GPUs and their uses (including those not related to graphics)

GPUs and their uses (including those not related to graphics).

1.1.2 Types of processor

Types of processor — Note 2 of 3

Candidates need to understand the purpose of GPUs and what applications they are used for (candidates need to understand how GPUs are used to aid graphics, but also other applications for example their use in modeling, data mining, etc.). Candidates should understand the benefits and using GPUs and why they are suited to certain tasks (specialist instructions, multiple cores and SIMD processing).

What you need to know

A Graphics Processing Unit (GPU) is a co-processor with a very large number of simple cores designed to carry out the same operation on many pieces of data at the same time. This makes it ideal for graphics, but also for any task that involves repeating the same calculation on large amounts of data.

What to know

Graphics use: calculating the colour and position of millions of pixels or polygons each frame, where each calculation is independent.
Non-graphics uses: machine learning and training neural networks, scientific modelling and simulation, data mining, video encoding, cryptocurrency mining, password cracking.

Key terms

Exam tipCommon mistake

GPU: A specialised processor containing hundreds or thousands of cores, designed to process large blocks of data in parallel.
GPUs have specialist instructions for maths operations common in graphics, such as matrix and vector calculations.
GPUs are less suited to tasks with lots of branching or sequential steps, where a CPU's fewer, more powerful cores perform better.
Exam tip: Explain WHY a GPU suits a task: the task involves the SAME operation on LARGE amounts of independent data, which can be split across many cores.
Common mistake: Saying a GPU replaces the CPU. A GPU is a co-processor; the CPU still runs the operating system and controls the system.
SIMDSingle Instruction, Multiple Data: One instruction is applied to many data items at the same time. GPUs use this approach.