Furball Factory
Practical answers on personal AI machines: choosing, building and upgrading one, running open-weight models on it, and fixing the thermal, power and driver problems that come with it.
Furball Factory is TechnoLynx's section for questions beyond our consulting work, written by the same engineers to the same technical standard. We compare hardware and approaches on equal terms. We do not endorse or sell third-party products.
Furball Factory Articles
- Lambda Vector One in XR Perception Pipelines: How It Works in Practice
- Vicuna-7B Explained: What It Is and Where It Fits (or Fails) in GenAI Projects
- 32B Models Explained: When a 32B LLM Fits Your GenAI Project (and When It Fails)
- How to Benchmark Your PC for AI: A Practical Protocol
- Why GPU Performance Is Not a Single Number — and What to Evaluate Instead of 'Best GPU for AI'
- Geekbench for AI Workloads: What It Measures and What It Misses
- Laptop GPU for AI: What Benchmarks Miss About Mobile Graphics Performance
- NVIDIA vs AMD GPU Performance: Why Software Stack Matters More Than Spec Sheets
- NVIDIA Linux Driver Installation: Correct Steps for AI Workloads
- Mac System Performance Testing for AI: Apple Silicon and Framework Constraints
- Best NVIDIA Driver for RTX 3090 and AI Workloads: Selection Criteria
- Thermal Throttling Meaning: Designed Behavior, Not Hardware Fault
- GPU Stress Test Linux Tools: What They Prove for AI Hardware
- AMD vs NVIDIA for AI Inference: When the Cost-Per-Inference Calculus Shifts
- The Three Reasons Why GPUs Didn't Work Out for You
- Intel Arc Linux Support: Driver Stack, oneAPI/SYCL, and What Works Today
- Intel and DeepSeek: What Running DeepSeek on Intel Hardware Means in Practice
- Lambda Labs GPU Workstations for XR Rendering: What They Handle and Where They Fit
- Lambda Vector Workstation for XR Pilots: What the Hardware Delivers Under Sustained Load
- Ollama Benchmark: How to Measure Local LLM Inference and Read the Numbers
- How to Check TensorFlow GPU Detection and Diagnose Common Failures
- Best Low-Profile GPUs for AI Inference: What Fits in Constrained Systems
- Low Cost GPU for AI Inference: When Cheaper Hardware Costs More
- What a 128GB GPU Means in Practice: Memory Capacity vs the Real Bottleneck
- A100 Workstation: What It Means for Portable GPU Code in Practice
- CPPC on AMD: What Collaborative Processor Performance Control Means for AI Workload Porting
- Ollama Benchmark: Measuring Local LLM Serving Before It Hits Production
- Ollama Benchmarks: Turning Local LLM Numbers into Regression Baselines
- llama.cpp Benchmark: How to Measure CPU/GPU Inference Performance in Practice
- llama.cpp Benchmark: Reading the Numbers Before You Port an LLM
- Running DeepSeek on Intel Hardware: What Profiling Tells You About Inference Cost
- Accelerate + DeepSpeed: Scaling Multimodal Training Without Rewriting Your Loop
- HIP vs CUDA: Moving a CUDA GPU Codebase Without a Rewrite
- DeepSeek R1 Benchmark: Reading Latency and Throughput for Real-Time GenAI
- NVIDIA DGX Spark Benchmarks: Reading Them as a Drift-Monitoring Baseline
- GPUs Are Part of a Larger System
- Power, Thermals, and the Hidden Governors of Performance
- How to Benchmark Your PC for AI: A Methodology That Goes Beyond Single Scores
- Evaluating AI Compute Candidates: Benchmarks, Pilots and Fair Comparisons
- DeepSeek on H100: What the Model-Hardware Pairing Means for Retail CV Cost
- How to Profile GPU Kernels to Find the Real Bottleneck
- OpenGL vs CUDA: Graphics API or Compute API for GPU Work?
- ACPI SRAT, L3 Cache as a NUMA Domain, and Why It Matters for GPU Data Movement
- GPU Parallel Computing Explained: How Thousands of Cores Solve Problems Differently
- Measuring GPU Benchmarks for AI
- GPU Coding Program: What an Inference-Focused Curriculum Actually Teaches in 2026
- Unified Virtual Memory Explained: How UVM Works for GPU Inference
- CPU vs CUDA: What the Comparison Actually Means in Practice
- CUDA Cores vs Stream Processors: What the Counts Actually Tell You
- Stream Processor vs CUDA Core: What the Spec Sheet Numbers Actually Mean
- Stream Processors vs CUDA Cores: What the Spec Sheet Numbers Actually Mean
- DirectCompute vs Vulkan Compute: Measuring What a Port Costs
- Diagnosing GPU Bottlenecks with Kernel Profiling, Not Vendor Specs
- Benchmarking Transcoding Cost and Quality for Launch-Day Capacity Planning
- Which LLM Has the Largest Context Window — and Why That Number Won't Decide It
- Imagination GPU Architecture: What It Means for Quant Compute Pipelines
Where to Go Next
If your question is how a specific machine performs on AI work, check the published results and the free Personal Edition on LynxBenchAI. If you are weighing GPU work for a team or a product rather than a personal machine, see GPU Performance Engineering.