WireByte
  • Log In
  • Sign Up
  • AI & Machine Learning
  • Technology
  • Science
  • Space & Astronomy
  • Gaming & Esports

Welcome Back

Log in to your WireByte account

Don't have an account? Sign Up

Create Account

Join WireByte for a personalized experience

Select your favorite categories:

Already have an account? Log In

Stay in the Loop

Get the stories that matter, delivered to your inbox.

You're all set! Check your inbox for a welcome email.

What do you want alerts for?

Search
Trending
WireByte
AI & Machine LearningTechnologyScienceSpace & AstronomyGaming & Esports

Home / AI & Machine Learning

Homelabber Runs 2.8T Kimi AI Model Using RTX 3090s and Optane
Image: © Tom's Hardware
AI & Machine Learning

Homelabber Runs 2.8T Kimi AI Model Using RTX 3090s and Optane

WireByte Staff · November 21, 2026

A tech enthusiast successfully runs the 2.8-trillion-parameter Kimi K3 AI model on a custom homelab setup featuring two RTX 3090 GPUs and Intel Optane memory. By offloading nearly half the model to a RAM drive, generation speed reached 3.11 tokens per second, a 41% improvement. The build requires significant cooling due to high heat output from the persistent memory.

Key points

  • A homelabber enhanced a second-generation Optane AI build by adding two NVIDIA RTX 3090 graphics cards to support large language model inference.
  • The system successfully runs the Moonshot Kimi K3 model, which contains 2.8 trillion parameters, achieving a generation speed of 3.11 tokens per second.
  • Transferring nearly half of the model data onto a RAM drive resulted in a 41% increase in generation speed and a 26% boost in prompt evaluation time.
  • The custom hardware configuration generates substantial heat, necessitating the installation of extra fans to maintain thermal stability during operation.

Custom Hardware Pushes Large Model Limits

A dedicated technology enthusiast has demonstrated the capabilities of a high-performance homelab setup capable of running extremely large artificial intelligence models locally. The build, described as a second-generation Optane AI configuration, utilizes consumer-grade hardware to tackle the Moonshot Kimi K3 model. This specific model is notable for its massive scale, containing 2.8 trillion parameters, which typically requires enterprise-grade infrastructure to operate efficiently.

The core of this experimental setup relies on the integration of two NVIDIA RTX 3090 graphics processing units. These GPUs provide the necessary computational power and memory bandwidth to handle the complex matrix operations required by such a large neural network. The addition of these high-end consumer cards marks a significant evolution from previous iterations of the build, allowing for more complex model architectures to be executed on personal hardware.

Performance Gains Through Memory Optimization

A critical component of the system's performance is the strategic use of Intel Optane persistent memory. The enthusiast reported that by transferring nearly half of the total model data onto a RAM drive, the system achieved substantial efficiency gains. This memory optimization technique allows for faster data access times compared to traditional storage solutions, directly impacting the speed at which the AI can process inputs and generate outputs.

The results of this architectural choice are quantifiable. The setup achieved a generation speed of 3.11 tokens per second. This metric represents the number of text tokens the model can produce in one second, a key indicator of usability for interactive applications. The move to utilize the RAM drive for a significant portion of the model data contributed to a 41% increase in this generation speed.

Thermal Challenges and System Stability

While the performance metrics are impressive, the physical demands of running such a dense model on consumer hardware present significant challenges. The use of persistent Optane memory, while beneficial for speed, generates considerable heat. The enthusiast noted that the hardware gets very hot during operation, requiring active cooling solutions beyond standard configurations.

To manage these thermal loads, the build includes additional fans. This cooling infrastructure is essential to prevent thermal throttling, which would otherwise degrade performance or cause system instability. The need for enhanced cooling highlights the trade-offs inherent in pushing consumer-grade components beyond their typical use cases for high-performance AI inference.

Implications for Local AI Deployment

This build illustrates the growing trend of individuals and small organizations attempting to run large language models locally. By achieving functional inference speeds with a 2.8-trillion-parameter model, the project suggests that the gap between cloud-based AI services and local hardware capabilities may be narrowing. However, the complexity and thermal requirements indicate that such setups remain niche and require significant technical expertise to maintain.

The successful operation of the Kimi K3 model on this hardware provides a reference point for other enthusiasts interested in local AI deployment. It demonstrates that with careful memory management and adequate cooling, it is possible to run models of significant size on non-enterprise hardware. The specific metrics of 3.11 tokens per second serve as a benchmark for similar future builds.

Key takeaways

  • A homelabber successfully runs the 2.8-trillion-parameter Kimi K3 model using two RTX 3090 GPUs.
  • The system achieves a generation speed of 3.11 tokens per second through optimized memory usage.
  • Offloading nearly half the model to a RAM drive improves generation speed by 41%.
  • Prompt evaluation time is reduced by 26% due to the same memory optimization strategy.
  • The build requires extra fans to manage the high heat output from persistent Optane memory.

Sources

  • tomshardware.com
WireByte Staff — Editorial Team

The WireByte editorial team synthesises technology news from multiple primary sources, verifies the facts, and links every source. Articles are produced with AI assistance and reviewed under our editorial policy.

