Domestic chips AI run sub-data appearance CES | Apple A11 directly at the end

For smart phones, Apple is indeed a leader in performance and system optimization, but at the AI ​​hardware level, a natural genetic defect is a bit sad reminder.According to reports, during the CES, for four mainstream AI chip ran sub-test, Apple A11 miserable cannon fodder appears to be behind an era.From the screenshots, the more representative of the VGG16, ResNet50, InceptionV4 several neural network model runs the sub-cost, Apple A11 directly bottom.

Almost by the domestic chip Kirin970, RK3399Pro spike rhythm.NVIDIA TX2 in the VGG16, ResNet50 test performed well, but did not give NVIDIA TX2 InceptionV4 data is not supported whether the test is not known.

Data show that in the most important NPU key computing performance, Apple A11 into a dilemma, Rockchip RK3399Pro 2.4TOPs computing power beyond the imagination.

In fact, AI chip run sub-test, Huawei Kirin 970 and Rockchip RK3399Pro these two domestic chip score higher is not surprising, because the two AI chips are used in a separate NPU design and AI hardware performance and CPU, GPU relationship Not too big.

Kirin 970 is the first integration of Huawei NPU using HiAI mobile computing architecture, its AI performance density significantly better than the CPU and GPU. Rockchip RK3399Pro CES2018 debut of the new debut. CPU using big.LITTLE size core architecture, dual-core Cortex-A72 Plus quad-core Cortex-A53 GPU Quad-core Mali-T860 also uses a dedicated NPU hardware architecture designed to support 8bit and 16bit computing, compatible with different software framework; solution cost more, which is conducive to mass production and commercial use. Also has high scalability and other characteristics.

AI hardware can not become a domestic chip corner overtaking, dominant AI artificial intelligence chip opportunities for ecological chain? Eye-catching opportunities for domestic refueling!