What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Most households need two roles, not one specific computer: an always-on Home Assistant host and a separate (or shared) machine that runs the local model server. Home Assistant can connect to Ollama over your network, while model fit depends on the model’s quantization, context length, concurrent requests and available GPU or unified memory. A dedicated GPU improves responsiveness in many configurations, but the documented guidance does not make one mandatory for every model.
The hardware architecture that works
Think of a local voice or text assistant as three layers. Home Assistant runs your automations, entity database and Assist interface. A model server such as Ollama loads the language model and answers prompts. Optional USB radios connect Home Assistant to Zigbee, Thread or Matter devices. These layers can share one computer, but separating them usually makes availability and troubleshooting easier.
As an Amazon Associate I earn from qualifying purchases.
Home Assistant host
Home Assistant recommends Home Assistant Green for most people and also supports a Raspberry Pi, mini PC, NUC or virtual machine. Its Raspberry Pi installation path specifies a Pi 4 or Pi 5 with at least 2 GB of RAM. That is a Home Assistant host recommendation, not a certified specification for running an LLM. Model inference has separate memory and performance requirements.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Clear out junk files and repair common Windows errors3Fix the driver behind crashes, sound loss and screen glitchesLLM host
Ollama runs on macOS, Linux and Windows as an external service. Home Assistant’s Ollama integration connects to that service over the network, so the model computer must remain powered on and reachable at the address you configure. The machine needs enough memory for the selected model, its context and any other workloads running at the same time.
#1 Best Overall
- EVOLUTION RYZEN AI MAX+ 395 MINI PC - GMKtec EVO-X2 is the next evolution in AI mini PC Ryzen Strix Halo series. Thanks to AMD Simultaneous Multithreading (SMT) the core-count is effectively doubled, to 32 threads. Ryzen AI Max+ 395 has 64 MB of L3 cache and can boost up to 5.1 GHz, depending on the workload. The Ryzen AI Max+ 395 is currently rated as the "most powerful x86 APU" on the market for AI computing.
- AI NPU with XDNA 2 ARCHITECTURE - Powered by 16 “Zen 5” CPU cores, 50+ peak AI TOPS XDNA 2 NPU and a truly massive integrated GPU driven by 40 AMD RDNA 3.5 CUs, the Ryzen AI MAX+ 395 is a transformative upgrade and delivers a significant performance boost over the competition. The Ryzen AI Max+ 395 excels in consumer AI workloads like the llama.cpp-powered application: LM Studio. Shaping up to be the must-have app for client LLM workloads, LM Studio allows users to locally run the latest language model without any technical knowledge required and unleash their creativity and productivity.
- AMD RADEON 8090S iGPU GAMING PC - The AMD Radeon RX 8060S offers all 40 CUs with up to 2.9 GHz graphics clock and uses the new RDNA 3.5 architecture. The powerful iGPU is positioned between an RTX 4060 and 4070 laptop GPU and therefore enables gaming in FHD at maximum details in most demanding games. The 8060S can also utilize the full 128GB pool, which is perfect for running LLMs such as Deepseek 70B Q8, which runs comfortably on this machine.
- EIGHT CHANNEL LPDDR5X - LPDDR5X is a new ground breaking memory small form factor installed on-board. With blazing speeds up to to 8000MT/s, it runs 1.5x faster than the DDR5 SODIMMs; 90% better performance over DDR5 SODIMMs in video conferencing and photo editing; 30% better performance in productivity apps; 12% better performance in digital content workloads.
- QUAD SCREEN 8K DISPLAY SUPPORT - EVO-X2 AI Mini PC support 4-screen 4K/8K output via HDMI 2.1 (8K@60Hz), DisplayPort 1.4 (4K@60Hz), and dual USB 4 40Gbps Transfer speed (supporting PD3.0/DP1.4/DATA). Ideal for gaming, video editing, and multitasking, it provides expansive and crisp multi-display support.
Smart-home radios are separate from compute
A USB adapter such as Home Assistant Connect ZBT-2 can provide Zigbee and Thread connectivity; Home Assistant’s FAQ also lists Matter support. It does not add CPU, GPU or model memory. Buy one only if your device network needs those protocols.
How much memory does a local LLM need?
There is no universal minimum. Start with the model and quantization you intend to run, then add headroom for context, Home Assistant descriptions, conversation history and concurrent requests. A quantized model generally requires less memory than a higher-precision build, but the exact allocation is model- and runtime-specific.
