A decade ago, AI looked very different from what it is now. Apart from the fact that it was not available to the common man, the old AI was also pretty slow. It didn’t respond as instantly as it does today - within literal milliseconds. But today’s modern laptops have built-in AI that’s always at your disposal.
Ever wonder what changed within a decade to make this drastic shift in AI? And what does the holy trinity of modern laptops (aka CPU, GPU, and NPU) have to do with this performance difference? They do have prominent roles behind the scenes, that’s for sure.
But which one has the spotlight for handling AI? That’s what we're going to disclose in this blog, so stay tuned.
What are Processing Units?
Modern AI laptops rely on not one, but three processing units to handle tasks efficiently. These processing units are unique and different, but still work alongside to run AI smoothly.
CPU - The Control Centre
The Central Processing Unit is the brain of any computer or laptop. From a general POV, it is responsible for overseeing the operations of a laptop. From managing apps and organising data to controlling other hardware components, the CPU coordinates everything. It even decides which processor between the GPU and NPU should handle which AI task based on efficiency potential. While it doesn’t perform the heavy AI math, a CPU ensures everything runs smoothly in an AI laptop.
GPU - The Processing Engine
The Graphics Processing Unit was originally designed to handle graphic tasks, such as rendering images and 3D models. But with time, the architecture of GPUs got better. As the number of cores in the GPU increased, this processing unit became eligible for carrying out AI computations. So, the GPU of an AI laptop handles parallel mathematical computations (in bulk). It essentially provides the computer with its speed and multitasking capabilities.
NPU - The AI Specialist
The Neural Processing Unit is specifically designed for AI laptops. Its main goal is to handle the AI and neural network workloads efficiently. You can modern-day AI laptop performing tasks like predictive typing and smart suggestions, where it acts in the capacity of an assistant or advisor - that’s thanks to the NPU. Plus, NPU makes sure that the other resources (like battery, etc.) are not being overused for AI tasks. If you compare the NPU with the CPU and GPU, it will definitely seem more hands-on with the AI tasks.
How AI Works in Modern Laptops?
When you enter an AI task on a laptop (like a voice command), the operating system gets triggered, and the CPU detects it. Since the CPU is the main coordinator in the panel, it decides where that particular task should be allocated. So, let’s say it’s a real-time task, the NPU will get it. But if it’s something heavy, like generating AI content, the CPU will allocate it to the GPU.
In the case of a logical task like detecting errors in a file, the CPU will keep performing the task by itself. Once the CPU assigns the task to the right processor, the NPU and GPU start performing the relevant mathematical computations. The GPU performs massive parallel operations. On the other hand, the NPU runs optimised neural networks efficiently (and non-stop). So it’s not like these processing units replace the CPU, but more like they are under its direction.
As soon as the processing part is complete, the NPU and GPU send the assignment back to the CPU. The CPU puts the complete task in the right place. So it basically integrates the work into the app or system, or the exact location where it needs to be for the user. For an AI-enhanced photo, the location will be the photo editing app. Similarly, a background blur will appear in your video call.
AI Workload Breakdown - CPU vs GPU vs NPU
Efficiency is the key to the success of the modern AI laptop. Credit for this efficiency goes to the three processing units, but there’s one of these which deserves more credit. Which one? Let’s find out.
AI Inference
It is the process by which the system makes predictions about the upcoming moves of the user. Common examples of AI inference include voice recognition and smart suggestions. The NPU is responsible for AI inference in modern laptops as it handles real-time neural network calculations efficiently. While the GPU and CPU also contribute in this department, the bulk of the work is done by the NPU.
AI Training
AI keeps evolving, that’s a fact. But did you know that your system uses your data to train the AI models? On modern laptops, the GPU is responsible for enabling AI models to perform different kinds of tasks efficiently. Since a GPU can perform massive parallel calculations, it helps the AI to learn patterns and behaviour to improve accuracy for future predictions. The better the GPU, the better the AI training.
Real-Time AI Processing
Real-time AI processing is when the AI is processing the result on the backend as you enter the command. The responses are always instant, without wasting even a second. Such tasks include facial recognition and AI camera features, among others. The NPU always handles such tasks like a pro. But it is not alone while performing such tasks. The CPU coordinates the workflow (as usual), and the GPU can also assist (if the computations are heavy).
Final Verdict
In modern laptops, AI won’t work if there’s no collaboration between the CPU, GPU, and NPU. The CPU is like the momager - always coordinating, managing, and integrating. Then we have the GPU, which handles the parallel computing for the heavy AI workload. And finally, the star of the show, the NPU, ensures all real-time and routine AI tasks are done without affecting battery. Together, these processing units rock. But if we were to give credit for most AI work done to one, it’d definitely be the NPU. It’s like the primary engine for AI (especially on-device AI).
