At CES 2023, AMD unveiled in more detail two upcoming chip lines for mobile devices: Dragon Range and Phoenix. Both are based on the Zen 4 architecture and have a chiplet design, but the production node changes. Dragon Range is in fact produced at 5nm, while Phoenix at 4nm and also integrates an RDNA3 GPU combined with the first Ryzen AI engine on XDNA architecture, which also exploits the know-how of Xilinx, a company acquired several months ago by AMD.
The Dragon Range HX series will equip notebooks aimed at the most demanding gamers and content creators. This new range of chips leverages existing Ryzen 7000 CPU chiplets by placing them inside a smaller BGA package specifically for laptops. This allows for over 55W processors using the Zen 4 architecture and integrated RDNA 2 GPUs.
The flagship model, the Ryzen 9 7945HX , reaches a maximum of 16 cores and 32 threads , just like the desktop models, giving AMD the highest number of 'performance cores' on the market. This chip has a boost clock of up to 5.4 GHz, is equipped with an impressive 80 MB of total cache and has an official TDP that starts at 55W and which on some selected models can even go up to 100W. According to AMD, the 7945HX is 29% to 62% faster in games than the Ryzen 9 6900HX. Additionally, AMD claims that the 7945HX is 18% faster in single-threaded and 78% faster in multi-threaded workloads than the Ryzen 9 6900HX.
Photo Credit: AMD
Photo Credit: AMD The Ryzen 7040 HS “Phoenix” series is aimed at slim and compact laptops. These 35-45W chips feature Zen 4 CPU architecture on a 4nm process paired with RDNA 3 GPU. The flagship model offers 8 cores and 16 threads operating at up to 5.2GHz, with a total of 40MB of cache. Undoubtedly, the most interesting feature of these processors is the integration of the new FPGA-based AI engine directly into the die, capable of handling up to 4 simultaneous AI streams, although it can be quickly reconfigured to handle varying amounts of streams. AMD claims that the “Ryzen AI Engine”, based on the XDNA architecture, is fully programmable and faster than even Apple's neural engine in the M2 SoCs.
Finally, AMD introduced the 6nm Rembrandt-R chips and Barcelo-R at 7nm, which are only an update of the predecessors, respectively, with Zen 3+ and Zen 3 architecture. Rembrandt uses integrated RDNA 2 GPUs, while Barcelo remains with the older Vega.