GPU
A graphics processing unit (GPU) is a processor designed for highly parallel computation, especially graphics and other data-parallel workloads.
Web Components in Action
A GPU is a processor specialized for performing highly parallel graphics and numerical computations.
- How it is used
- Motivates GPU-backed rendering for real-time 3D and video effects when CPU or JavaScript processing may be slow.
- How deep it goes
- Explains the CPU-to-GPU shift through shaders and canvas contexts, but leaves GPU internals and extensive 3D math aside.
- Examples
- Uses a radial-fade image shader and a WebGL video component with greyscale, sepia, and edge-detection filters.
- What you can do after
- Apply it via WebGL-backed components or 3D libraries to use GPU rendering without writing every low-level shader.
GPU Programming with Triton
A GPU device is the complete graphics-processing hardware containing the streaming multiprocessors and their connected memory system.
- How it is used
- Grounds GPU kernel programming in the device hardware and its performance constraints.
- How deep it goes
- Explains throughput design, grid scheduling, and the memory hierarchy, then provides a hardware-inspection checklist.
- Examples
- Uses element-wise tensor operations and an H100 grid with 256-thread blocks to illustrate parallel work and scheduling.
- What you can do after
- Evaluate a GPU's compute resources and memory characteristics when reasoning about kernel performance.
Fast Python: High performance techniques for large datasets
- How it is used
- Positions GPUs as a platform for data-parallel analysis and AI workloads.
- How deep it goes
- Explains a simplified execution model and NVIDIA processor and cache organization, while deferring tensor cores and detailed architecture.
- Examples
- Illustrates parallel execution with array doubling and describes the RTX 2070’s processor and cache configuration.
- What you can do after
- Evaluate GPU suitability by identifying independent, similar operations and accounting for a different execution model.
Read how Fast Python: High performance techniques for large datasets explains GPU →
Deep Learning with R, Third Edition
A GPU is a massively parallel processor designed primarily for graphics computation and also widely used for numerical machine learning workloads.
- How it is used
- Positions GPUs as key deep-learning compute resources and motivates preprocessing fast enough to keep them supplied.
- How deep it goes
- Explains GPU tensor computation, model training, and CPU-side preprocessing constraints.
- Examples
- Shows CPU-side text tokenization and parallel preprocessing, and traces consumer gaming GPUs into deep-learning hardware.
- What you can do after
- Understand why tokenization stays on the CPU and how parallel preprocessing helps prevent GPUs waiting for input.
Grokking GPU Programming
A GPU is a processor designed to execute many operations concurrently, especially data-parallel operations.
- How it is used
- Frames GPUs as throughput-focused partners to CPUs for large jobs with many independent operations.
- How deep it goes
- Explains GPU origins, workload fit, and CPU-GPU roles, while deferring detailed hardware execution mechanics.
- Examples
- Uses pixel rendering, matrix multiplication, AI training, and summing millions of numbers as examples.
- What you can do after
- Evaluate whether a workload’s independent, regular operations make it a good fit for GPU parallelism.
Starting Data Analytics with Generative AI and Python
A GPU is a massively parallel processor originally designed for graphics that is also used to accelerate general-purpose numerical and machine-learning workloads.
- How it is used
- Accelerates numerical data-processing functions when parallel hardware may improve execution time.
- How deep it goes
- Demonstrates a CuPy implementation and execution-time comparison, without detailing GPU architecture or kernel internals.
- Examples
- Uses optimized and GPU-utilizing bootstrap functions, CuPy installation, and execution-time figures across different DataFrame sizes.
- What you can do after
- Apply it via CuPy to move bootstrap sampling and statistics onto GPU-backed arrays, then compare execution time with an optimized baseline.
Read how Starting Data Analytics with Generative AI and Python explains GPU →
Inside Deep Learning: Math, Algorithms, Models
A graphics processing unit used as a parallel compute device for computation.
- How it is used
- Accelerates PyTorch tensor calculations for neural-network workloads and provides compute for deep-learning experiments.
- How deep it goes
- Explains tensor acceleration, transfer overhead, device mismatches, and benchmarking, but does not detail GPU hardware internals.
- Examples
- Demonstrates matrix-multiplication timing, CUDA availability, device-mismatch errors, and moving nested objects to a device.
- What you can do after
- Apply it via PyTorch by checking CUDA, moving tensors to a device, and benchmarking workloads against CPU execution.
Read how Inside Deep Learning: Math, Algorithms, Models explains GPU →
Graph Neural Networks in Action
A graphics processing unit (GPU) is a massively parallel processor designed to accelerate graphics, numerical, and machine-learning computations.
- How it is used
- Accelerates large-scale GNN training through parallel computation in PyTorch and PyG workflows.
- How deep it goes
- Explains parallelism, memory constraints, multi-GPU distribution, and trade-offs, but does not detail GPU architecture.
- Examples
- Uses NVIDIA Tesla T4, A100, and V100 GPUs available through Colab as training examples.
