Reproducibility data for LLM inference benchmarks run on AMD Strix Halo hardware with 128 GB unified memory. Raw logs from llama-bench and llama-server outputs are provided as a stable citation target for papers and technical discussions. The dataset was authored by NorthstarAurora and last updated on May 25, 2026.
Use Cases
- Benchmarking LLM inference speed and memory usage based on raw performance logs
- Reproducing hardware performance claims for AMD gfx1151 architecture
- Comparing inference efficiency across different hardware configurations based on documented benchmark runs
- Citing specific performance numbers in technical papers and blog posts
Strengths
- Serves as a stable citation target for technical publications referencing Strix Halo inference numbers
- Benchmarks are documented as part of a production run linked to a finetune guide
- Data is associated with specific hardware: AMD Strix Halo (Ryzen AI MAX+ 395 / Radeon 8060S iGPU / gfx1151)
Limitations
- Column-level documentation is absent; field semantics must be inferred after download
- Row count is unknown, which may limit suitability assessment
- Description metadata is limited; actual data quality requires manual inspection after download
Provenance
- Source
- NorthstarAurora on Hugging Face
- Collection Method
- Raw llama-bench and llama-server outputs from production benchmark runs
- Time Range
- null
- Freshness
- Last updated 2026-05-25 22:46:28; freshness should be verified
- Geography
- null