ARMdevices.net

MSI Coolers presented a range of thermal and power solutions at Computex 2026, showcasing new designs in liquid and air cooling. Among the key cooling exhibits is the MEG CORELIQUID E15 360 all-in-one liquid cooler, which features a cooling safeguard function that halts operation and flashes a warning indicator during a fan or pump failure before executing a recovery sequence once the obstruction is removed. The cooler utilizes a Pogo-pin magnetic fan connection scheme that eliminates individual fan cables, routing power, PWM control, ARGB synchronization, and pump monitoring through a single Type-C connection to a JF2 motherboard header.

—
HDMI® Technology is the foundation for the worldwide ecosystem of HDMI-connected devices; integrated with displays, set-top boxes, laptops, audio video receivers and other product types. Because of this global usage, manufacturers, resellers, integrators and consumers must be assured that their HDMI® products work seamlessly together and deliver the best possible performance by sourcing products from licensed HDMI Adopters or authorized resellers. For HDMI Cables, consumers can look for the official HDMI® Cable Certification Labels on packaging. Innovation continues with the latest HDMI 2.2 Specification that supports higher 96Gbps bandwidth and next-gen HDMI Fixed Rate Link technology to provide optimal audio and video for a wide range of device applications. Higher resolutions and refresh rates are supported, including up to 12K@120 and 16K@60. Additionally, more high-quality options are supported, including uncompressed full chroma formats such as 8K@60/4:4:4 and 4K@240/4:4:4 at 10-bit and 12-bit color.
—

The MPG CORELIQUID P22 360 liquid cooler provides system metrics, custom animations, or image backgrounds on its integrated 2.1-inch circular IPS display. To streamline system assembly, it is equipped with a universal bracket compatible with both Intel and AMD processor sockets. For users seeking a different aesthetic, the MAG CORELIQUID A23 360 W offers a lifestyle-oriented design with a woven fabric pump cover that conceals a digital temperature display until it is activated. The A23 uses engineered fan blades designed to lower noise output while maintaining high static pressure.

For air-cooled systems, the MPG COREFROZR AP15 dual-tower air cooler features an integrated digital display for real-time temperature and metric monitoring. The cooler is designed for easy DIY installation, utilizing a mounting design that allows users to secure the heatsink using a long screwdriver without removing the cooling fans. The display and fans coordinate to provide thermal management without requiring complex mounting disassembly.

In the PC chassis segment, MSI Coolers displayed the MEG MAESTRO 900R, a large-format enthusiast case with a centered interior layout and triple-sided tempered glass panels designed for custom liquid cooling loops. The MPG VIXTA 300 series is split into an Airflow edition with pre-installed dual 160mm front intake fans and a Tempered Glass edition that supports dual 140mm side intake fans. Both cases utilize tool-free panels, magnetic components, and Pogo-pin fan connections to reduce cable routing complexity and improve overall assembly efficiency.

The exhibits also featured power supply units from the MPG series equipped with Safeguard+ technology, which offers per-pin power monitoring for 12V-2×6 graphics card cables to shut down the system and prevent overheating in the event of a power spike or bad connection. These cooling and power components were demonstrated alongside enthusiast motherboards like the MEG Z890 GODLIKE and the UNIFY-X, which includes a remote control module that allows users to adjust overclocking parameters directly without entering the BIOS.

source https://www.youtube.com/watch?v=GrN3kYJjLYI

Samsung Display showcased its latest panel technologies at Computex 2026, highlighting next-generation display developments designed for professional content creation, gaming, and robotics. The exhibition focused on expanding the capabilities of QD-OLED and OLED panels, featuring improvements in peak brightness, color uniformity, and form factor flexibility.

