Jason Cross, editor of MacWorld, analyzed the upcoming Apple A14 chip. He said that moving from a 7-nanometer process to a 5-nanometer process may not sound like much advancement, but can potentially be considered a major upgrade. The 5 nm process means that the A14 chip can have 12.5 billion transistors, which is more than desktop and server CPUs. Apple may cut the total chip area to about 85 mm square, which will dramatically improve its performance, especially in terms of multi-core performance.
Jason Cross believes that the multi-core benchmark score of the A14 chip may be around 4,500, or it may even exceed 5,000. At present, the highest multi-core score of an Android phone is around 3000. A score of 5,000 means that it will be close to the score of 6-core mainstream desktop CPU or high-end laptop CPU, and it can already match the 15-inch MacBook Pro.
In addition, more transistors in the GPU and rumored 6 GB of RAM are expected to improve gaming performance by about 50 percent.
Cross suggests that the A14 chip will have bigger progress in neural engines. Because the 5nm manufacturing process provides higher transistor density, he also believes that Apple will add neural engine cores and may also make other architectural improvements. It is expected that A14 machine learning will be at least twice as fast as A13.
Jason Cross believes that the multi-core benchmark score of the A14 chip may be around 4,500, or it may even exceed 5,000. At present, the highest multi-core score of an Android phone is around 3000. A score of 5,000 means that it will be close to the score of 6-core mainstream desktop CPU or high-end laptop CPU, and it can already match the 15-inch MacBook Pro.
In addition, more transistors in the GPU and rumored 6 GB of RAM are expected to improve gaming performance by about 50 percent.
Cross suggests that the A14 chip will have bigger progress in neural engines. Because the 5nm manufacturing process provides higher transistor density, he also believes that Apple will add neural engine cores and may also make other architectural improvements. It is expected that A14 machine learning will be at least twice as fast as A13.
Post a Comment