1.1.1 // Structure & Function of the Processor: VN, Harvard, Contemporary Processor Architecture, and Factors Affecting CPU Performance

Question 1

What are the principles of the ‘Stored Program Concept?’

Answer 1

• Programs and Data are stored in the same memory (allowing for manipulation during execution)
• Instructions are FDE’d sequentially (one at a time)

Question 2

Von Neumann Architecture is the most common implementation of the SPC.

Describe a simple layout for Von Neumann Architecture.

Answer 2

• Instructions and Data are stored in a common Main Memory with a single, simple Control Unit.
• They are transferred back and forth using a single shared bus.

(It’s the System Bus, which contains all 3 types of buses.)

Question 3

Von Neumann Architecture has many applications…

What sort of computers might contain Von Neumann Architecture?

Answer 3

1. High-Performance Computers
2. Mobiles, Tablets
3. Laptops, PCs

Question 4

The Von Neumann Architecture has it’s problems.

What issues might occur with Von Neumann Architecture?

Answer 4

1. Memory cannot be written/read from the CPU simultaneously, because of the single bus route.
2. Accidental Overwrites can happen because Data and Instructions share the same memory.

(1) Is known as the Von Neumann Bottleneck.

Question 5

The Von Neumann Architecture has it’s problems.

What’s the solution to the Von Neumann Bottleneck?

Answer 5

• Cache: Fast acting memory inside the CPU chip that holds frequently used data/instructions.
• Cache decreases the usage of RAM, increasing efficiency & speed.

Cache has it’s tradeoffs, however.

Question 6

There are more than one type of CPU Architecture…

Describe a simple layout for Harvard Architecture.

Answer 6

• Separates Data and Instructions into two distinct memories, accessible with a dedicated bus of a necessary width.
• A more complex CPU with more circuitry, as well as seperate caches.)

Question 7

Harvard Architecture has many applications…

What sort of computers might contain Harvard Architecture?

Answer 7

1. Specialist Embedded Systems
2. Digital System Processing (DSP), where speed is prioritised over complexity (Calculators, Oscilloscopes)

Question 8

What are the Advantages of a CPU with Von Neumann Architecture?

Answer 8

1. Less Circuitry in the CU, thus cheaper to manufacture.
2. Data and Instructions are accessed from memory in the same way.

Both of these advantages decrease power usage, increasing longevity.

Question 9

What are the Disadvantages of a CPU with Von Neumann Architecture?

Answer 9

1. Memory Latency & Bandwidth: Since Data & Instructions use the same memory and bus, the latency can increase due to a bottleneck cause by a demand to accessing memory.
2. Cache Inefficiency: Since Cache is handling 2 types of information, the effiency can decrease because of cache contention.

Question 10

What are the Advantages of a CPU with Harvard Architecture?

Answer 10

1. Simultaneous access to both Data and Instructions: This reduces the bottleneck as there is less competition for the same buses.
2. Better Memory Stability/Security: Data is less likely to be overwritten, reducing any accidental Overwrites.

Question 11

What are the Disadvantages of a CPU with Harvard Architecture?

Answer 11

1. Increased Manufacturing Cost: Due to the CU’s complex circuitry, as well as exclusive buses and specialised CPU components.
2. Less Flexibility: The size of Data and Instruction memory are fixed; This can be inefficient if a program needs more memory for one, and less for another.

In contrast, VNA can dynamically allocate memory to Instructions or Data

Question 12

What is Contemporary Processor Architecture?

Answer 12

Modern CPUs feature:
- Multiple cores
- Out-of-order execution
- hyper-threading, and a multi-level cache hierarchy
all designed to enhance performance, handle multiple tasks simultaneously, and efficiently manage data.

Most are 64-bit, supporting more memory and complex computations.

Question 13

What is Pipelining?

Answer 13

A technique in CPU design where instruction processing is divided into stages (like Fetch, Decode, Execute) to allow multiple instructions to be processed simultaneously at different stages, increasing overall throughput and efficiency.

Question 14

What exactly is Clock Speed, and what is it measured in?

Answer 14

• The number of FDE cycles a CPU completes in a second.
• Measured in Hz, it’s usually 2-4GHz

Question 15

How does the Clock in the CPU work, and how does it affect performance?

Answer 15

• The Clock produces signals oscillating between a low and high voltage to regulate circuits including the CPU.
• A cycle is the time period between 2 oscillations, which the CPU uses to determine the timing of operations in the computer.

Thus, the faster the cycles are (rate), the more instructions can be executed per second; Processing is faster.

Question 16

Explain what Overclocking is.

Answer 16

Overclocking increases the Clock Speed of the C/GPU above their factory standard. Although beneficial, it is not to be done carelessly:

• Higher processing rates means heat is generated more and faster, which could damage the components.
• Electricity consumption increases, affecting the electricity bill.
• Some components being pushed past their operating limits can cause corruption or crashes.

Question 17

What is Cache Size, and how does it affect CPU performance?

Answer 17

Cache exists to speed up the fetch cycle; by storing frequently used instructions inside it, the CPU saves time fetching information from RAM.

As Cache size increases, more instructions can be accessed faster. This is because the CPU is more likely to get a ‘cache hit.’ However, there are drawbacks, as latency also increases with size.

Question 18

[Cache Type] Describe all 3 levels of Cache.

Answer 18

• L1: With the lowest capacity, but the highest transfer speeds, it’s located inside the CPU chip. It’s the most likely to get a cache hit.
• L2 & L3: Usually physically situated between the RAM and CPU; They’re larger and slower than L1 and exist to ‘back up’ L1.
• Cache hits are the least likely here.
• Less frequently used data is stored here.

Question 19

Explain what cores are, and what they have with them.

Answer 19

• Cores are essentially multiple processors on a CPU chip. Each core may have it’s own ALU, CU & Registers.
• As long as software is optimised (written to support) by the OS, multiple instructions can be performed at once (Simultaneously.)
• Each core also has it’s own L1 Cache, sometimes L2, and all share a common L3.

Question 20

Do more cores necessarily mean faster CPU performance?

(Talk about Amdahl’s Law)

Answer 20

Yes, most of the time.
- Amdahl’s Law states that the potential speedup of a task through parallelism is limited by the fraction of the task that cannot be parallelised.
- This is because some tasks must be performed in order (sequentially) and so will not benefit from the multi-core processing ability.