—
HDMI® Technology is the foundation for the worldwide ecosystem of HDMI-connected devices; integrated with displays, set-top boxes, laptops, audio video receivers and other product types. Because of this global usage, manufacturers, resellers, integrators and consumers must be assured that their HDMI® products work seamlessly together and deliver the best possible performance by sourcing products from licensed HDMI Adopters or authorized resellers. For HDMI Cables, consumers can look for the official HDMI® Cable Certification Labels on packaging. Innovation continues with the latest HDMI 2.2 Specification that supports higher 96Gbps bandwidth and next-gen HDMI Fixed Rate Link technology to provide optimal audio and video for a wide range of device applications. Higher resolutions and refresh rates are supported, including up to 12K@120 and 16K@60. Additionally, more high-quality options are supported, including uncompressed full chroma formats such as 8K@60/4:4:4 and 4K@240/4:4:4 at 10-bit and 12-bit color.
—

A key highlight of the showcase is a tandem OLED display capable of achieving a peak brightness of 4,500 nits. The booth also featured QD-OLED monitors tailored for video editing and financial operations, emphasizing low reflections and high black level stability for legibility. Samsung Display demonstrated Sync Chroma technology, which maintains color consistency across different device classes, including televisions, laptops, and smartphones.

For gaming applications, the manufacturer displayed 34-inch QHD+ and 31.5-inch 4K QD-OLED monitors featuring a 360 Hz refresh rate and Clear MR certification to minimize motion blur. Additionally, the exhibit included an ultra-slim gaming display prototype and various laptop displays. Samsung Semiconductor also co-exhibited at the booth, showing HBM4e memory modules designed for integration alongside next-generation GPUs.

The showcase concluded with demonstrations of emerging display architectures. These included a glasses-free 3D light field display designed to minimize surface reflections, and a stretchable display prototype integrated onto a mobile robot platform that physically expands and contracts on the sides.

source https://www.youtube.com/watch?v=fd4_uWOY2tI

Samsung Display presented its latest display technology at Computex 2026, showcasing the Samsung 31 point 5 inch 4 K 360 hz panel. The display combines a high resolution with a fast 360 Hz refresh rate to target demanding gaming and professional creative workflows. As an OEM panel supplier, the company demonstrated the hardware to show its capabilities ahead of adoption by third-party display brands in the coming months.

—
HDMI® Technology is the foundation for the worldwide ecosystem of HDMI-connected devices; integrated with displays, set-top boxes, laptops, audio video receivers and other product types. Because of this global usage, manufacturers, resellers, integrators and consumers must be assured that their HDMI® products work seamlessly together and deliver the best possible performance by sourcing products from licensed HDMI Adopters or authorized resellers. For HDMI Cables, consumers can look for the official HDMI® Cable Certification Labels on packaging. Innovation continues with the latest HDMI 2.2 Specification that supports higher 96Gbps bandwidth and next-gen HDMI Fixed Rate Link technology to provide optimal audio and video for a wide range of device applications. Higher resolutions and refresh rates are supported, including up to 12K@120 and 16K@60. Additionally, more high-quality options are supported, including uncompressed full chroma formats such as 8K@60/4:4:4 and 4K@240/4:4:4 at 10-bit and 12-bit color.
—

The Samsung 31 point 5 inch 4 K 360 hz panel features OLED technology with True Black 600 certification, enabling deep black levels and high contrast ratios. The integration of 4K resolution at a 31.5-inch form factor yields a high pixel density, providing sharp text and detailed image rendering. This combination of resolution and a 360 Hz refresh rate represents the next generation of high-speed, high-density display panels for the consumer and professional markets.

Color performance and viewing angles are key aspects of this panel design. The OLED structure of the Samsung 31 point 5 inch 4 K 360 hz maintains consistent color accuracy and luminance even at wide off-axis viewing angles, which is critical for multi-monitor setups and collaborative work environments. The panel is designed to deliver vibrant color reproduction and rapid pixel response times, minimizing motion blur during fast-paced sequences.

Samsung Display’s demonstration at Computex highlights the ongoing transition of high-refresh-rate panels into higher resolutions like 4K. By providing the Samsung 31 point 5 inch 4 K 360 hz to various monitor manufacturers, the company is establishing the foundation for a new class of premium displays. Consumer products utilizing this technology are expected to enter the market from various brands in the near future.

source https://www.youtube.com/watch?v=WPvB9qCIMcA

Media Tek demonstrated a micro-LED optical data transfer technology for data center applications at Computex 2026. This technology, developed as a co-development proof of concept in collaboration with Microsoft, aims to replace copper connections in high-performance computing environments. By integrating micro-LEDs directly onto silicon, the project explores new architectures for high-bandwidth data transmission.

