Recently, the AMD Ryzen AI 5 430, codenamed "Gorgon," has surfaced in the BAPCo database, confirming the existence of this quad-core Zen 5 APU. It is equipped with 8MB of L3 cache and 4MB of L2 cache, mirroring the cache size of its predecessor, the Ryzen AI 5 330. The most significant enhancement in this release is the upgrade in integrated graphics, from a 2 CU Radeon 820M to a 4 CU Radeon 840M.

The product ID, stepping number, and the reference motherboard designation "Korat Plus-GPT3" suggest that both the "Gorgon" and "Krackan" belong to the same core architecture. They are differentiated by unit activation, frequency tweaking, and power consumption adjustments to form distinct SKUs. AMD's approach of deploying the same silicon across the 15-28W TDP range while distinguishing models by CU count and boost frequency illustrates a standard mobile APU binning strategy. For configurations with smaller caches like the 430, potential defects in earlier production stages likely result in entry-level SKUs by deactivating specific CCXs or CUs, thereby optimizing the wafer yield.

The doubling of integrated CUs from 2 to 4 delivers an immediate increase in computational power. The Radeon 820M previously experienced bottlenecks in texturing and rasterization; the 840M alleviates these bottlenecks, improving latency for 3D and multimedia tasks with lighter loads. While frequency details remain undisclosed, the RDNA 3.5's scheduling and caching architecture brings tangible benefits particularly for low-CU designs on bandwidth-constrained mobile platforms.

A CrossMark test platform equipped with 64GB of DDR5-5600 memory indicates that testers aimed to minimize memory-related performance bottlenecks to accurately evaluate the core's standalone capabilities. The Ryzen AI 5 430's performance improvement—showing a 19% gain over the 330—is explainable by potential boosts in frequency or scheduling enhancements. With the quad-core Zen 5 gaining similar improvements in IPC within the same power envelope, primarily if boosted by approximately 100-200 MHz, these results align with prior expectations.

The shared designation of Family 26 Model 104 Stepping 0 across "Gorgon" and "Krackan" confirms that internal iterations remain minimal, with distinctions primarily in enabled unit counts and voltage configurations. The presence of 14 PCIe Gen4 lanes indicates that this SKU maintains connectivity options for NVMe storage and discrete GPUs, thus offering full APU functionality rather than a scaled-back feature set.

Analyzing the leaked Ryzen AI 400 series architecture reveals a strategic tiered structure: with configurations cascading from 12 CU down to 2 CU, each step corresponds to a varying number of CPU cores. This diversity in product density from the same physical wafer not only bolsters yield rates but also affords OEMs more versatility within the 15-45W power bracket. Positioned at the low end of the complete GPU spectrum, the Ryzen AI 5 430 fits snugly between the 2 CU Ryzen AI 3 and the 4 CU Ryzen AI 5 440 within this hierarchy.

The primary enhancements to the Ryzen AI 5 430 focus on doubling the CU count and increasing the frequency. The merger of quad-core Zen 5 processing and 4 CU RDNA 3.5 graphics narrows the graphical performance gap seen in the original 330, especially in the 15-28W mobile power segment. For OEMs, it presents an option for more robust graphics performance without scaling up the silicon size. For AMD, these SKUs enhance overall wafer efficiency by decreasing active unit count while controlling costs.