Definition
A GPU is designed with a large number of independent processors that work in parallel, making it exceptionally effective for tasks that require the same operation to be repeated on large sets of data.
Key Point
Key Characteristics of GPUs:
• Composed of thousands of smaller, simpler cores that can handle many operations simultaneously
• Optimised for tasks involving parallel processing (image rendering, video processing, deep learning)
• Act as co-processors, working alongside the CPU to offload specific workloads
• Use SIMD (Single Instruction Multiple Data) architecture
• Composed of thousands of smaller, simpler cores that can handle many operations simultaneously
• Optimised for tasks involving parallel processing (image rendering, video processing, deep learning)
• Act as co-processors, working alongside the CPU to offload specific workloads
• Use SIMD (Single Instruction Multiple Data) architecture
Example
GPU Applications:
• 3D graphics rendering for gaming
• Machine learning and neural network training
• Scientific simulations
• Cryptocurrency mining
• Video encoding/decoding
• 3D graphics rendering for gaming
• Machine learning and neural network training
• Scientific simulations
• Cryptocurrency mining
• Video encoding/decoding
SIMD Architecture
Definition
Single Instruction Multiple Data (SIMD) - An architecture where one instruction is applied to multiple data points simultaneously. This is ideal for tasks like image processing where the same operation needs to be performed on every pixel.
Exam Tip
Why GPUs excel at graphics/ML: The parallel nature of these tasks (same operation on many pixels/data points) matches the GPU's SIMD architecture perfectly. A CPU would process pixels one by one; a GPU processes hundreds simultaneously.
Co-processors
Definition
A co-processor is an extra processor that works alongside the main CPU and takes over one specialised type of task, which it carries out more efficiently, leaving the CPU free for other work.
Example
Examples of co-processors:
• A GPU carrying out graphics rendering and other calculations on large sets of data
• A floating point unit (FPU) carrying out floating point arithmetic
• A GPU carrying out graphics rendering and other calculations on large sets of data
• A floating point unit (FPU) carrying out floating point arithmetic