Release 2.0.1 benchmarks CPU/iGPU/NPU vs Ollama CPU/Vulkan/CUDA/IPEX-LLM · Issue #43 · SearchSavior/OpenArc @SearchSavior I ran a series of benchmarks to compare the performance of OpenArc against Ollama using its various backends (CUDA, VULKAN, and IPEX-LLM). Benchmarking Script: To ensure a consistent ...

Gone full LLM-benchmarking-nut on laptop: iGPU/dGPU/NPU/CPU
Verdict: RTX 4070 king @ 45 t/s. OpenArc's iGPU is the dark horse 16.5 t/s & <1s TTFT! Smokes default Ollama. My year-idle Intel NPU? 9.6 t/s. Full torture report: github.com/SearchSavior...

#IntelNPU #Benchmark #LLMs #Ollama #OpenArc #AI

testing NPU · Issue #38 · SearchSavior/OpenArc This issue helps test 1.0.6 with NPU devices. To keep things consistent please use Qwen3-8B-int4-cw-ov for LLM and distil-whisper-large-v3-int8-ov for whisper. Loading models in openarc add use --d...

Personal challenge: Dusted off the year-idle **Intel NPU**.

Got **OpenArc** running on v1.0.6 and squeezed out a **benchmark**! Feels good to see that silicon finally working after judging my life choices for a year.

Proof: github.com/SearchSavior...

#IntelNPU #OpenVINO #OpenArc #Benchmark #AI

Do you know about #OpenARC, the Open University’s new Arts & Humanities Research Centre? Find out more below👇🏼
@sharpweb.org @crasshlive.bsky.social @torchoxford.bsky.social @ies-sas.bsky.social @artscouncilengland.bsky.social @beinghumanfest.bsky.social @ukri.org @britishacademy.bsky.social