive on July 16th, 2026, as a gaming-focused option for the AM5 platform, keeping 3D V‑Cache accessible at a lower tier.
Zen 6 Medusa: Next-Generation Architecture and Core Counts
Beyond Zen 5, AMD is lining up Zen 6 Medusa as its next CPU architecture for mobile platforms. Reports point to two major families: Medusa Point and Medusa Halo, both built for a larger FP10 socket and designed to increase core counts per CCD. Zen 6 is expected to allow up to 12 cores in a single CCD, compared with 8 cores on current Ryzen designs, which opens the door to much denser high-end chips. Medusa Halo, the flagship Ryzen AI Max 500 series, targets up to 24 cores and 48 threads plus as much as 96 MB of L3 cache, paired with an RDNA 5 iGPU and support for LPDDR6 memory. This combination signals a future where mobile processors deliver desktop-class multi-threaded performance and far higher memory bandwidth, especially for AI and GPU-heavy workloads.
X3D Gaming CPUs, AM4 Revival and Platform Longevity
AMD is also using Computex to underline its X3D cache strategy for gaming. The Ryzen 7700X3D brings 3D V‑Cache to a slightly lower clocked, more affordable AM5 chip, while the iconic Ryzen 7 5800X3D returns as an AM4 Anniversary Edition. The revived 5800X3D keeps its original specifications but comes back at USD 349 (approx. RM1,640), down from its 2022 launch price of USD 449 (approx. RM2,110). For existing AM4 users, this is a clear upgrade path that extends the life of older systems. At the same time, AMD is extending AM5 motherboard support to 2029, far beyond its initial guidance. This confirms that Zen 6 and likely later generations will land on the same socket, giving builders a long upgrade runway and making the platform safer for long-term investment.
EXPO Ultra Low Latency Memory and Radeon RX 9070 GRE
To support its CPUs, AMD is updating the memory ecosystem and adding a new GPU. AMD EXPO Ultra Low Latency is a new class of memory modules with tuned EXPO profiles that promise lower latency settings and automatic memory overclocking. AMD claims these kits can provide “an additional 4% FPS (on AVG) vs non-ULL EXPO memory” while also improving 1% lows in games. For graphics, AMD is preparing the Radeon RX 9070 GRE, a new card positioned for mainstream gaming builds that will naturally pair with Zen 5 and Zen 6 systems plus the new X3D chips. Together, the EXPO ULL memory and RX 9070 GRE signal that AMD’s Computex 2026 announcement is not only about CPU architecture, but about closing the loop on platform performance from memory tuning to GPU horsepower for gaming PCs.
