llm-kb
← Back to social
Community

User tests Intel Arrow Lake iGPU with Llama.cpp: SYCL works at 12 tok/s, Vulkan fails

A user tested Intel's Arrow Lake iGPU with Llama.cpp for local LLM inference. Vulkan support was broken (1 tok/s), while SYCL achieved ~12 tok/s on Qwen3.6 35B models. CPU-only inference was more consistent at 14 tok/s, suggesting the iGPU offers no practical benefit.

6 engagement·1 source·Sun, Jul 12, 2026, 08:45 PM
The user ran benchmarks on an Arrow Lake system using Llama.cpp with Vulkan and SYCL backends. Vulkan was essentially non-functional, delivering at best 1 tok/s. SYCL worked reasonably well, running Qwen3.6 35B models at around 12 tok/s with prefill at ~20 tok/s, but suffered from intermittent freezes. CPU-only inference was the most reliable, achieving 14 tok/s on the same models and 30-40 tok/s on prefill. The user concluded the iGPU is currently useless for LLM inference, as it neither improves speed nor consistency over the CPU.

Entities

Llama.cpp(tool)Intel Arrow Lake(concept)Qwen3.6 35B(model)Vulkan(tool)SYCL(tool)

Related

CommunitySun, Jul 12, 2026, 04:38 AM

User tests LLMs on Intel 285HX CPU-only mini-PC, finds Llama-Swap incompatible with SYCL

A user set up a homelab server on an MS-02 mini-PC with an Intel Core Ultra 285HX and 64GB RAM, testing Qwen3, Qwen3.6, and Gemma4 via Llama.cpp. They found Llama-Swap's quick swapping helpful but incompatible with SYCL, complicating testing. The post shares early impressions of running LLMs on a CPU-only system.

4 engagement·1 source·reddit
ProductfeaturedSun, Jul 5, 2026, 08:45 PM

Rust+CUDA LLM inference engine for RTX 5090, optimized with NVFP4, MoE, and speculative decoding

A from-scratch LLM inference engine written in Rust and CUDA, specifically tuned for the RTX 5090 Laptop GPU (Blackwell sm_120a). It supports NVFP4 quantization, mixture-of-experts (MoE), and multi-token prediction (MTP) speculative decoding, achieving up to 1.6x speedup over llama.cpp on target models. Designed for developers who need maximum inference performance on a single high-end laptop GPU.

260 engagement·1 source·github
BenchmarkSun, Jul 12, 2026, 10:47 AM

User benchmarks 4x RTX 5060 Ti with SGLang for Qwen3.6 27B, finds better concurrency than vLLM

A user shared benchmark results showing that SGLang handles higher concurrency better than vLLM when running Qwen3.6 27B (INT8 with bf16 KV cache) on a 4x RTX 5060 Ti (64GB VRAM) setup. The test achieved 200 successful requests at 8 concurrency over 348.87 seconds, processing 61,870 input tokens and generating 44,525 tokens. This provides a practical reference for others considering multi-GPU configurations with these consumer cards.

3 engagement·1 source·reddit
CommunitySat, Jul 11, 2026, 01:03 PM

User benchmarks AMD EPYC 9374F for LLM inference, finds 48-thread sweet spot

A user replaced their EPYC 9135 with a cheap 9374F (8 CCDs) for LLM inference. Initial benchmarks showed no decoding advantage until they used 48 threads; 64 or 32 threads performed worse than the 9135 in some scenarios. The 9374F is worse for gaming.

6 engagement·1 source·reddit
Tool ReleaseMon, Jul 6, 2026, 07:28 PM

Reame: CPU-first LLM inference server built on llama.cpp released

Reame is a new LLM inference server designed to run efficiently on cheap CPU hardware, including shared vCPUs, free tiers, and 2-core ARM boxes. Built on llama.cpp, it features disk KV cache, self-regulating speculation, generation archive, and interleaved multi-user support. The project emphasizes treating CPU hardware as a first-class citizen rather than a fallback.

60 engagement·1 source·github