—
HDMI® Technology is the foundation for the worldwide ecosystem of HDMI-connected devices; integrated with displays, set-top boxes, laptops, audio video receivers and other product types. Because of this global usage, manufacturers, resellers, integrators and consumers must be assured that their HDMI® products work seamlessly together and deliver the best possible performance by sourcing products from licensed HDMI Adopters or authorized resellers. For HDMI Cables, consumers can look for the official HDMI® Cable Certification Labels on packaging. Innovation continues with the latest HDMI 2.2 Specification that supports higher 96Gbps bandwidth and next-gen HDMI Fixed Rate Link technology to provide optimal audio and video for a wide range of device applications. Higher resolutions and refresh rates are supported, including up to 12K@120 and 16K@60. Additionally, more high-quality options are supported, including uncompressed full chroma formats such as 8K@60/4:4:4 and 4K@240/4:4:4 at 10-bit and 12-bit color.
—

The proof of concept integrates more than 150 micro-LEDs on a single CMOS-compatible silicon die. Each individual micro-LED operates at a bandwidth of 1 to 2 Gbps. While the single-channel speed is modest, the array allows aggregating hundreds of signals to achieve high total bandwidth, demonstrated at 100 to 200 Gbps, with scaling capability up to 800 Gbps total throughput across eight fibers in the active optical cable module.

Using micro-LEDs as a light source instead of traditional semiconductor lasers offers key advantages in manufacturing cost, reliability, and temperature coefficient. Unlike lasers, these micro-LEDs do not require external light sources and can be embedded directly. However, physical scaling faces constraints from thermal dissipation and electromagnetic signal coupling, requiring careful spatial layout within the module form factor.

This optical engine is designed for co-packaged optics, near-packaged optics, and direct CPU or AI accelerator integration. The setup utilizes standard optical fiber, but requires a specialized, customized fiber attachment technique to couple the fiber array to the micro-LEDs and photodetectors. The design path focuses on scaling SerDes speeds and total bandwidth for future production readiness.

source https://www.youtube.com/watch?v=8wdBt5kUitk

Deep X showcased its edge AI hardware and software solutions at Computex 2026, highlighting the next-generation XM2 silicon alongside its established system-on-chip and accelerator module portfolio. Head of Sales for Deep X Taiwan, Jack Horn, introduced the company’s Neural Processing Unit (NPU) technologies designed for low-power, high-performance edge computing applications. The company, headquartered in South Korea and founded by former Apple systems researcher Lokwon Kim, specializes in designing dedicated accelerators optimized for matrix operations to support machine learning workloads on-device.

—
HDMI® Technology is the foundation for the worldwide ecosystem of HDMI-connected devices; integrated with displays, set-top boxes, laptops, audio video receivers and other product types. Because of this global usage, manufacturers, resellers, integrators and consumers must be assured that their HDMI® products work seamlessly together and deliver the best possible performance by sourcing products from licensed HDMI Adopters or authorized resellers. For HDMI Cables, consumers can look for the official HDMI® Cable Certification Labels on packaging. Innovation continues with the latest HDMI 2.2 Specification that supports higher 96Gbps bandwidth and next-gen HDMI Fixed Rate Link technology to provide optimal audio and video for a wide range of device applications. Higher resolutions and refresh rates are supported, including up to 12K@120 and 16K@60. Additionally, more high-quality options are supported, including uncompressed full chroma formats such as 8K@60/4:4:4 and 4K@240/4:4:4 at 10-bit and 12-bit color.
—

The upcoming XM2 silicon, scheduled for fabrication tape-out around March 2027, is a five-watt NPU designed to deliver 80 TOPS of performance. This processor targets edge-based Large Language Models (LLMs) and Generative AI, supporting models with parameter counts ranging from 20 billion up to 100 billion utilizing Mixture of Experts (MoE) architectures. Operating within a power budget of five watts, the XM2 achieves token generation speeds of 20 to 30 tokens per second and supports memory configurations of up to 64 gigabytes to run large models directly on edge devices, including models such as Gemma.

