A central processing unit (CPU) is a general-purpose processor optimized for executing varied instructions quickly, whereas a graphics processing unit (GPU) is optimized to perform many similar calculations simultaneously. The CPU prioritizes low latency and control; the GPU prioritizes parallelism and high throughput.
How Their Architectures Differ
A CPU normally contains a small number of powerful cores. Each core includes a control unit, arithmetic logic unit (ALU), registers and cache, enabling it to execute complex instruction sequences and manage the computer system.
A GPU contains many smaller processing units that apply the same or similar instructions to large data sets concurrently. This parallel processing is effective when a problem can be divided into independent calculations.
| Feature | CPU | GPU |
|---|---|---|
| Primary design goal | Low-latency execution of varied instructions | High-throughput execution of repeated calculations |
| Core structure | Fewer powerful cores | Many smaller processing units |
| Best suited to | Sequential logic, branching and system control | Parallel numerical processing |
| Memory emphasis | Large caches reducing access latency | High bandwidth for large data sets |
| Typical tasks | Operating systems and application logic | Graphics rendering, machine learning and scientific simulation |
For example, a game CPU processes player input, rules and artificial intelligence. The GPU simultaneously calculates the colour and position of millions of pixels or vertices. The CPU coordinates the program and sends suitable workloads to the GPU.
A common misconception is that a GPU is simply a faster CPU. It is not: performance depends on the workload. Frequent branching or sequential dependencies often favour a CPU, while highly parallel tasks may run much faster on a GPU.
IB Exam Technique
For A1.1 Computer hardware and operation, explain differences through architecture and processing style, then link each design to an application. Do not only count cores; explain how the design affects latency, throughput and suitability for parallel tasks.
