AMD blog: AMD Ryzen AI MAX+ 395 Processor: Breakthrough AI Performance (Radeon 8060S LLM token-throughput measurements)

AMD first-party blog, used for the two GPU facts the product page omits. Both are AMD's own words: (a) architecture and CU count - 'a truly massive integrated GPU driven by 40 AMD RDNA 3.5 CUs'. Note AMD calls it RDNA 3.5 here, not RDNA 3 and not RDNA 4, and the product page's 'Graphics Core Count 40' agrees on the CU count; (b) the memory bandwidth that matters for LLM decode - 'the ASUS ROG Flow Z13 - powered by the integrated Radeon 8060S and taking full advantage of the 256 GB/s bandwidth'. The 256 GB/s figure is the LPDDR5x system-memory bandwidth available to the iGPU on that machine, so it is a platform-level ceiling for one SKU's memory configuration, not a fixed property of the GPU block; the record says so. The blog also supplies the LLM context for this part: 'system memory options ranging from 32GB all the way up to 128GB of unified memory', and AMD's own 2.2x-over-Intel-Arc-140V token-throughput comparison on the Flow Z13. Because this is vendor-authored measurement on vendor-selected hardware rather than a spec sheet, figures it supplies are still recorded with vendor_claim true, but the record's status rests on the product page, not on this blog. Accessed 2026-10-03.