How Quant
Local AI Benchmarks, Hardware & Inference Guides
Exact commands. Measured results. No filler.
Local AI & Compute
The one-page answer for Strix Halo: which backend to build, when MTP is worth it, where to start.
ROCm 7.14 on Strix HaloInstall the ROCm userspace stack on Ryzen AI Max+ while keeping the kernel amdgpu driver.
Vulkan / RADV on Strix HaloSet up the Mesa RADV compute path from the Kisak PPA and verify it actually sees the iGPU.
Run llama.cpp on Strix HaloWhich backend to build, device selection, memory behavior, and the MTP speedup — from a tested run.
ROCm vs Vulkan benchmarkDense + MoE models on llama-bench and a realistic MTP coding workload. Raw data included.
RAM, GTT and GPU memoryWhat the BIOS “VRAM” number actually is, what the kernel really gives the iGPU, and how to watch it.
GPU model-fit calculatorWill this quantized model fit on that GPU at that context length? Check before you download.
Recently tested
- Check Your GPU's Real PCIe Bandwidth on Linux
See what PCIe link width and generation your GPU actually negotiated (lspci vs nvidia-smi), what the numbers should be, and when a downgrade actually matters.
- RTX PRO 5000 Blackwell — llama.cpp Benchmarks
Prompt processing and token generation speeds for popular GGUF models on dual RTX PRO 5000 Blackwell, with full reproduction setup.
- PCIe 5.0 x8/x8 Isn't Enough: Testing GPU P2P Bandwidth on Z890
Two CPU-attached Gen5 x8 GPUs that move only 5.3 GB/s GPU-to-GPU in one direction and 2.0 GB/s in the other — measured with a custom probe and NVIDIA's p2pBandwidthLatencyTest, and why PCIe generation and width alone mislead.
- Run MiniMax H3 Locally with Diffusers
MiniMax H3 generates video and audio together and needs two ~62 GB BF16 components. Run it on one high-memory GPU or split it across two 48 GB cards — Modular Diffusers, no ComfyUI.
- llama.cpp Common Options
The llama.cpp flags you will actually use — GPU offloading, context and KV cache, performance, and sampling — grouped and one line each.
- Install ROCm 7.14 on AMD Strix Halo / Ryzen AI Max+
Install and verify ROCm 7.14 with gfx1151 support on a Strix Halo (Ryzen AI Max+) laptop — repository, the amdrocm-core-dev package set, permissions, and rocminfo/HIP/llama.cpp verification.
ToolsAll tools →
Calculate whether an LLM fits in GPU memory. Uses exact GGUF artifact sizes for preset models, parameter-count estimates with provenance labels otherwise, and includes measured GTT usage from the How Quant test machine.
Video Resolution CalculatorFind video and image generation sizes divisible by your model's required multiple (8, 16, 32 or 64) for a target aspect ratio — or check whether a resolution you already have divides cleanly.
Browse by category
Local LLMs, CUDA, inference, image and video generation, TTS.
Software & DevelopmentProgramming, frameworks, libraries, debugging, Windows/WPF.
Linux & NetworkingLinux configuration, NetworkManager, Thunderbolt networking, system administration.
HardwareGPUs, storage, memory, PCIe, power, benchmarks.