battlemage

Open recipes, engine patches, and benchmark harnesses for LLM inference on Intel Arc Pro B60/B70 (Battlemage, Xe2). MoE 35B at 160 t/s decode / 7.5K t/s prefill single-stream, 27B at 50~ t/s decode / 1.7K t/s prefill single stream. vLLM XPU MTP unlocked. Muse Glimmer recipe added!!

Benchmark results and performance data for the Intel Arc Pro B70 GPU (Xe2/Battlemage) - LLM inference, video generation, dual-GPU scaling.

Resizable BAR for Intel Arc GPUs on firmware without ReBAR support — boot-time fix + diagnostics, no firmware mod (Debian .deb)

Custom fan curves + GPU power/clock tuning for Intel Arc Pro B60/B70 (Battlemage) on Linux. Confirmed on Arc Pro B60 (8086:e211).

Plug-and-play llama.cpp runtime for Intel Arc GPUs. Auto-detects your card, picks safe SYCL defaults, and exposes an OpenAI-compatible API.

Run LLMs on Intel Arc (B60, B580, B570, B50, B70, A770) with llama.cpp Vulkan: benchmarks, tuning, agent reliability evals, raw data.

C++/SYCL local LLM inference for Intel Arc: Qwen3.8, DeepSeek-V4.1, MiMo-V2.6, GLM-5.3 and LTX-2.5 text-to-video on Arc Pro B70 cards. XMX kernels, quantized models, multi-GPU.

videogen-b70: ComfyUI + PyTorch XPU recipes for MiniMax H3 T2V on Intel Arc Pro B70. Auto-detects one or two cards (CLIP on GPU1, UNet on GPU0). Not tensor parallel.

Field-tested guide: multi-GPU vLLM tensor-parallel (TP=2/TP=4) on Intel Arc Pro B70 (Battlemage BMG-G31, Xe2) on Linux. Driver setup (xe force_probe=e223), bare-metal vLLM + oneAPI 2025.3, the compute-runtime multi-root USM + triton-xpu init_devices fixes, FP8/int4-AutoRound quant, root-cause error reports. AI-agent readable (AGENTS.md).

Fix small BAR / 'Resizable BAR not detected' for Intel Arc Pro B60 on Dell Precision 5820 (Purley/C422): one hidden BIOS setting, MMIO High Granularity 1G -> 64G

Automatically re-enable your eGPU after hot-plugging on Windows Never manually enable your eGPU from Device Manager again! This tool monitors your external GPU and automatically enables it whenever you reconnect it after safe-removal.

Community-maintained llama.cpp SYCL Docker optimized for Intel Arc Pro B70 (Battlemage). Latest oneAPI/compute-runtime/IGC, B70-specific AOT build, all features enabled (Flash Attention, MTP, reorder kernels, no GGML_SYCL_DISABLE_OPT).

llama.ccp on Intel Arc Battlemage, that actually works

Run Krea 2 (12B DiT) on an Intel Arc B580 in Docker — working PyTorch-XPU/ComfyUI setup for Battlemage on kernel 7.x, plus the fix for the Level-Zero abort, benchmarks, and troubleshooting

Experimental Intel Xe workaround for stale Battlemage package temperatures and fan speed

Everything a new Intel Arc Pro B70 owner needs for local LLMs: setup check, which engine, known driver/engine bugs with workarounds (live status), benchmarks, tested models

Fix for Intel Arc VAAPI SIGBUS during vaInitialize on small-BAR systems (Debian 13 .deb + patch)

llama.cpp SYCL patches and tuning for Intel Arc B580 12 GB: 35B-A3B MoE at 47-50 t/s with 160-200K context, and ternary Bonsai 2 27B at 196K - on a single 12 GB card

Run MiniMax Music 3 on Intel Arc Pro B70/Battlemage GPUs with a focused local web studio