A concrete Ollama example
Ollama’s current quickstart uses Gemma 4 E2B as an example. The download is about 7.2 GB, and Ollama recommends 8 GB of available VRAM or Apple unified memory for that example. This is not a floor for every model: a larger model, different quantization or longer context can require substantially more.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Context length changes the bill
Context includes the instructions, entity descriptions, previous messages and tool results sent with a request. Ollama documents these defaults by available VRAM:
| Available VRAM | Documented Ollama context default |
|---|---|
| Below 24 GiB | 4k tokens |
| 24–48 GiB | 32k tokens |
| 48 GiB or more | 256k tokens |
Ollama says larger context consumes more memory and suggests at least 64,000 tokens for tasks such as agents and coding tools. Those are runtime defaults and recommendations, not a guarantee that every model can use those lengths effectively. If your home assistant only handles short commands, a smaller context may be a sensible trade-off.
Concurrency matters as much as model size
Ollama’s FAQ states that required RAM scales with parallel requests multiplied by context length. One household request is very different from several dashboards, automations and users querying the model simultaneously. Reserve memory for the operating system, Home Assistant if co-hosted, the model server and any vector database or monitoring process you add.
Do you need a GPU to run Ollama?
No blanket rule requires a dedicated GPU. Ollama can use system RAM when VRAM is insufficient, although it warns that responses may be slower. CPU or system-memory inference can therefore be practical for smaller models, experimentation or an existing computer you already leave on.
Rank #2
- EVOLUTION CORE ULTRA 9 285H MINI PC - GMKtec EVO-T1 is the next evolution in AI mini PC Ultra 9 series. The Core Ultra 9 285H offers 16 cores (six P-cores + eight E-cores + two LPE-cores) and 16 threads with a turbo clock of 5.4 GHz. It is currently one of the best value for performance AI mini PC computers.
- AI NPU - The 285H features an Intel AI Boost NPU, capable of up to 13 TOPS (Tera Operations per Second) for INT8 calculations, which is designed to accelerate AI tasks.
- INTEL ARC 140T GAMING PC - The Arc 140T GPU includes 8 Xe cores and supports features like DirectX 12, OpenGL 4.5, and OpenCL 3, making it capable of handling modern games and creative applications. It also supports Quick Sync Video for efficient video encoding and decoding, as well as AV1 encoding and decoding.
- 64GB DDR5 RAM + 1TB SSD - The EVO-T1 is equipped with Dual 32GB (Total 64GB) SO-DIMM DDR5 5600MHz memory sticks. 2TB PCIE 4.0 SSD Drive with 3x M.2 2280 Expansion slots. Each slot capable of reading up to 4TB. (12TB MAX)
- QUAD SCREEN 8K DISPLAY SUPPORT - EVO-T1 AI Mini PC support 4-screen 4K/8K output via HDMI 2.1 (8K@60Hz), DisplayPort 1.4 (4K@60Hz), and USB Type-C Transfer speed (supporting PD3.0/DP1.4/DATA). Ideal for gaming, video editing, and multitasking, it provides expansive and crisp multi-display support.
When a GPU is worth prioritising
- Lower latency: GPU acceleration can produce tokens faster than CPU-only execution for supported models.
- Longer context: more VRAM gives the runtime room for larger context windows without spilling into slower system memory.
- Multiple users or services: additional memory and throughput help when requests overlap.
- Predictable responsiveness: keeping the model in supported accelerator memory avoids the performance swing caused by RAM fallback.
Supported acceleration paths
Ollama documents NVIDIA GPU support subject to the required compute capability and drivers. On Apple devices it can use Metal, and it documents Vulkan support for additional Windows and Linux hardware. Check the supported software path for the exact operating system and accelerator before buying; a large specification on paper is not useful if the runtime cannot use it.
What the published benchmark does—and does not—show
Ollama reported a configuration-specific comparison using Gemma 3 12B at 128k context on one NVIDIA GeForce RTX 4090: 52.02 tokens per second before a model-scheduling change and 85.54 tokens per second after it. VRAM use was 19.9 GiB before and 21.4 GiB after. The result demonstrates that scheduling and context settings affect throughput and memory; it is not a general GPU ranking or a promise of those speeds on another machine.
Can Home Assistant and Ollama run on different computers?
Yes. This is the most straightforward arrangement for many homes. Home Assistant remains on an always-on Green, Pi, mini PC, NUC or virtual machine, while Ollama runs on a computer with the memory and acceleration required by your chosen model. The two machines communicate over your local network; Home Assistant requires the Ollama server to be reachable.