In addition to the XM2, Deep X demonstrated its current hardware implementations, including the DX-H1 PCIe accelerator board and the DX-M1 series. The DX-H1 board integrates four DX-M1 chips, providing a combined performance of 100 TOPS with 16 gigabytes of system memory distributed as four gigabytes per chip. The M1M variant packages the memory on-chip with two gigabytes of integrated RAM. These modules are integrated into various industrial PCs and edge servers through partnerships with local IPC companies such as Advantech, AAEON, Avalue, and DFI, utilizing standard M.2 slots to expand hosting system capabilities.

A notable storage-focused integration includes a collaboration with Apacer, combining a four-terabyte PCIe SSD and two DX-M1 processors on a single M.2 board to deliver simultaneous storage and AI processing. The hardware ecosystem also features the DX-M1+ M.2 module, which comes in a quad-chip 100 TOPS configuration or a dual-chip 50 TOPS version that pairs two NPU chips with two Rockchip SoCs for video transcoding and decoding. Software demonstrations at the booth highlighted system compatibility across Linux, Windows, and Android, showcasing mobile OS integration for physical AI alongside medical diagnostic tablets built in partnership with Powertip.

Architecturally, the NPU design is optimized specifically for matrix operations, which are central to modern machine learning workloads, providing competitive performance-per-watt metrics. To support deployment, Deep X offers a custom software compiler and quantizer that converts models from FP32 precision to INT8 precision. This optimization process compresses the model weight representations to minimize processing latency and power consumption at the edge. The software tools allow developers to perform fine-tuning on quantized models to recover accuracy loss typically associated with precision reduction, ensuring high-accuracy inference during deployment.

source https://www.youtube.com/watch?v=iuHR0PNQ1TE

Ugreen demonstrated its latest network-attached storage and smart home technology at Computex 2026, highlighted by the debut of the AI NAS iDX series. This family includes the iDX Pro and non-Pro models, which run large language models locally on the device to enable private chat applications without uploading data to cloud services. The processing is driven by Intel Core Ultra 7 processors utilizing built-in AI acceleration to handle local data analysis. Cliff from Ugreen presented the hardware details, showcasing how these systems serve as a secure hub for local artificial intelligence and data management.

—
HDMI® Technology is the foundation for the worldwide ecosystem of HDMI-connected devices; integrated with displays, set-top boxes, laptops, audio video receivers and other product types. Because of this global usage, manufacturers, resellers, integrators and consumers must be assured that their HDMI® products work seamlessly together and deliver the best possible performance by sourcing products from licensed HDMI Adopters or authorized resellers. For HDMI Cables, consumers can look for the official HDMI® Cable Certification Labels on packaging. Innovation continues with the latest HDMI 2.2 Specification that supports higher 96Gbps bandwidth and next-gen HDMI Fixed Rate Link technology to provide optimal audio and video for a wide range of device applications. Higher resolutions and refresh rates are supported, including up to 12K@120 and 16K@60. Additionally, more high-quality options are supported, including uncompressed full chroma formats such as 8K@60/4:4:4 and 4K@240/4:4:4 at 10-bit and 12-bit color.
—

The flagship iDX Pro version is equipped with either 32GB or 64GB of RAM and features a front-facing status monitor, an 8K HDMI port for high-resolution video playback, and an OCuLink port designed for external graphics card expansion. In comparison, the non-Pro version lacks the front display and comes with 32GB of RAM, while still supporting high-definition video output. Both models support optional HDMI 2.1 cables from Ugreen to enable 8K playback on compatible displays.

In addition to the Intel-based iDX series, Ugreen showed its existing 6-bay and 8-bay Pro NAS models, which are equipped with Thunderbolt 5 connectivity and dual 10GbE network ports on the back panel. A new AMD-based NAS model was also highlighted, featuring support for enterprise-grade U.2 and U.3 SSDs in its first two drive bays. This capability allows power users to configure high-speed storage pools or utilize NVMe SSDs as fast cache to accelerate system performance.

