Semiconductor company ARM, CPUs for next-generation mobile devices supporting 5G, which will be popularized in 2020, are Cortex-A78 and X1, and Mali-G78 and NPU. Introduced Ethos-N78.
First, the Cortex-A78 uses the same architecture as the previous generation CPU, Cortex-A77. However, by optimizing the micro-architecture, it aimed not only to improve performance, but also to increase efficiency in both power and area size. The Cortex-A78 single-core performance, which outputs 1W, is improved by 20% compared to the A77, and power consumption is reduced by 50%, greatly improving energy efficiency. In addition, the DynamIQ cluster, which combines 4 Cortex-A78s and 4 Cortex-A55s, has a 20% higher performance than the existing cluster consisting of 4 A77s + 4 A55s.
Cortex-A78 also has excellent efficiency per milliwatt by reducing the L1-I cache and the L1-D cache, and can exhibit high computing power even in mobile devices that require miniaturization, such as miniature smartphones. It has started spreading this year, and it is revealing that it is possible to respond to 5G.
The Cortex-X1, which was announced together, has a 30% higher peak performance than the existing A77. Combining 1 Cortex-X1, 3 A78s, and 4 A55s can achieve the highest performance than the Daina Mi-Q cluster. And can be expected to be used in large screen devices. In addition, single-threaded integer arithmetic performance increased by 22%, and machine learning performance improved by 2 times compared to the A77.
The Mali-G78 adopts the same Valhalla architecture as the previous Mali-G77. It supports up to 24 cores and increases the performance by 25%. ARM emphasizes that high-quality mobile game experiences and immersive entertainment experiences are possible, and that mobile devices also provide more complex games, videos, cameras, and security functions.
Mali-G78 also increases drawing performance by 17% in 3D games that require more complex CG or rendering than the previous generation and by 15% in machine learning-related computing performance using GPUs. ARM also announced the Mali-G68, a low-cost GPU targeting the mid-range market.
Ethos-N78 is an NPU specialized for machine learning apart from the CPU. It provides 30% higher space efficiency than the previous generation N77, doubles peak performance, 25% higher performance efficiency, and 40% higher DRAM bandwidth efficiency. This way, the battery can be used longer and the software application can perform machine learning related operations efficiently. Related information can be found here .