Computer graphics is the long story of turning an integral nobody can solve in closed form into pixels that ship in 16 milliseconds. From Sketchpad's light pen to Phong's specular highlight, from Kajiya's rendering equation to Pixar's REYES, and from the GeForce 256 to RTX ray-tracing cores, this is how we got from vector scopes to real-time path tracing that still cheats with denoisers.
Gpu
-
The History of Computer Graphics -
Distributed Training Parallelism A map of the five ways to split a training job across GPUs — data, tensor, pipeline, expert, and sequence parallelism — when each one pays off, how they compose into 3D and 4D parallelism, and the communication-versus-memory math that decides the whole thing.
-
Training Cluster Networking: NVLink, Rails, and the Bandwidth Budget That Bounds Model Scale Why the network, not the GPU, decides how big a model you can train: the NVLink and NVSwitch scale-up domain versus the InfiniBand and RoCE scale-out fabric, rail-optimized topology, in-network reduction, and the bandwidth-budget math that maps tensor, pipeline, data, and expert parallelism onto real cables.
-
Why GPUs Beat CPUs at Matrix Math A GPU does not win at matrix multiplication because its cores are faster than a CPU's. They are slower, dumber, and clocked lower. It wins because it has thousands of them, schedules them in warps, and hides memory latency by oversubscribing the machine so aggressively that there is always work to run. Here is how SIMT, occupancy, coalescing, and tensor cores actually work.
-
HBM4 and the Memory Wall for AI Why every modern AI accelerator is pin-limited on bandwidth rather than starved for math, how High Bandwidth Memory actually works, what each generation from HBM2 to HBM4 buys, why LLM inference is a memory problem in disguise, and the three-company supply story that makes HBM the real binding constraint on AI economics.
-
Liquid-Cooling a Homelab Liquid cooling in a homelab is either the thing that finally lets you run a quiet 600 W GPU under your desk or the slow-moving disaster about to short a $4,000 rig. We walk the actual thermodynamics, AIO versus custom loops, what changes when the heat is GPU-class instead of CPU-class, the failure modes that bite 24/7 hardware, and the honest cost-benefit against better airflow.
-
Chip Export Controls: Why You Can't Buy an H100 in Shanghai How US export controls turned the AI compute supply chain into a theater of geopolitics: the TPP and performance-density thresholds, the A800/H800 workaround, ASML's EUV monopoly as the real chokepoint, China's domestic scramble, and why the rules keep flip-flopping under your feet.
-
Self-Hosted Image Generation: Flux.1, SDXL, and ComfyUI Workflows The complete self-hosted image generation stack: Flux.1 vs SDXL architecture and quality tradeoffs, hardware requirements by model tier, ComfyUI node-based workflow construction, the HTTP and WebSocket API for automation, LoRA fine-tuning, and Docker deployment behind a reverse proxy.
-
vLLM vs SGLang vs TensorRT-LLM: Inference Engine Shootout 2026 A thorough comparison of the three dominant LLM inference engines: how each one works internally, where each wins on benchmarks, and which to reach for given your hardware, workload, and operational tolerance.
-
FSDP and DDP: Distributed Training Patterns That Actually Scale The distributed-training patterns that actually scale in PyTorch — DDP versus FSDP, how each shards parameters, gradients, and optimizer state, and the collective-communication and precision knobs that decide whether more GPUs help.
-
Triton Inference Server: Production ML Serving That Actually Scales NVIDIA Triton Inference Server for production ML serving — dynamic batching, concurrent model execution, multiple backends, and the GPU-utilization and model-management problems it solves between a trained model and serving at scale.
-
GPU Passthrough End-to-End A complete guide to PCIe GPU passthrough with VFIO and KVM — IOMMU groups, binding the device, firmware quirks, NVIDIA's Code 43, and the SR-IOV and vGPU options — so you can hand a GPU to a VM and debug it when it breaks.
-
NCCL Deep Dive: Multi-GPU Collectives, Ring vs Tree, and Debugging Distributed Training How NVIDIA's NCCL moves gradients across multi-GPU and multi-node training — ring versus tree collectives, NVLink and InfiniBand topology awareness, and how to debug the silent hangs and bandwidth cliffs that take training down.
-
OpenMP: Threading Without Pthreads (And Without the Nightmares) OpenMP as the sane way to parallelize C, C++, and Fortran — pragmas instead of pthread plumbing for parallel loops, tasks, and GPU offload — plus the data-sharing and scheduling details that decide whether you get speedup or bugs.
-
Bend: Automatic Parallelism via Interaction Combinators Bend is a Python-like language that automatically parallelizes on CPU and GPU without threads or locks, powered by the HVM2 interaction combinator runtime. Here's what the novel ideas actually are, what works today, and what's still research-grade.
-
GPU Programming Without CUDA: OpenCL, ROCm, Vulkan, Metal, and WebGPU The full landscape of non-NVIDIA GPU compute: OpenCL for portability, ROCm on AMD hardware, Vulkan Compute shaders, Metal for Apple Silicon, and WebGPU for cross-platform GPU work in the browser and beyond.
-
GPU Infrastructure for ML: CUDA, MIG, Kubernetes Device Plugins, and Cost-Efficient Training A practical guide to GPU infrastructure for machine learning — CUDA fundamentals, NVIDIA MIG for multi-tenancy, sharing GPUs in Kubernetes with device plugins and time-slicing, building cost-efficient training clusters, and monitoring GPU utilization.
-
Self-Hosted AI Inference: vLLM, llama.cpp, and Running Your Own OpenAI-Compatible API A comprehensive guide to running your own LLM inference servers — covering vLLM, llama.cpp, OpenAI-compatible APIs, batching strategies, quantization, and benchmarking throughput so you can make informed hardware and software decisions.
-
Gaming on Linux in 2026: Better Than You Think A comprehensive guide to gaming on Linux in 2026 — Steam and Proton compatibility, native games, Lutris, Wine, GPU drivers, performance tuning, and the distributions best suited for gaming.
-
Local LLM Deep Dive: Ollama, Quantization, and Running AI on Your Own Hardware A comprehensive technical guide to running large language models on your own hardware — covering Ollama setup, quantization formats, hardware selection, Apple Silicon and NVIDIA GPU configuration, and building a private local coding assistant.