For home users, Ugreen introduces home automation integration with upcoming IP cameras. These security cameras connect to the central NAS to store video recordings locally. Ugreen also showcased its entry-level NAS model running a Rockchip ARM CPU. Designed for household media and phone backup, this entry-level device provides simple operations with a single gigabit LAN port, allowing family members to share and centralize their storage.

Finally, Ugreen previewed an upcoming 40Gbps portable storage enclosure that features an interactive touchscreen. The display provides real-time readouts of transfer speed, drive health, remaining capacity, and overall lifecycle status. The device supports passcode security directly on the screen and can connect to laptops or directly to a NAS. Ugreen continues to update its operating system and software applications to enhance user convenience and expand functionality across its entire storage lineup.

source https://www.youtube.com/watch?v=o2YHj0P6jOo

At Computex 2026 in Taipei, Taiwan, faytech Sales Director Michael Foss presented the HLD (Hololuminescent Display) series, developed in partnership with Looking Glass. The HLD series recently won the Display of the Year award from the Society for Information Display (SID) at their annual conference. These glasses-free holographic displays are engineered for retail, digital signage, and experiential interactive systems.

—
HDMI® Technology is the foundation for the worldwide ecosystem of HDMI-connected devices; integrated with displays, set-top boxes, laptops, audio video receivers and other product types. Because of this global usage, manufacturers, resellers, integrators and consumers must be assured that their HDMI® products work seamlessly together and deliver the best possible performance by sourcing products from licensed HDMI Adopters or authorized resellers. For HDMI Cables, consumers can look for the official HDMI® Cable Certification Labels on packaging. Innovation continues with the latest HDMI 2.2 Specification that supports higher 96Gbps bandwidth and next-gen HDMI Fixed Rate Link technology to provide optimal audio and video for a wide range of device applications. Higher resolutions and refresh rates are supported, including up to 12K@120 and 16K@60. Additionally, more high-quality options are supported, including uncompressed full chroma formats such as 8K@60/4:4:4 and 4K@240/4:4:4 at 10-bit and 12-bit color.
—

The product family includes 16-inch, 27-inch, and large-format 86-inch versions, which are already in production or entering deployment phases. In addition to these sizes, faytech is developing 43-inch and 55-inch versions to expand the lineup. All initial units, including the 16-inch, 27-inch, and 86-inch models, feature native 4K display panels to output high-fidelity content.

From a hardware and integration standpoint, the HLD series supports various external media players, including single-board computers like the Raspberry Pi, as well as standard Android devices and Windows-based PCs. The displays support both native 2D and 3D content. For media that is not originally designed for light-field or holographic playback, the system can upscale the content to render it in 3D.

Initial manufacturing runs of the large 86-inch display, which started with a mass production batch of 20 units for global distributors, have sold out, prompting additional production runs. This commercial interest spans digital signage integrators, marketing agencies, and rental providers looking to deploy the displays at trade shows, exhibitions, and retail storefronts. Engagement is also growing among content creation and graphic design agencies that render custom professional 3D assets for brands.

Future iterations of the HLD technology aim to combine these light-field displays with interactive AI systems. While the current displays feature built-in speakers, upcoming developments include integrating camera systems and directional microphone arrays to enable real-time communication with 3D AI avatars, providing an immersive, conversational interface for retail and public installations.

source https://www.youtube.com/watch?v=TpN46ghJBpg

faytech, in collaboration with Napster, demonstrated their smart AI-powered quick service restaurant ordering kiosk at Computex 2026. The system, referred to as Mc Nap, combines faytech’s interactive hardware with Napster’s software integration to facilitate natural language food ordering. Managing founder Arne Weber presented the physical kiosk, highlighting how it integrates touch display technology with voice interaction to streamline customer service in retail and fast food environments.

