The memory hierarchy: registers, cache, main memory and storage

The levels

From fastest, smallest and most expensive per byte, to slowest, largest and cheapest:

Typical sizes vary between processors; the order is what matters.

How it operates

When the CPU needs an item it checks the fastest level first. If the item is there (a hit) it is used straight away. If not (a miss), the next level down is checked, and a copy is brought up into the faster levels so that it will be close next time.

This works because programs tend to reuse the same items (loops, frequently called functions) and to use items stored near each other (arrays, sequential instructions).

Benefits and drawbacks

Benefits: performance close to that of the fastest memory, at a cost close to that of the cheapest; large non-volatile capacity.

Drawbacks: misses cost extra time; keeping copies in different levels consistent adds complexity; cache is costly and takes up space on the chip; a hierarchy only helps if data is reused.

Worked example

Explain why a computer with more level 3 cache may run a program faster even though RAM speed is unchanged.

1. More L3 cache can hold more of the program's frequently used instructions and data.
2. More requests are hits in cache rather than misses that go to RAM.
3. Cache is much faster than RAM, so the CPU waits less.

Answer: A larger cache gives a higher hit rate, so fewer slow trips to RAM.

Check yourself

1. Which level of the memory hierarchy is fastest?
   • CPU registers
   • Level 1 cache
   • Main memory
   • Secondary storage
   Show answer

   CPU registers. Registers are inside the CPU and accessed directly.

2. Which cache level is usually shared by all cores of a multi-core processor?
   • Level 1
   • Level 0
   • Level 3
   • Level 2 only on the first core
   Show answer

   Level 3. L1 (and usually L2) are per core; L3 is shared.

3. Which list is in order from smallest capacity to largest?
   • Registers, L1 cache, L2 cache, RAM, secondary storage
   • L1 cache, registers, RAM, L2 cache, secondary storage
   • Secondary storage, RAM, L3 cache, L1 cache, registers
   • RAM, registers, L1 cache, secondary storage, L2 cache
   Show answer

   Registers, L1 cache, L2 cache, RAM, secondary storage. Capacity grows as you move down the hierarchy.

4. Explain why computers do not use only the fastest type of memory.
   Show answer

   The fastest memory is very expensive per byte and physically small, so a large amount would cost too much; a hierarchy gives near top speed at lower cost.. Give the cost or size reason and say what a hierarchy achieves.

5. What happens on a cache miss?
   • The program stops with an error
   • The item is fetched from a slower level and a copy is placed in the cache
   • The CPU skips the instruction
   • The item is deleted from RAM
   Show answer

   The item is fetched from a slower level and a copy is placed in the cache. A miss means going further down the hierarchy and bringing the item closer for next time.

6. A program makes 200 memory requests and 170 are found in cache. What percentage is the hit rate?
   Show answer

   85 %. 170 / 200 x 100 = 85%.

7. Which level of the hierarchy is non-volatile?
   • Level 2 cache
   • Secondary storage
   • Main memory (RAM)
   • CPU registers
   Show answer

   Secondary storage. Only secondary storage keeps data with the power off.

Practise with instant feedback Print this lesson

Sources

• Pearson Edexcel International Advanced Subsidiary/Advanced Level in Computer Science specification (2026)

Last updated 8 November 2026. Spotted a mistake? Tell us.