CPU Performance and Embedded Systems

Two CPUs can run the same program in very different times. The difference comes from three key characteristics:

Characteristic	What it measures	Unit
Clock speed	How many fetch-execute cycles the CPU performs per second	GHz (gigahertz = billion cycles per second)
Cache size	How much fast on-chip storage is available to hold frequently used data/instructions	MB (megabytes)
Number of cores	How many independent processing units the CPU contains	Count (e.g. dual-core = 2, quad-core = 4)

Each of these directly affects how quickly a CPU can run programs. Exam questions may ask about any one characteristic, a combination of two, or how changing all three together affects performance.

The clock speed is the frequency at which the CPU performs its fetch-execute cycles. A higher clock speed means more cycles per second — more instructions executed per second.

Clock speed	Cycles per second
1 GHz	1,000,000,000
3.2 GHz	3,200,000,000
4.8 GHz	4,800,000,000

Effect of increasing clock speed: The CPU executes more instructions per second → programs run faster.

Limitation: Higher clock speeds generate more heat. A CPU running at 5 GHz produces significantly more heat than one at 2 GHz. Cooling hardware (heatsinks, fans) becomes larger and more expensive. Without adequate cooling, a CPU will throttle (reduce its own speed) or overheat.

(Extra context — overclocking is the practice of running a CPU faster than its rated speed; not required by OCR J277 1.1.2.)

Worked example: A 2 GHz CPU takes twice as long to execute the same program as a 4 GHz CPU (all else being equal). If the 4 GHz CPU finishes in 5 seconds, the 2 GHz CPU takes approximately 10 seconds.

Cache is fast memory built into the CPU. When the CPU needs data or an instruction, it checks the cache before going to the slower main RAM. A larger cache can hold more data — fewer trips to RAM are needed.

Effect of increasing cache size:

• More frequently used data/instructions can be held in cache
• Fewer cache misses (times the CPU must wait for RAM)
• Programs run faster, especially those that use the same data repeatedly (loops)

Limitation: Cache is expensive to manufacture. Increasing cache size increases the CPU's cost and physical size.

Worked example: A CPU running a loop that processes 1,000 values benefits greatly from a larger cache — if all 1,000 values fit in cache, no RAM accesses are needed during the loop. With a smaller cache, values are repeatedly fetched from (slower) RAM.

A core is an independent processing unit within the CPU. A dual-core CPU has two cores; a quad-core has four. Each core can execute its own fetch-execute cycle independently.

Effect of increasing number of cores:

• Multiple programs (or multiple parts of one program) can run simultaneously
• Overall throughput increases — the system feels more responsive when running several tasks at once
• Programs specifically written to use multiple cores (multithreaded programs) run significantly faster

(Extra context — multithreading is the programming technique of splitting a task into concurrent threads, each assigned to a different core; not required by OCR J277 1.1.2.)

Limitation: Not all programs benefit from multiple cores. A program that must complete each step before starting the next (sequential program) cannot split its work across cores. Adding more cores does not help it run faster.

Worked example: A web browser and a music player running simultaneously on a dual-core CPU can each use one core. On a single-core CPU, the operating system must rapidly switch between them, creating occasional pauses.

A general-purpose computer (like a laptop or desktop) can run many different programs and be used for a wide variety of tasks.

An embedded system is a computer built into a larger device to perform a specific, dedicated function. The embedded system's program (stored in ROM) is fixed and designed for exactly one purpose.

Typical characteristics of embedded systems:

• Single-purpose — designed to perform one specific task
• Low power consumption — often battery-powered or in devices left on continuously
• Real-time operation — must respond to inputs within strict time limits (e.g. a car's ABS brake controller)
• Compact — physically small, integrated into the device
• Low cost — manufactured in very high volumes
• Program stored in ROM — does not need to be changed by the end user

Examples of embedded systems:

Device	Embedded system function
Washing machine	Controls wash cycle timings, temperature, water level
Digital camera	Processes image capture, applies settings
Car engine management system	Monitors and adjusts fuel injection and timing
Smart thermostat	Reads temperature sensor, controls heating
Traffic light controller	Manages signal timing sequences
Medical pacemaker	Monitors heart rhythm, delivers electrical pulses
Microwave oven	Controls power level, timing, and display

1. Saying "faster clock speed = always better"

Higher clock speed means more cycles per second, but it also means more heat generated. At very high speeds, cooling becomes a bottleneck. Performance may not increase proportionally beyond a certain point.

2. Saying more cores always makes everything faster

More cores help when running multiple programs simultaneously or when programs are written to use multiple cores. A single sequential program with no multithreading does not run faster on 8 cores than on 1 — it still executes one step at a time.

3. Confusing cache size with RAM

Cache is fast memory on the CPU chip used to hold frequently used data. RAM is separate, much larger, and slower. Increasing cache size does not increase RAM.

4. Thinking embedded systems can run general software

An embedded system is designed for one specific task. You cannot install a different operating system or run arbitrary programs on most embedded systems — the program is fixed in ROM.

Mistake	Correction
"A larger cache means more RAM"	Cache and RAM are separate; larger cache means more fast on-chip storage, not more main memory
"More cores always makes programs run faster"	Only programs designed to run across multiple cores (multithreaded) benefit from more cores
"Embedded systems are like small PCs"	Embedded systems are single-purpose; they run fixed programs and are not general-purpose computers