Separate hosts: best for role isolation
| Arrangement | Best fit | Main trade-offs |
|---|---|---|
| Home Assistant host plus GPU-capable LLM computer | You want dependable automation and responsive inference | Higher purchase cost, power draw and noise; you must size the model computer correctly. |
| Existing computer as LLM server plus separate Home Assistant host | You already own suitable hardware | Little additional hardware cost, but the computer must stay on and its availability and speed determine response quality. |
| One computer for both services | You prefer fewer devices and can manage services or virtualization | Model inference and automation compete for memory, CPU, storage and power. No cited source certifies a combined minimum specification. |
| CPU/system-memory inference | You prioritise reuse and accept smaller models or slower replies | It can work when VRAM is short, but Ollama warns that system-memory use may slow responses. |
One host: check contention before committing
A combined host must hold the model and context while Home Assistant handles automations, databases, add-ons and dashboards. Leave enough capacity for bursts: a voice command can arrive while a scheduled automation and another conversation are running. Virtual machines or containers can improve separation, but they do not create additional memory or GPU capacity.
How Home Assistant uses the local model
The Home Assistant Ollama integration adds a conversation agent that talks to an external Ollama server. You can allow it to use the Assist API and provide information about, or control, entities exposed to it. Control is labelled experimental. Only models that support tools can control Home Assistant, and smaller models are more likely to make mistakes.
Limit the first experiment
Home Assistant recommends exposing fewer than 25 entities while experimenting. Begin with a small, low-risk set such as a test light or temperature sensor. Do not expose locks, garage doors, alarms or other safety-critical devices until you have observed the model’s behaviour and added conventional automations for hard safety limits.
Alternative: an OpenAI-compatible server
Home Assistant’s llama.cpp integration accepts an OpenAI-compatible chat-completions API. Home Assistant lists llama.cpp, llama-cpp-python and vLLM as compatible backend examples. If you choose Ollama, Home Assistant recommends using its official Ollama integration rather than treating Ollama as a generic endpoint.
Rank #3
- 【Low Power for Always-On AI Workflows】At just 15W TDP, the GEEKOM A7 uses far less power than a traditional 350W desktop, helping reduce electricity costs, heat, and cooling noise during extended operation. That efficiency makes it ideal for keeping cloud AI assistants and AI Agent tasks running in the background—automating document summaries, email polishing, meeting notes, content rewriting, research, and scheduled workflows throughout the day. The energy savings can help recoup the device cost in about 1 year, making A7 a practical choice for 24/7 AI task hosting and efficient everyday computing.
- 【Ryzen 7 7730U – More Than a Low-Power PC】Think low power means less performance? Not here. The Ryzen 7 7730U mini computer packs 8 cores, 16 threads, and up to 4.5GHz, giving you the power to handle multitasking, dozens of tabs, video calls, and creative work smoothly. AMD Radeon Graphics supports 4K playback, multi-display work, photo editing, and casual gaming without a dedicated GPU. Compared with the Ryzen 7 5825U and Ryzen 5 7430U, it delivers up to 20% higher performance for faster response and smoother everyday computing—all in a compact, energy-efficient Mini desktop.
- 【Lock In More Memory Before It Costs More】32GB gives you the headroom most demanding tasks need today—and room to grow tomorrow. Built for heavy multitasking, content creation, large projects, and AI-assisted workloads, the GEEKOM mini pc starts you with twice the memory of a typical 16GB setup, so you can skip an immediate upgrade. With AI driving greater demand for memory, starting with 32GB is a smarter way to stay ready for what’s next. The 500GB PCIe Gen4 x4 SSD delivers fast storage, with support for up to 64GB RAM and 4TB SSD storage when you need more.
- 【Premium Metal Design & 3-Year Warranty】Why settle for plastic? The GEEKOM mini desktop features a premium aluminum alloy chassis that resists daily wear and helps dissipate heat during extended use. Rigorous quality testing and CE, FCC, and RoHS compliance support dependable performance, backed by a 3-year limited warranty and professional support for long-term peace of mind.
- 【One Mini PC, All Your Ports】Stay connected with dual USB-C ports, 5 USB 3.2 ports, dual HDMI 2.0, and a 2.5G LAN port for fast, flexible connectivity. The USB-C ports support high-speed data transfer, display output, and peripheral power, while Wi-Fi 6E keeps streaming, file transfers, and online work fast and reliable. From multiple peripherals to high-resolution displays, everything you need stays within easy reach.
A practical sizing and buying process
- Define the workload. Decide whether you need short voice commands, long conversations, tool-using agents, coding assistance or several simultaneous users. Write down an acceptable first-token and full-response delay.