- What you can do after
- Evaluate GPU suitability, memory needs, cost trade-offs, and multi-GPU scaling for GNN training.
The Complete Obsolete Guide to Generative AI
A massively parallel processor originally designed for graphics workloads and widely used for numerical machine-learning computation.
- How it is used
- Supports training and running LLMs by supplying compute capacity for AI workloads.
- How deep it goes
- Explains GPU acceleration through parallel AI operations and hardware-selection advice without detailing GPU architecture or setup.
- Examples
- Uses consumer computers choosing CPU-only models or rented cloud GPUs as the concrete hardware scenario.
- What you can do after
- Operate/configure it by matching an LLM to CPU-only local hardware or rented cloud GPU capacity.
Read how The Complete Obsolete Guide to Generative AI explains GPU →
CUDA for LLMs
A GPU is a massively parallel processor designed to perform graphics and general-purpose computations.
- How it is used
- Positions GPUs as throughput-oriented hardware for deep-learning workloads with many similar operations.
- How deep it goes
- Explains host-device roles, CUDA threads, and memory hierarchy, while deferring deeper streaming-multiprocessor architecture.
- Examples
- Uses million-element ReLU and 1,000-token self-attention to illustrate parallel workloads.
- What you can do after
- Apply CUDA's workflow to allocate device memory, transfer data, launch a kernel, and copy results back.
Sutskever's List
A graphics processing unit (GPU) is a massively parallel processor designed for graphics and other computations involving many similar operations.
- How it is used
- Scales neural-network training through parallel computation and can accelerate inference with reduced-precision arithmetic.
- How deep it goes
- Explains model and data parallelism, memory limits, and communication trade-offs without detailing GPU hardware internals.
- Examples
- Compares AlexNet’s two-GPU model split with Deep Speech 2’s multi-GPU batch processing and custom all-reduce.
- What you can do after
- Evaluate how memory limits and GPU communication affect the benefits of scaling neural-network training.
Bayesian Optimization in Action
A graphics processing unit is a hardware processor optimized for highly parallel numerical operations such as matrix multiplication.
- How it is used
- Accelerates parallel model training, batch optimization queries, and matrix-heavy Gaussian-process training.
- How deep it goes
- Defines GPU parallel matrix multiplication and demonstrates practical object transfer, without detailing GPU hardware internals.
- Examples
- Uses parallel neural-network training, batch BayesOpt, and CUDA-based Gaussian-process training as practical examples.
- What you can do after
- Apply it via PyTorch and GPyTorch by transferring data, models, and likelihoods to the GPU for training.
Deep Learning with Python, Third Edition
A graphics processing unit used here as accelerated hardware for parallel computation in deep learning workloads.
- How it is used
- Accelerates tensor computations and supports distributed deep-learning training across multiple devices.
- How deep it goes
- Introduces GPU execution and scaling context, but stops short of configuration procedures or hardware internals.
- Examples
- Uses CUDA, consumer GPUs, single-GPU training, and multi-GPU training as concrete reference points.
- What you can do after
- Evaluate/understand how GPUs fit into framework-based model training and scaling decisions.
Read how Deep Learning with Python, Third Edition explains GPU →
Machine Learning for Tabular Data: XGBoost, Deep Learning, and AI
A GPU is a massively parallel processor designed to execute many computational operations concurrently.
- How it is used
- Speeds gradient-boosted decision-tree training when datasets contain many cases or features.
- How deep it goes
- Explains acceleration, configuration, hardware requirements, and GPU tree growth, but stops short of detailed GPU architecture.
- Examples
- Uses XGBoost device settings, CUDA device ordinals, GPU prediction, and SHAP contribution calculations.
- What you can do after
- Operate/configure GPU-accelerated XGBoost training and prediction while checking supported hardware and CUDA requirements.
Read how Machine Learning for Tabular Data: XGBoost, Deep Learning, and AI explains GPU →
Hugging Face in Action
A graphics processing unit is a massively parallel processor designed to accelerate computation, including machine learning workloads.
- How it is used
- Accelerates PyTorch training and inference and runs Transformers pipelines on CUDA-compliant NVIDIA GPUs.
- How deep it goes
- Explains CUDA integration, memory movement, device checks, monitoring, and multi-GPU model parallelism without deriving GPU internals.
- Examples
- Demonstrates CUDA availability checks, GPU property inspection, GPUtil monitoring, and sentiment classification on a CUDA pipeline.
- What you can do after
- Operate/configure PyTorch and Transformers workflows to detect GPUs, inspect devices, and place inference on CUDA.
LLMs in Production
A GPU is a processor designed for highly parallel numerical computation, commonly used to accelerate machine-learning workloads.
- How it is used
- Accelerates LLM training and inference when CPU-only execution cannot meet latency demands.
- How deep it goes
- Explains parallel processing and numerical benefits, then covers memory, provisioning, containers, and metrics; stops short of hardware internals.
