Independent experiments, source-level analysis, and practical guides to the AI running on your hardware.
Featured investigation
A controlled RTX 4090 + RTX 3060 experiment traced a severe Strata Q2_0 multi-GPU prefill slowdown to the Linux expert-copy path. Allowing full CUDA host registration raised 32K prefill from 665 to 2,110 tok/s.
RAMGPT QuantBench
Verified GGUF comparisons against pinned BF16 references. Start with the results, then inspect the method.
Selected research
Independent RTX 4090 testing finds UD-IQ3_XXS only ~2% slower than UD-Q3_K_XL at matched context, while reducing the 200K GPU memory budget from 16 to 15 GiB cuts prompt processing by 19%.
An RTX 4090 reproduction of Qwen3.8-27B at ~2.5 bpw finds 1.93% lower WikiText-2 perplexity for GSQ-RCO than Unsloth UD-IQ2_S, while tensor dumps reveal a far more heterogeneous precision allocation and slightly lower throughput.
An RTX 4090 reproduction of Qwen3.8-Flash-Next PLE paging shows a 38.4GB table reaching only 221.7MiB resident after an 8K high-diversity prompt sweep.
A cumulative learning series
Build your understanding, one idea at a time. Beginner-friendly explanations from contributors learning AI systematically.
Explore all 33 lessons ↗Continue learning
Continue learning
From understanding to practice
Learn how to estimate LLM VRAM requirements from model size, quantization, KV cache, context length, runtime overhead, and GPU offloading.
A simple step-by-step llama.cpp guide for beginners: compile from source, run a GGUF model, use GPU acceleration, and tune the main inference settings.
Learn why a local LLM runs slowly and how model size, GPU offloading, context length, memory bandwidth, CPU threads, and runtime settings affect speed.
The people behind the work
Infrastructure support
Independent AI testing can require substantial compute. We acknowledge infrastructure providers that help RAMGPT run larger experiments at practical cost.
Infrastructure relationships do not determine RAMGPT's benchmark methodology, measurements, rankings, or editorial conclusions.