- Choose the model and quantization first. Record its download size, runtime memory guidance and tool-use support. Do not choose a computer before this step.
- Estimate context. Include system instructions, Home Assistant entity names and areas, conversation history and tool results. If you plan agent-style tasks, account for Ollama’s 64,000-token suggestion rather than assuming a 4k context is enough.
- Calculate concurrency. Multiply the context requirement by the number of requests that may overlap, then add operating-system and service overhead. This is why a machine that runs one test comfortably may struggle in a busy home.
- Compare accelerator memory and support. Check available VRAM or Apple unified memory, driver requirements and whether Ollama can use the accelerator through NVIDIA, Metal or Vulkan on your platform.
- Decide on host separation. Use a separate Home Assistant host when automation uptime matters or when the model computer is noisy, power-hungry or shared with other work.
- Plan network availability. Give the model server a stable local address, keep it powered when Assist is needed and verify that Home Assistant can reach the configured port.
- Measure your actual workload. Test representative commands, long contexts and overlapping requests. Vendor benchmark figures should inform expectations, not replace testing on your model and configuration.
Common failure modes and fixes
The model loads but replies are very slow
- Check whether the runtime is spilling from VRAM into system RAM; Ollama says this can slow responses.
- Reduce context length and remove unnecessary entity descriptions or conversation history.
- Use a smaller or more aggressively quantized model, or move inference to a host with more supported accelerator memory.
- Check drivers and the documented backend path for your GPU or integrated accelerator.
Home Assistant cannot connect to Ollama
- Confirm Ollama is running as a server, not only as an interactive local command.
- Verify the server address, port, firewall rules and that both hosts are on reachable networks.
- Test from the Home Assistant host itself; a browser test from another computer does not prove that Home Assistant has access.
- If the model computer sleeps or shuts down, Assist requests will fail until the service returns.
Entity control is unreliable
- Confirm the selected model supports tool use; chat-only models cannot reliably execute Home Assistant actions.
- Expose fewer than 25 entities while evaluating, and use clear names and areas.
- Start with non-critical devices and retain deterministic automations for safety-sensitive behaviour.
- Expect smaller models to make more mistakes and to lose conversational continuity when control is enabled.
A combined host runs out of memory
Measure total usage while Home Assistant, the model, dashboards and add-ons are active. Lower the context, reduce parallel requests, move the model to another computer or add memory that the platform and runtime can actually use. Buying a radio adapter or faster storage will not solve a model-memory shortage.
What a sensible first setup looks like
For a typical home, put Home Assistant on Green, a Pi 4 or Pi 5 meeting its 2 GB host recommendation, a mini PC, NUC or virtual machine. Put Ollama on an existing macOS, Linux or Windows computer if it has enough memory for your selected model and can remain available. If responsiveness matters, choose a supported accelerator with sufficient memory rather than selecting by brand alone. Add a Zigbee, Thread or Matter USB adapter only when your devices require it.
This arrangement keeps automation available even when you replace or tune the model server. It also lets you scale the LLM host independently: begin with CPU or system-memory inference, then move to a GPU-capable machine when latency, context length or concurrent use justifies the cost and power.
Bottom line
Hardware requirements follow the model and workload, not the Home Assistant logo. Provide Home Assistant with a dependable always-on host, provide Ollama with enough memory for the model plus context and concurrency, and choose GPU acceleration when its supported software path and latency benefits match your needs. Treat entity control as an experimental tool-use feature, expose a deliberately small set of devices, and validate speed and reliability on your own model before expanding the system.
Frequently Asked Questions
Can a Raspberry Pi run the local LLM as well as Home Assistant?
The cited Home Assistant guidance recommends a Raspberry Pi 4 or 5 with at least 2 GB of RAM for the Home Assistant installation path. It does not certify that specification for LLM inference. Run the model on a separately sized computer unless testing proves the Pi and chosen model meet your latency and memory requirements.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →How much VRAM do I need for a local LLM?
There is no single answer. Ollama’s Gemma 4 E2B example is a roughly 7.2 GB download with an 8 GB available VRAM or Apple unified-memory recommendation. Larger models, longer context and parallel requests need more. Use the selected model’s guidance and add headroom for context and other services.
Will Home Assistant control devices with any Ollama model?
No. Home Assistant says only models supporting tools can control entities, and it labels the control feature experimental. Start with fewer than 25 exposed entities and non-critical devices; smaller models are more prone to mistakes.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
