How MediaTek’s chips save smartphone memory – 2nm flagship Dimensity 9600 Pro unveiled

Taiwanese company MediaTek decided to promote its latest mobile processor Dimensity 9600 Pro in a rather extraordinary way, announcing the advantages offered by the AI ​​technology built into the chip. They can reduce memory load, which cannot be considered a minor factor in low memory situations.

Image source: MediaTek

The Dimensity 9600 Pro is produced using Taiwan’s TSMC 2nm process, which also contributes to the energy efficiency of this smartphone component. JC Hsu, senior vice president of MediaTek, said that the average price of smartphones has grown significantly As memory prices rise, the company is working immediately with multiple partners to reduce the need for large amounts of memory in the CPU without affecting overall performance levels. He also added that MediaTek was one of TSMC’s first customers, providing them with services to produce chips using 2nm technology. MediaTek representatives said the cost of flagship processor models has also increased due to the shift to more advanced production technology. The Dimensity 9600 Pro, also the flagship, concentrates 33 billion transistors and a complete set of functional blocks on a small chip area.

For example, the neural co-processor reduces the chip’s power consumption by 40% when processing AI models compared to its predecessor, while improving performance by 50%. The company supplies processors to Samsung Electronics, Google, Oppo, Vivo, Amazon and Xiaomi. Partners will demonstrate smartphones based on the new chip models this month and may be available in October.

Dimensity 9600 Pro adopts a 2+3+3 layout, which can combine 2 Arm C2-Ultra cores, 3 Arm C2-Pro cores with 1 MB L2 cache, and 3 Arm C2-Pro cores with 512 KB L2 cache. In fact, the new MediaTek processor increases the size of the L2 cache by 21% compared to its predecessor. In single-core measurements, the Dimensity 9600 Pro is 17% faster and 37% more energy efficient than its predecessor. In the multi-core dimension, the new chip is 15% faster than the previous generation chip, and the energy efficiency level has increased by 61%. The processor generates tokens 40% faster in large language models and runs AI models natively 27% faster.

In addition, the chip supports LPDDR6 and UFS 5.0, speeding up memory work. In terms of communication, it supports 5G networks, modern versions of Wi-Fi and Bluetooth 6.2.