Parallel Processing and Multi-Core Processors: GCSE Computing Guide

Modern computers do not work through problems one instruction at a time — they divide work across multiple cores and execute tasks simultaneously. Parallel processing is the reason a phone can stream music, run a navigation app, and receive messages all at once without grinding to a halt.

What is parallel processing?

Parallel processing means carrying out multiple computations at the same time rather than one after another (sequentially). Instead of a single worker completing each task before starting the next, parallel processing employs many workers each handling part of the total workload simultaneously.

The opposite is serial (sequential) processing: one task completes, then the next begins. Early processors were entirely serial — fast, but fundamentally limited by the single stream of instructions.

A simple analogy: washing up ten bowls one at a time is serial. Handing two bowls each to five people and washing simultaneously is parallel. If each person takes 10 seconds per bowl, the serial approach takes 100 seconds; the parallel approach takes 20 seconds — a fivefold speedup.

What are CPU cores and how do they enable parallelism?

A core is an independent processing unit within the CPU. It has its own fetch-decode-execute pipeline and its own set of registers and cache. A dual-core processor contains two such units; an octa-core processor contains eight.

Each core can execute a different instruction stream simultaneously. The operating system divides running programs (and their threads) across the available cores, scheduling which core handles which task at each moment.

What is pipelining?

Pipelining is a different form of parallelism that operates within a single core. Rather than waiting for one instruction to fully complete before fetching the next, the CPU overlaps the stages of multiple instructions simultaneously.

The classic five-stage pipeline:

1. Fetch — retrieve the instruction from memory
2. Decode — identify what operation it represents
3. Execute — perform the operation in the ALU
4. Memory access — read from or write to RAM if needed
5. Write-back — store the result in a register

While instruction 1 is in the Execute stage, instruction 2 can be in Decode and instruction 3 can be in Fetch. Five instructions move through the pipeline simultaneously, each at a different stage — like an assembly line where different cars are at different stations at the same time.

Without pipelining, a 5-stage process on 5 instructions takes 25 clock cycles. With pipelining, it takes just 9 (5 to fill the pipeline + 4 more for the remaining instructions). A modern CPU pipeline has 10–20+ stages.

What limits the benefit of adding more cores?

Not every task can be split across cores equally. Some parts of a program are inherently sequential — step 2 cannot begin until step 1 finishes because it depends on step 1's result. Only the parts that are independent can run in parallel.

Amdahl's Law quantifies this. If 90% of a task can be parallelised and 10% must remain sequential, then even with infinitely many cores the maximum speedup is:

Maximum speedup = 1 / (1 - 0.9) = 1 / 0.1 = 10×

Adding more than about 10 cores gives diminishing returns for that task. This is why simply buying a processor with more cores does not automatically make all software faster — the software itself must be written to exploit parallelism.

What is the difference between multi-core and multi-processor systems?

Modern consumer devices almost universally use multi-core single-CPU designs because on-chip communication between cores is far faster than bus communication between separate chips.

Frequently asked questions

Does more cores always mean a faster computer?

Not necessarily. Clock speed, cache size, memory bandwidth, and the nature of the workload all matter. For single-threaded tasks (where only one execution stream is running), a processor with fewer, faster cores often outperforms one with many slower cores. For tasks that can be parallelised — video encoding, 3D rendering, compiling code — more cores provide genuine speedup up to the limit imposed by the sequential portions of the work.

What is a thread and how does it relate to cores?

A thread is the smallest unit of execution that the operating system can schedule — a sequence of instructions within a program. A program may have many threads running concurrently. Modern CPUs support hyper-threading (Intel's term) or simultaneous multithreading (SMT), where each physical core presents two logical cores to the operating system by keeping two sets of registers. This allows the core to switch between two threads on every clock cycle, hiding memory access latency. A quad-core CPU with hyper-threading appears as eight logical cores to the OS.

Why did processor clock speeds stop increasing dramatically around 2004?

Increasing clock speed requires more voltage, which generates more heat. Around 2004, processors reached the practical limit of heat dissipation — running faster would require cooling beyond what was commercially viable. Engineers responded by adding more cores rather than increasing individual core speed. This is why modern CPUs have higher core counts than processors from 20 years ago, even though individual core clock speeds are not dramatically higher.

How does parallel processing relate to the GPU?

The GPU is an extreme example of parallel processing: rather than having 4–32 powerful cores like a CPU, a GPU has thousands of simpler cores all executing the same operation on different data simultaneously. This SIMD (Single Instruction, Multiple Data) approach is ideal for graphics rendering and AI training, where the same calculation must be applied to millions of pixels or matrix values at once. The GPU and CPU work together: the CPU handles serial coordination and the GPU handles bulk parallel computation.

Explore how processors are designed to maximise speed — Professor Turing at aitutors.me will connect hardware architecture to the algorithms you write.