Hardware catalog sources
GPU and CPU records are stored as structured catalog data. Source names and verification dates are shown where available, and unsupported or missing hardware data remains visible.
Methodology
How AIPCFit turns hardware data into guidance.
AIPCFit combines hardware catalog data, verified compatibility rules, and conservative planning heuristics. Official requirements stay separate from AIPCFit estimates.
Compatibility rules
Tool compatibility is evaluated with deterministic rules for known operating system, CPU, GPU, RAM, VRAM, and accelerator requirements. Unknown inputs do not become supported by assumption.
Model and workflow fit
Ollama and LM Studio model fit, plus ComfyUI workflow fit, use cataloged model or workflow data with planning headroom. These are practical heuristics, not official vendor minimums.
File size is not a VRAM requirement
Model file size is treated as one memory signal. It is not presented as an official VRAM requirement because runtime behavior, context, offloading, precision, and workflow settings can change memory use.
PyTorch is split into hardware and runtime checks
AIPCFit can verify a Windows and NVIDIA CUDA hardware path for PyTorch. That is separate from checking whether the local Python, PyTorch build, CUDA availability, device, and driver are configured correctly.
No fabricated performance
AIPCFit does not invent benchmark speeds, tokens per second, image generation times, or quality rankings. If evidence is missing, the result says so.
Inspect the evidence.
The homepage Evidence & trust section links to the current source records used by the compatibility and fit engines.