More in AI & Machine Learning

Rockstar Details NPC AI and World Design in GTA 6 Preview

Rockstar Details NPC AI and World Design in GTA 6 Preview

Rockstar Games has revealed new details on Grand Theft Auto VI's non-player character (N…

Google Launches Gemini 4 Neon AI Model for Complex Workflows

Google Launches Gemini 4 Neon AI Model for Complex Workflows

Google has released Gemini 4 Neon, an AI model optimized for software engineering and kn…

Signal Blocks Screenshots to Shield Privacy from AI Agents

Signal Blocks Screenshots to Shield Privacy from AI Agents

Encrypted messaging service Signal has restricted screenshots on mobile devices to preve…

Trump Proposes Ban on US AI Data Centers Overseas

Trump Proposes Ban on US AI Data Centers Overseas

President Trump announced a potential ban on US companies building AI data centers abroa…

artificial intelligence technology

OpenAI Expands Dots Agent Access to ChatGPT Plus Users

OpenAI is testing its always-on AI agent, Dots, with a select group of ChatGPT Plus subs…

Latest

Samsung Galaxy Watch 7 Hits 40% Discount on Amazon for Black Friday

Samsung Galaxy Watch 7 Hits 40% Discount on Amazon for Black Friday

Samsung’s Galaxy Watch 7 is available at a 40% discount on Amazon during Black Friday sa…

cinema entertainment

Quentin Tarantino's The Hateful Eight Now Free on Tubi

Quentin Tarantino's 2015 Western thriller The Hateful Eight is officially available for …

Rockstar Details NPC AI and World Design in GTA 6 Preview

Rockstar Details NPC AI and World Design in GTA 6 Preview

Rockstar Games has revealed new details on Grand Theft Auto VI's non-player character (N…

electric car

Harbinger Raises $240M to Target Commercial EV Gap

Electric truck maker Harbinger secured $240 million in Series D funding, following a $30…

Real Madrid Hosts Celta Vigo in La Liga Showdown

Real Madrid Hosts Celta Vigo in La Liga Showdown

Real Madrid faces Celta Vigo at the Santiago Bernabéu on November 21 at 3 p.m. ET. The h…

UK Gov Abolishes Tech Dept, Creates AI Taskforce

UK Gov Abolishes Tech Dept, Creates AI Taskforce

The UK government abolished the Department for Science, Innovation and Technology on Jul…

Related articles

Rockstar Details NPC AI and World Design in GTA 6 Preview AI & Machine Learning Rockstar Details NPC AI and World Design in GTA 6 Preview WireByte

Rockstar Games has revealed new details on Grand Theft Auto VI's non-player character (NPC) systems, emphasizing unique personalities and reactive behaviors to create a living world. The studio dedicated a specific team to ensure authenticity in dialogue and interactions, with over 100 explorable interiors adding narrative depth to the setting of Leonida.

WireByte Staff · November 21, 2026

Google Launches Gemini 4 Neon AI Model for Complex Workflows AI & Machine Learning Google Launches Gemini 4 Neon AI Model for Complex Workflows WireByte

Google has released Gemini 4 Neon, an AI model optimized for software engineering and knowledge work. Priced at $3 per million input tokens, it claims superior performance against rivals like Claude Opus 5.5 and GPT-6.2 Sol. The launch targets enterprise efficiency, offering enhanced capabilities for long-running agentic tasks while maintaining strict safety safeguards.

WireByte Staff · November 21, 2026

Signal Blocks Screenshots to Shield Privacy from AI Agents AI & Machine Learning Signal Blocks Screenshots to Shield Privacy from AI Agents WireByte

Encrypted messaging service Signal has restricted screenshots on mobile devices to prevent AI agents from reading private chats. The company warns that deep OS-level AI integrations by Apple, Google, and Microsoft erode privacy guarantees. Signal urges tech giants to provide developers with concrete tools to block agent access, citing an existential threat to user data security.

WireByte Staff · November 21, 2026

Trump Proposes Ban on US AI Data Centers Overseas AI & Machine Learning Trump Proposes Ban on US AI Data Centers Overseas WireByte

President Trump announced a potential ban on US companies building AI data centers abroad, threatening heavy taxes for violations. The move aims to keep AI infrastructure domestic to counter China. Details remain vague, with Trump admitting he had not yet discussed the policy with his team. Enforcement and legal feasibility are currently unclear as the administration considers the proposal.

WireByte Staff · November 20, 2026

artificial intelligence technology AI & Machine Learning OpenAI Expands Dots Agent Access to ChatGPT Plus Users WireByte

OpenAI is testing its always-on AI agent, Dots, with a select group of ChatGPT Plus subscribers, marking a shift from its previous restriction to higher-tier plans. The rollout also includes all ChatGPT Business Standard users in the US. Dots, powered by GPT-6 models, operate continuously in the background to automate tasks. Access is currently limited and unavailable in Europe, the UK, and Switzerland.

WireByte Staff · November 20, 2026

Samsung Galaxy Watch 7 Hits 40% Discount on Amazon for Black Friday Space & Astronomy Samsung Galaxy Watch 7 Hits 40% Discount on Amazon for Black Friday WireByte

Samsung’s Galaxy Watch 7 is available at a 40% discount on Amazon during Black Friday sales. The 40mm Bluetooth model in Green is priced at $149.99, down from its standard $249.99 list price. This marks the lowest price for the device on Amazon since last year's holiday season. The smartwatch runs Wear OS with Samsung’s One UI Watch interface.

WireByte Staff · November 21, 2026

WireByte

Tech & AI news, synthesised and to the point.

Delivering original syntheses and breaking news with cited sources. Stay informed with WireByte.

Sections

  • AI & Machine Learning
  • Technology
  • Science
  • Space & Astronomy
  • Gaming & Esports

Company

  • About Us
  • Editorial Policy
  • Privacy Policy
  • Contact: [email protected]

Newsletters

Get the stories that matter, delivered to your inbox.

© 2026 WireByte. All rights reserved. Original syntheses with cited sources.

Home Search