Samsung: second generation NPU in artificial intelligence technology
Samsung: second generation NPU
Nowadays, the development of artificial intelligence technology in hardware may still be in its infancy, but the main players are not willing to leave the team and have entered the ranks. Samsung’s ai chip business started relatively late, but it is also unambiguous. Recently, a linkedin on the LinkedIn profile shows users of Samsung’s former employees, saying that Samsung may have completed its second-generation npu (neural processing unit). The architecture will be built with the exynos 9820 processor.
Samsung is expected to launch a dedicated second-generation npu in the exynos 9820 soc, which will be unveiled with the upcoming galaxy s10 and galaxy s10+ smartphones, which will replace the current vpu (vision processing unit) chip with new optimized multi-function hardware. And can complete all the tasks that the vpu chip can handle. At the same time, some of Samsung’s products still use Qualcomm’s chips. According to recent industry news, Qualcomm is also expected to be equipped with independent npu for the first time in the next generation of flagship chipsets, supporting audio, imaging and machine learning based on the Snapdragon 845. The function of the hexagon dsp is compressed as much as possible, synthesizing all parties, this processor may be named Snapdragon 8150.
Frankly speaking, Samsung’s first ai chip with built-in npu was released at the beginning of this year, and it is not surprising that the second-generation npu was launched at this time. On the contrary, this move has been quite conservative. Samsung exynos and Qualcomm Xiaolong have long been known at the mobile processor level. Every breakthrough of Qualcomm chips is alerting Samsung that there is no stagnation in the chip layout of exynos.
In fact, the entire mobile phone industry is investing a lot of work in ai processing and related hardware technologies to solve the mobile workload problem. At this point, both Huawei and Apple are committed to developing the corresponding npu in order to gain an edge in the fierce competition: Kirin 980 is equipped with a new dual npu, which can process 4,500 images per minute; Apple claims to be the first commercial 7nm process The mobile processor a12 bionic (a12 bionic), with an 8-core neural engine, can run 5 trillion times per second. Although there is currently no unified workload scenario or suitable metrics to directly compare the advantages and disadvantages of these two chips, it is clear that the gate of this arena is completely open. In the near future, the application of artificial intelligence in technology The field, especially mobile devices, will last a long time.
Although many vendors tend to associate artificial intelligence with voice assistants, artificial intelligence is widely used and can be combined with multiple technologies to achieve novelty and even rare features. In general, in recent years, artificial intelligence technology has made significant progress in the newly developed market clock, optimizing various functions such as image recognition and graphics rendering.
If you are fing one stop PCB supplier for your AI products, we are here to support you in bare PCB manufacturing and PCB assembly, including SMT and DIP, and components sourcing, we have professional team to save your time and energy and avoid risk. Please feel free to contact us for more information: