Your AI Destination

Pushing the limits of AI requires the right hardware. At Micro Center, you’ll find everything from Blackwell-based GPUs to high-performance systems capable of running 405B-parameter large language models locally. Whether you’re driving open-source research and development or building next-generation AI applications, we’re the retailer that supplies innovators, researchers, and enthusiasts with the tools to achieve cutting-edge artificial intelligence performance.

Jump to

NVIDIA DGX Spark

A Grace Blackwell AI supercomputer on your desk.

Where to Buy

Find the best purchase option for you on the NVIDIA Marketplace.

Reserve Now

Datasheet Specifications

Features

Enhance Your Devices with AI

  • Next Generation Chips

    • Beyond NPU and iGPU, CPUs now share AI workloads
    • More PCIe lanes and memory channels improve data flow
    • Higher core and thread counts boost AI inferencing
    • Built to scale with future AI applications

  • GPU/ VRAM

    • Powers generative image models like Stable Diffusion from text promp
    • More VRAM handles larger datasets and longer context windows
    • GPUs speed up training and inferencing for AI workflows
    • NVIDIA Blackwell architecture delivers efficiency and scale

  • System Memory

    • Rule of thumb: 2.5× your GPU’s VRAM
    • More memory means smoother AI performance
    • Speed and bandwidth directly impact results
    • Low latency keeps inferencing responsive
    • Extra capacity supports multitasking and larger models

  • OS/ Software

    • Windows supports a wide range of tools and workflows
    • NVIDIA DGX delivers a turnkey AI ecosystem
    • ROCm provides open-source GPU acceleration on AMD hardware
    • NVIDIA CUDA stack powers GPU compute and integrated graphics
    • Open-source frameworks ensure flexibility and community support

  • Storage

    • PCIe Gen 5 SSDs deliver top-tier throughput and responsiveness
    • Storage needs scale with LLM size and dataset volume
    • NVMe drives reduce latency and speed up inferencing
    • RAID or multi-drive setups can separate datasets from OS for efficiency and reliability

  • Use Case / LLM Selection

    • Fine-tuning for domain-specific tasks
    • Development and testing across frameworks
    • Run open LLMs like GPT, Llama, DeepSeek, Qwen, and Mistral
    • Emerging use cases include multi-agent systems and RAG

About AI Pro Workstations

AI Pro workstations span a wide range of form factors, use cases, and software certifications. At Micro Center, we help you navigate the options between mobile and desktop systems, recommending configurations that fit everything from casual inferencing to the most demanding AI workloads. The first decision is where your work will run: fixed in one location, on the go, or a hybrid of both. Mobile points to a notebook workstation, fixed performance favors a robust desktop build, and hybrid setups often pair a portable notebook with solutions like the NVIDIA DGX Spark to offload heavier AI tasks to a compact “micro” desktop.

Lenovo ThinkPad P1 Gen 6 mobile workstation laptop
Lenovo ThinkPad P1 Gen 6 Mobile Workstation 16" Laptop Computer - Black

Mobile AI Systems

  • Run LLMs locally from a notebook workstation
  • Manage AI data effectively and efficiently across locations
  • Balance portability with strong computer performance
  • Ready for edge AI and on-site inferencing in dynamic environments
Shop Mobile AI Systems
MSI Aegis RS2 AI gaming desktop PC
MSI Aegis RS2 AI C2NVP7-1443US Gaming PC

Desktop/ Hybrid AI Systems

  • Designed for fixed work environments requiring scalability
  • Support robust configurations with powerful graphics cards
  • Handle large LLMs and demanding AI workloads
  • Maximize network throughput for faster data movement
  • Scales toward cluster-level performance for enterprise AI demands
Shop Desktop / Hybrid AI Systems

Features

Inside AI

Exploded view of an AI workstation showing CPU cooling, GPUs, RAM, CPU, motherboard, power supply, cooling and drives CPU Cooling GPUs Ram CPU Cooling Drives PSU GPUs Motherboard

Compare Devices

Choosing the right hardware for training or inference depends on your model size, precision requirements, and GPU memory capacity. This comparison makes it easy to see how different system builds perform across fine-tuning and inferencing tasks. Review VRAM needs, recommended GPUs, and system tiers (Good, Better, Best) to match your project’s scale.

AI model types with recommended VRAM, GPUs and system tiers
Model Type Model Size GPU VRAM Needed Recommended GPU Token/s Good System Better System Best System
FINE TUNING
Full-fine tuning 258.92 GB1.01 TB8xH100 - SXM5 (x2) N/AN/AN/AN/A
Full-fine tuning 258.92 GB517.84 GB8xH100 - SXM5 N/AN/AN/AN/A
LoRa (2% trainable) 258.92 GB279.63 GB3x RTX 6000 Blackwell MaxQ N/AN/AN/ALenovo PX Xeonx2/3x6000BQ/1TB/8/P
LoRa (2% trainable) 129.46 GB139.82 GB3x RTX 6000 Blackwell MaxQ N/ANVIDIA DGX SPARK (FE)Dell Prec 7885 9965WXLenovo P8 9975WX/2x6000B
LoRa (2% trainable) 64.73 GB69.91 GB1x RTX 6000 Blackwell N/AHP 22 G1A Mini 395+/128/2/PNVIDIA DGX SPARK (FE)PowerSpec AI 200 9975X/6000B
dLoRa (2% trainable) 32.36 GB34.95 GB1x RTX 5000 Blackwell N/AApple M4 Max/128/2Lenovo P3 Tower G2/CU9/5000BPowerSpec AI 200 9970X/5000B
INFERENCING
float32 258.92 GB310.7GB4x RTX 6000 Blackwell MaxQ Enter ValueN/AN/ALenovo PX Xeonx2/4x6000BQ
float16/bfloat16 129.46 GB155.35GB4x RTX 6000 Blackwell Enter ValueNVIDIA DGX SPARK (FE)Dell Prec 7885 9965WXLenovo P8 9975WX/2x6000B
int8 64.73 GB77.68GB1x RTX 6000 Blackwell Enter ValueHP 22 G1A Mini 395+/128/2/PNVIDIA DGX SPARK (FE)Dell Prec 7885 9965WX
int4 32.36 GB38.84 GB1x RTX 5000 Blackwell Enter ValueApple M4 Max/128/2Lenovo P3 Tower G2 CU9/5000BPowerSpec AI 200 9970X/5000B

FINE TUNING

INFERENCING

Our Top Picks for AI Systems

Shop our top picks for AI-ready systems, hand-selected to deliver the performance you need for everything from LLM inferencing to generative AI development.

Shop All AI Pro Systems

Shop AI Pro Products

AI News

View All Articles

  • Can Your PC Run OpenAI's New GPT-OSS Large Language Models?

    These new AI models from the maker of ChatGPT are available for anyone to download and run, if you have enough RAM and processing power.

    View Full Article

  • How to Build Your Own AI Personal Assistant

    The key to customizing an AI tool to better understand you? Write it a job description.

    View Full Article

  • How to Use an AI Agent

    Operator, is available as a research preview for ChatGPT Pro subscribers. As promised, it can surf the web independently to accomplish tasks like online shopping. But how well does it work, and will it save you time?

    View Full Article

AI FAQ: Choosing the Right AI Pro System at Micro Center

  • Micro Center carries a full range of AI-ready systems and components. You can build your own or choose from complete OEM systems by HP, Dell, Lenovo, Apple, NVIDIA, and more. We are also the exclusive retail partner for the NVIDIA DGX Spark, a compact desktop companion for offloading AI workloads. For DIY builders, we stock Intel and AMD CPUs with NPUs, motherboards, RAM, storage, graphics cards, power supplies, and cases. We also offer custom build services if you prefer a system assembled for you. Our BYO and Systems experts are ready to help in-store.

  • The best system depends on the size of the LLM you plan to use, whether you will run it locally or in the cloud, and your application stack. Key guidelines include:

    • CPU: Impacts core and thread count, memory channels, and capacity.
    • GPU: Controls the size of LLMs (in billions of parameters) you can run locally. Multiple GPUs or remote GPU software can expand performance.
    • RAM: Rule of thumb is 2.5 times your GPU VRAM for LLM loading and computation.
    • OS and Software Stack: NVIDIA CUDA, AMD ROCm, or open-source tools like Hugging Face and Ollama. Linux and Ubuntu are common. Windows requires WSL, which adds overhead.
    • Storage: PCIe Gen 5 drives perform best for loading and unloading LLMs. Ensure enough space for models, apps, and instances.

  • There is no universal AI benchmark yet. Performance depends on the workload. In general, VRAM is the most important factor, followed by RAM, CPU, OS, and storage. Independent reviews and YouTube benchmarks can show how similar systems perform under different workloads. For one-on-one advice, visit our BYO, Systems, or AI associates in-store.

  • Your choice depends on your coding languages, workloads, budget, and hardware. The main options are:

    • NVIDIA: Blackwell GPUs, CUDA software, and DGX Spark. Strong ecosystem with robust support. Higher cost but widely used.
    • Integrated Graphics (AMD or Intel): CPUs with onboard GPUs, such as AMD Ryzen AI with ROCm. Lower cost, more setup complexity, and open-source focus.
    • Open Source: Flexible and community-driven. Platforms like Hugging Face, GitHub, and Ollama let you run open LLMs on Intel, AMD, or NVIDIA. Best for developers and smaller teams.

  • Micro Center updates our AI landing page, blog, and social channels with product releases, articles, and influencer reviews. For broader news, join active developer communities such as Kaggle and Anaconda, or follow AI leaders like OpenAI, Anthropic, xAI, and Google. AI is evolving quickly, from generative models to agentic AI. Staying connected helps you keep pace.