BMAX tests Intel Core Ultra mini PCs for local AI workloads
BMAX says its Intel Core Ultra mini PCs can run local LLMs and image-generation models offline while staying within a 28W to 45W power envelope. The company’s benchmarks point to practical use for private coding, document summarization, and generative graphics without cloud costs or data-transfer risk.
Why it matters: - Local AI on compact PCs can reduce cloud subscription costs, network latency, and exposure of proprietary code or confidential data. - BMAX positions Intel Core Ultra mini PCs as self-contained systems for on-device inference, not as training servers. - The company says the machines can support developers, analysts, and enterprises that want data sovereignty and offline workflows.
What happened: - BMAX published a performance review of its AI mini PC lineup built around Intel Core Ultra hardware. - The evaluation covered quantized large language models, diffusion image generation, and thermal-acoustic behavior under sustained workloads. - The tested systems included the BMAX B12 Power and B11 Pro, with mention of the MaxMini B11 Pro and MaxMini B9 Power in the product lineup. - BMAX pointed readers to its official portal for enterprise AI hardware specifications, OpenVINO deployment guides, and evaluation units: the official portal.
The details: - The systems use Intel Meteor Lake architecture on Intel 4 process technology with 3D Foveros disaggregated packaging. - The design splits the chip into CPU, GPU, SoC, and I/O tiles connected through high-density micro-bumps. - The CPU tile is described as supporting up to 16 physical cores and 22 execution threads, including 6 P-cores, 8 E-cores, and 2 low-power E-cores. - The Intel AI Boost NPU delivers 11 TOPS of INT8 performance with sparsity support. - The Intel Arc integrated GPU uses 8 Xe-cores and XMX matrix engines for FP16 and INT8 tensor work. - The memory configuration goes up to 24GB of LPDDR5 unified memory at 6400 MT/s. - BMAX says the platform supports host CPU orchestration for prompt processing, Python environments, PyTorch and ONNX graph parsing, and token scheduling. - The company says the NPU can handle continuous background AI tasks such as audio noise suppression, eye-contact tracking, and local vector embeddings. - BMAX says the NPU reduced video conferencing AI power consumption by 38% and operated at about 12W system power. - The GPU delivered up to 4x faster local image generation than legacy integrated graphics, according to BMAX. - The benchmark stack used Intel Distribution of OpenVINO 2024.x and IPEX-LLM under Windows 11 Pro 64-bit. - Llama-3-8B-Instruct with INT4 weight-only quantization used about 5.6GB of system RAM/VRAM, reached a 420 ms time to first token, and generated 8.5 to 11.2 tokens per second. - Mistral-7B-Instruct-v0.3 with INT4 quantization used 4.8GB and produced 9.8 to 12.5 tokens per second. - In Stable Diffusion 1.5 testing at 512x512 resolution, the system generated one image in 3.2 seconds, which BMAX says is a 4x speedup over CPU-only execution. - In SDXL 1.0 testing at 1024x1024 resolution, image generation took 14.8 seconds and used a peak shared memory footprint of 7.2GB. - The cooling design, called Space Capsule, uses dual copper heatpipes, copper fin arrays, and a centrifugal turbofan with fluid-dynamic bearings. - Thermal testing took place in an ISO-compliant hemi-anechoic chamber at 25°C with a 20 dB(A) noise floor. - Idle and office multitasking produced 26.5 dB(A) at about 1200 RPM and a 39°C package temperature. - A 30-minute full-load test running Llama-3-8B token generation and SDXL synthesis held the system at 33.2 dB(A) at 30cm. - Core temperatures stabilized between 72°C and 76°C with zero thermal throttling, according to the test results. - BMAX lists the B12 Power with an Intel Core Ultra 9 Processor 185H, 24GB LPDDR5 memory, a 1TB PCIe 4.0 NVMe SSD, WiFi 6, Bluetooth 5.2, and native triple-display output. - The MaxMini B11 Pro uses an Intel Core Ultra 5 115U with 16GB LPDDR5 memory and a 512GB SSD. - The MaxMini B9 Power uses an Intel Core i9-12900H with 24GB LPDDR5 memory and a 1TB SSD. - All of the listed BMAX AI mini PCs include triple 4K display support, USB Type-C, USB 3.2, WiFi 6, Bluetooth 5.2, and Gigabit Ethernet.
Between the lines: - BMAX is framing compact PCs as a practical middle ground between cloud AI and high-power workstation hardware. - The benchmarks emphasize responsiveness, silence, and offline privacy more than raw model-training scale. - The product pitch targets teams that need local inference for coding, summarization, and image generation without ongoing API fees.
What's next: - BMAX says interested customers can request engineering consultations and commercial deployment support. - The company is also steering buyers toward enterprise specification review and evaluation-unit testing before broader rollout.
The bottom line: - BMAX is betting that Intel Core Ultra mini PCs can make local AI good enough for everyday professional use while staying quiet, compact, and power efficient.
Disclaimer: This article was produced by AGP Wire with the assistance of artificial intelligence based on original source content and has been refined to improve clarity, structure, and readability. This content is provided on an “as is” basis. While care has been taken in its preparation, it may contain inaccuracies or omissions, and readers should consult the original source and independently verify key information where appropriate. This content is for informational purposes only and does not constitute legal, financial, investment, or other professional advice.
Sign up for:
The Consumer News Network
The daily local news briefing you can trust. Every day. Subscribe now.
Check Your Email!
We sent a one-time activation link to: .
Confirm it's you by clicking the email link.
If the email is not in your inbox, check spam or try again.
Welcome back!
is already signed up. Check your inbox for updates.