—
HDMI® Technology is the foundation for the worldwide ecosystem of HDMI-connected devices; integrated with displays, set-top boxes, laptops, audio video receivers and other product types. Because of this global usage, manufacturers, resellers, integrators and consumers must be assured that their HDMI® products work seamlessly together and deliver the best possible performance by sourcing products from licensed HDMI Adopters or authorized resellers. For HDMI Cables, consumers can look for the official HDMI® Cable Certification Labels on packaging. Innovation continues with the latest HDMI 2.2 Specification that supports higher 96Gbps bandwidth and next-gen HDMI Fixed Rate Link technology to provide optimal audio and video for a wide range of device applications. Higher resolutions and refresh rates are supported, including up to 12K@120 and 16K@60. Additionally, more high-quality options are supported, including uncompressed full chroma formats such as 8K@60/4:4:4 and 4K@240/4:4:4 at 10-bit and 12-bit color.
—

The kiosk is equipped with an eight-microphone array designed to establish a localized audio zone directly in front of the terminal. This array uses physical directional capturing to isolate the speaker’s voice from ambient exhibition hall noise. The hardware-level audio processing includes automatic echo cancellation, active noise cancellation, and voice isolation to ensure the software receives clear voice inputs.

On the software side, the kiosk leverages Napster’s natural language processing algorithms to interpret conversational orders, allowing customers to add, modify, or remove items dynamically during the ordering process. The AI assistant can handle adjustments such as altering portion sizes or removing specific items like salads and nuggets based on direct spoken instructions. The conversational flow is designed to mimic human interaction while updating the order on screen in real-time.

To enhance user engagement and expedite transactions, the kiosk supports a hybrid interface. Customers can interact with the system simultaneously using the physical faytech touchscreen and voice commands, allowing them to browse menus, confirm choices, or make changes manually. Additionally, the system incorporates user recognition capabilities, enabling opt-in members to receive personalized greetings and pre-configured discounts based on past preferences, subject to data privacy authorizations.

For payment processing, the terminal integrates contactless card reading capabilities, permitting users to finalize their transactions immediately at the kiosk. The entire system is built to target fixed-service restaurants and quick-service retail locations seeking to deploy digital order processing solutions. This hybrid design allows operators to customize the conversational style and branding of the digital assistant to align with specific franchise requirements.

source https://www.youtube.com/watch?v=niT-n_7Q8xc

Nvidia Nemoclaw GTC Taipei demonstrated its open-source AI agent technologies, showcasing development with NeMo Claw and Open Shell. These tools are designed to facilitate the creation of customized AI agents capable of performing automated tasks on local edge hardware or within secure enterprise networks. The demonstration highlighted how developers can build and run these assistant agents for daily workflows or complex industry applications without relying on cloud-based compute environments.

—
HDMI® Technology is the foundation for the worldwide ecosystem of HDMI-connected devices; integrated with displays, set-top boxes, laptops, audio video receivers and other product types. Because of this global usage, manufacturers, resellers, integrators and consumers must be assured that their HDMI® products work seamlessly together and deliver the best possible performance by sourcing products from licensed HDMI Adopters or authorized resellers. For HDMI Cables, consumers can look for the official HDMI® Cable Certification Labels on packaging. Innovation continues with the latest HDMI 2.2 Specification that supports higher 96Gbps bandwidth and next-gen HDMI Fixed Rate Link technology to provide optimal audio and video for a wide range of device applications. Higher resolutions and refresh rates are supported, including up to 12K@120 and 16K@60. Additionally, more high-quality options are supported, including uncompressed full chroma formats such as 8K@60/4:4:4 and 4K@240/4:4:4 at 10-bit and 12-bit color.
—

Edge compute applications were shown running on the Nvidia Jetson Nano, demonstrating the capability of lightweight hardware to execute local AI tasks. One specific implementation included a local podcast generation system that runs entirely on the edge device. By utilizing edge hardware, developers can deploy responsive AI workflows while maintaining data privacy and reducing latency associated with external API calls.

For enterprise deployments, the integration of Open Shell with NeMo Claw provides a policy security layer to enforce safety guidelines. The multi-agent demonstrations illustrated how multiple independent agents, each with specific skills and personalities, communicate within isolated sandboxes. These sandboxes apply strict execution policies that limit the agent’s action boundaries, ensuring that even if an agent possesses extensive capabilities or tools, its compute execution is restricted according to pre-defined organizational policies.