- Examples
- Uses regional GPU exhaustion, CPU-versus-GPU deployment, and a GPT-3-to-GPT-4 migration with increased timeout errors.
- What you can do after
- Operate/configure GPU-backed LLM systems while anticipating scarcity, workspace failures, container changes, and timeout trade-offs.
Latency
A GPU is a processor designed with many parallel execution resources for graphics and general-purpose data-parallel computation.
- How it is used
- Positions GPUs as hardware for massive parallelism, especially data-parallel workloads applying similar calculations across datasets.
- How deep it goes
- Defines GPU parallel execution and contrasts it with concurrency, but stops short of implementation details or performance measurement.
- Examples
- Uses interleaved single-core threads, multicore execution, and tens of thousands of GPU hardware threads as comparison examples.
- What you can do after
- Evaluate whether a workload fits GPU execution by identifying repeated calculations across independent datasets.
Systems Programming with Zig
A graphics processing unit is a processor specialized for highly parallel graphics and compute operations.
- How it is used
- Accelerates graphics with specialized hardware and uses parallel processing to manipulate geometry.
- How deep it goes
- Traces rasterization from draw call through vertex processing, fragment generation, shading, and image assembly.
- Examples
- Uses floating-point vertices, draw calls, vertex transformations, and per-pixel color shaders as pipeline examples.
- What you can do after
- Understand how geometry moves through a rasterization pipeline to become an image.
Machines that Think
A graphics processing unit (GPU) is a parallel processor designed for graphics workloads and widely used for numerical and machine-learning computation.
- How it is used
- Supplies parallel compute for training and running modern AI models.
- How deep it goes
- Explains parallel suitability, software support, flexibility tradeoffs, and supply constraints without detailing chip implementation.
- Examples
- Uses image rendering, AI workloads, NVIDIA’s CUDA platform, and tensor processing units to illustrate GPU roles and tradeoffs.
- What you can do after
- Understand why GPUs suit AI workloads and how flexibility and supply constraints shape their use.
Also covered in
- Software Testing with Generative AI
- Grokking Concurrency
- Build an Advanced RAG Application (From Scratch)In early access (MEAP)
- Vibe EngineeringIn early access (MEAP)
- Introduction to Generative AI, Second Edition
- Fabulous Adventures in Data Structures and Algorithms
- Kubernetes for Developers
- Time Series Forecasting Using Foundation Models
- HTML5 in Action
- GitHub Actions in Action: Continuous integration and delivery for DevOps
- Transformers in Action
What to know first
- language modelA language model is a computational model that represents patterns in linguistic sequences to assign probabilities, predict, generate, or evaluate language.
- CPUA central processing unit (CPU) is a computer processor that executes program instructions and performs general-purpose computational operations.
- CUDACUDA is NVIDIA’s parallel computing platform and programming model for performing general-purpose computations on NVIDIA GPUs.
- deep learningDeep learning is a machine-learning approach that uses neural networks with multiple computational layers to learn representations from data.
Related concepts
- CPUcontrasts with · A central processing unit (CPU) is a computer processor that executes program instructions and performs general-purpose computational operations.
- batch sizeused by · Batch size is the number of items grouped for processing together in a single batch or computational operation.
- acceleratorkind of · A hardware accelerator is specialized computing hardware designed to perform particular classes of computation more efficiently than a general-purpose processor.
- L4 GPUincludes · The NVIDIA L4 GPU is a data-center accelerator designed for energy-efficient artificial-intelligence inference, graphics, video processing, and other parallel workloads.
- shaderused by · A shader is a program executed by graphics hardware to process graphics data or compute visual properties during rendering.
- streaming multiprocessorhas part · A streaming multiprocessor is a GPU execution unit that schedules and executes groups of threads using its processing and on-chip memory resources.
Questions about GPU
What is GPU?
A graphics processing unit (GPU) is a processor designed for highly parallel computation, especially graphics and other data-parallel workloads.
Which Manning books explain GPU?
Web Components in Action, GPU Programming with Triton, Fast Python: High performance techniques for large datasets, Deep Learning with R, Third Edition, Grokking GPU Programming, Starting Data Analytics with Generative AI and Python, Inside Deep Learning: Math, Algorithms, Models, Graph Neural Networks in Action and 11 more.
What should I know before learning GPU?
Manning books that teach GPU usually introduce language model, CPU, CUDA and deep learning first.
What is the difference between GPU and CPU?
A graphics processing unit (GPU) is a processor designed for highly parallel computation, especially graphics and other data-parallel workloads. A central processing unit (CPU) is a computer processor that executes program instructions and performs general-purpose computational operations.
What kinds of GPU are there?
Kinds of GPU include L4 GPU.
What is GPU a kind of?
GPU is a kind of accelerator. A hardware accelerator is specialized computing hardware designed to perform particular classes of computation more efficiently than a general-purpose processor.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.


















