PC WITH AI (NPU)
Hachman had the opportunity to test a notebook using a real Core Ultra 7 165H and try running the Procyon benchmark to measure AI processing performance. and found that the GPU is still the core The scores for each part are as follows: Procyon (OpenVINO) NPU: 356 Procyon (OpenVINO) GPU: 552 Procyon (OpenVINO) CPU: 196 Points out that having an NPU can actually help improve the efficiency of some AI tasks. But in the end, the GPU is still more important. Plus, the real benefit of NPU is not performance. But it is better to use energy. Helps the notebook battery last longer. (A clear example of usage is Video calling in Microsoft Teams for a long time and let AI help cut out the noise Blurring the background like this, the NPU helps the battery slow down) Hachman concludes that we are only just beginning to enter the AI PC era, and it is good to have a built-in NPU chip. But it's not a prerequisite. Because we can always run AI on the GPU. Plus, most users probably run AI through the cloud. In addition, the hardware will continue to evolve. Intel's own example is its Lunar Lake chips. The next generation will be released by the end of 2024 and will have 3x better NPU performance than before.