The demonstrations highlighted the capabilities of autonomous agent teams in software development. A researcher agent can scrape the web for design details and program games or web applications using 3.js, Blender-style rendering, or 8-bit graphics. Additionally, operator agents can control interactive applications, utilizing neural graphs and decision trees to learn from execution failures and optimize performance dynamically over successive trials.

Nvidia Nemoclaw GTC Taipei is actively contributing to the open-source open-claw ecosystem by developing sandboxing policies and enhancing agent frameworks. Practical use-cases include community management automation, where multiple agents scan online platforms like Reddit, X, and Hacker News to compile trend reports and suggest business strategies. These workflows demonstrate how natural language interfaces are replacing traditional programming methods, allowing a wider range of users to customize and deploy AI assistants.

source https://www.youtube.com/watch?v=hAUFN62uOaA

Nvidia solution architects are utilizing autonomous AI agents to automate and accelerate complex chip design and system design workflows. By deploying agentic tools such as Nemoclaw and OpenCLAW, engineers can offload repetitive simulation setup and geometry generation tasks. These agents run locally on laptops or on remote servers, interfacing with engineering applications through Model Context Protocol (MCP) servers and APIs to streamline the design iteration cycle.

—
HDMI® Technology is the foundation for the worldwide ecosystem of HDMI-connected devices; integrated with displays, set-top boxes, laptops, audio video receivers and other product types. Because of this global usage, manufacturers, resellers, integrators and consumers must be assured that their HDMI® products work seamlessly together and deliver the best possible performance by sourcing products from licensed HDMI Adopters or authorized resellers. For HDMI Cables, consumers can look for the official HDMI® Cable Certification Labels on packaging. Innovation continues with the latest HDMI 2.2 Specification that supports higher 96Gbps bandwidth and next-gen HDMI Fixed Rate Link technology to provide optimal audio and video for a wide range of device applications. Higher resolutions and refresh rates are supported, including up to 12K@120 and 16K@60. Additionally, more high-quality options are supported, including uncompressed full chroma formats such as 8K@60/4:4:4 and 4K@240/4:4:4 at 10-bit and 12-bit color.
—

In a demonstration of thermal management optimization, the AI agent is tasked with optimizing a liquid-cooled heat sink design to minimize chip surface temperature. The agent autonomously builds the required geometry, generates meshes, and interfaces directly with ANSYS simulation software. By programmatically controlling the simulation runs, the agent collects performance data and analyzes the results to guide subsequent design adjustments.

After evaluating the thermal simulation outputs, the agent autonomously iterates on design parameters to optimize the heat sink geometry. The simulation results and thermal profiles are visualized using Nvidia Omniverse, providing real-time rendering of the temperature distribution across the chip surface. This loop of design generation, simulation, and analysis continues until the target cooling efficiency is achieved.

The architecture relies on Nvidia large language models, including Nemotron, to drive the reasoning and planning capabilities of the agent. Rather than using computer vision or direct mouse emulation, the agent connects via MCP servers to access tool APIs, allowing robust command-line and programmatic control of CAD and simulation suites. This integration enables domain experts to interact with the design system through natural language while maintaining precise execution.

Integrating agentic workflows allows chip designers to focus on high-level architecture and system goals rather than repetitive CAD and simulation setup tasks. As simulation and CAD tools increasingly integrate AI agents directly, designers will utilize APIs and command-line interfaces to orchestrate multi-application workflows. This shift empowers domain experts with greater operational efficiency and accelerates the timeline for physical design verification.

source https://www.youtube.com/watch?v=8RDqXBw4PH4

Nvidia Alpamayo is a proprietary reasoning model designed for autonomous driving applications. The model is built on the open-source Cosmos 2.5 Reason foundation and subsequently post-trained for self-driving vehicle stacks. During development and testing at headquarters, the real-time autonomous driving stack runs on a cluster of 16 VR Ruben systems.

