Computer hardware and operation practice questions.

Which task is better suited to a GPU than to a CPU?

• running the branching logic of an operating system
• applying the same brightness change to millions of pixels
• handling a single stream of keyboard input events
• coordinating the reads and writes of one open file

Answer: applying the same brightness change to millions of pixels

A GPU contains thousands of simple cores that all execute the same operation on different data at once, which is exactly the pixel case. Branching, sequential and coordination-heavy work belongs on the CPU's few fast, flexible cores.

A laptop must start quickly and keep its files when powered off. Which component should hold the operating system?

• a solid-state drive, which is non-volatile and fast to read
• main memory, which the processor can address directly
• cache memory, which sits closest to the processor
• a magnetic tape drive, which offers the largest capacity

Answer: a solid-state drive, which is non-volatile and fast to read

The requirement combines persistence (rules out RAM and cache, which are volatile) with speed (rules out tape). At start-up the operating system is then copied from the SSD into main memory to be executed.

A processor's clock speed is doubled, but most programs barely run faster. Which hardware limit best explains this?

• the arithmetic logic unit stops performing logical operations
• instructions and data cannot be fetched from main memory fast enough
• the address bus starts carrying data values in both directions
• secondary storage devices lose their stored contents at high speed

Answer: instructions and data cannot be fetched from main memory fast enough

Processor and main memory share one path for instructions and data, and memory is far slower than the CPU — the von Neumann bottleneck. A faster clock just makes the processor wait more cycles for each fetch, which is exactly the gap cache exists to bridge.

Which list orders storage from fastest to slowest access time?

• cache, registers, solid-state drive, main memory
• main memory, cache, registers, solid-state drive
• registers, main memory, cache, solid-state drive
• registers, cache, main memory, solid-state drive

Answer: registers, cache, main memory, solid-state drive

Speed falls with distance from the processor core: registers are inside it, cache sits beside it, main memory is across the buses, and secondary storage is slowest but the only non-volatile level. The same order reversed gives capacity per cost.

A team moves a calculation from the central processing unit (CPU) to a graphics processing unit (GPU) and finds that it now runs more slowly. Which feature of the calculation best explains this?

• Each step of the calculation needs the result of the step before it.
• The same calculation is applied independently to every element of a large array.
• The calculation uses only addition and multiplication.
• The calculation reads its input values from main memory.

Answer: Each step of the calculation needs the result of the step before it.

A GPU gains its speed from very many simple cores running the same operation on different data at the same time. A strictly sequential chain of dependencies leaves almost every core idle, so only one slow core does the work, and the cost of copying the data across is added on top.

Three extra lines are added to the address bus of a processor and nothing else is changed. By what factor does the number of distinct memory locations that the processor can select change?

• It is multiplied by three.
• It is multiplied by eight.
• It is increased by three.
• It is multiplied by six.

Answer: It is multiplied by eight.

Every extra line can be either 0 or 1, so each one doubles the number of different patterns the bus can carry. Three extra lines multiply the count by 2 x 2 x 2 = 8; the growth is exponential in the number of lines, not proportional to it.