The intel processors ranked, previously known under the codename Tiger Lake, introduced new features for processors that enhance the performance of AI applications by taking advantage of fewer data representations, which can provide faster results for the inference task. We want to guide you through the process of running benchmarking tests yourself using Intel(r) DevCloud for the Edge by using the Deep Learning Workbench provided in the OpenVINO(tm) toolkit.

In this article, we're conducting our tests using three different systems located in the DevCloud and present an evaluation of performance on AI applications using FP16 as well as IN8 date representations. If you're interested in learning more about rigorous benchmarks for performance, you can find them on the OpenVINO Benchmarks page.

System Comparison

To conduct our research, We'll be using hardware with similar characteristics in three distinct generations of the models available in the DevCloud. We'll examine the performance from Intel(r) Core(tm) I7-1185G7E, Intel(r) Core(tm) the i7-10710U processor, as well as best intel processor. The comparison of the three processors is provided on the Intel specifications page for the product. The table below shows that the intel processors, previously known as Comet Lake, has a higher maximal turbo frequency as well as more CPU threads and cores. We should use systems that have identical configurations. However, it is still possible to see trends in the efficiency of the systems in the outcomes of tests using different data formats.

A significant difference between the i7-1185G7E versus the previous generation processors is the addition of new features for processors: Intel(r) AVX-512, Intel Deep Learning Boost, and Vector Neural Network Instructions. These functions allow for additional optimizations, which were impossible with earlier-generation processors when used with Intel's INT8-based precision.