—
HDMI® Technology is the foundation for the worldwide ecosystem of HDMI-connected devices; integrated with displays, set-top boxes, laptops, audio video receivers and other product types. Because of this global usage, manufacturers, resellers, integrators and consumers must be assured that their HDMI® products work seamlessly together and deliver the best possible performance by sourcing products from licensed HDMI Adopters or authorized resellers. For HDMI Cables, consumers can look for the official HDMI® Cable Certification Labels on packaging. Innovation continues with the latest HDMI 2.2 Specification that supports higher 96Gbps bandwidth and next-gen HDMI Fixed Rate Link technology to provide optimal audio and video for a wide range of device applications. Higher resolutions and refresh rates are supported, including up to 12K@120 and 16K@60. Additionally, more high-quality options are supported, including uncompressed full chroma formats such as 8K@60/4:4:4 and 4K@240/4:4:4 at 10-bit and 12-bit color.
—

While the Alpamayo model can fit onto smaller graphics processing units such as the GB300, executing the full simulation and autonomous driving stack in real time currently requires high-performance hardware. The high computing demand during testing stems from simulating multiple camera views and generating the surrounding environment in real time. Once the system is deployed on production physical vehicles, only the Alpamayo reasoning model is executed, eliminating the need for real-time world generation on the vehicle itself.

Nvidia Alpamayo is currently deployed on test vehicles at the company headquarters. The autonomous driving model is designed to scale down to smaller chips for commercial vehicles. To train the autonomous model, the system uses simulated environments to accumulate driving experience. This training process is supported by Omni Dreams, a world-generation model built on the open-source Cosmos 2.5 Predict model.

Omni Dreams generates synthetic environment data to train the autonomous driving model under diverse road conditions, including rain, snow, and varying traffic densities. The simulation runs locally on a DGX Station using the Alpasim platform. This local simulation generates virtual environments in real time, allowing test vehicles to navigate maps without collision constraints, thereby augmenting physical road testing.

The core models Alpamayo and Omni Dreams remain proprietary. However, their underlying backbone models, Cosmos 2.5 and Cosmos 3, are open source. This allows developers to utilize the open backbone models to train custom autonomous systems for specific deployment requirements. Nvidia is currently updating its autonomous driving stack to leverage the recently released Cosmos 3 model.

source https://www.youtube.com/watch?v=xdfudTRNJu4

Microsoft Surface Laptop Ultra is showcased running on the Nvidia RTX Spark system-on-chip, which integrates an Arm CPU and an Nvidia Blackwell GPU. This architecture allows devices to be thinner and lighter with longer battery life while maintaining high performance. Details on native Arm application support are tracked at worksonwa.com.

—
HDMI® Technology is the foundation for the worldwide ecosystem of HDMI-connected devices; integrated with displays, set-top boxes, laptops, audio video receivers and other product types. Because of this global usage, manufacturers, resellers, integrators and consumers must be assured that their HDMI® products work seamlessly together and deliver the best possible performance by sourcing products from licensed HDMI Adopters or authorized resellers. For HDMI Cables, consumers can look for the official HDMI® Cable Certification Labels on packaging. Innovation continues with the latest HDMI 2.2 Specification that supports higher 96Gbps bandwidth and next-gen HDMI Fixed Rate Link technology to provide optimal audio and video for a wide range of device applications. Higher resolutions and refresh rates are supported, including up to 12K@120 and 16K@60. Additionally, more high-quality options are supported, including uncompressed full chroma formats such as 8K@60/4:4:4 and 4K@240/4:4:4 at 10-bit and 12-bit color.
—

The platform supports native execution for major creative and productivity tools. Creative applications such as Adobe Photoshop, Adobe Premiere, and DaVinci Resolve run natively to handle demanding tasks like 4K60 video editing. Productivity suites including Microsoft Office, Microsoft Edge, and Google Chrome also feature full native Arm compatibility.

For applications that do not run natively on Arm, Windows utilizes the Prism translation layer. Prism automatically translates x86 instruction sets into Arm instructions, allowing older or unoptimized software to run on the system. This translation layer ensures compatibility across a broader ecosystem of legacy software.

The unified memory architecture of the Nvidia RTX Spark chip provides up to 128GB of shared memory across the CPU and GPU. This shared memory pool allows the Blackwell GPU to load large AI models, supporting configurations of up to 120 billion parameters without the limits of discrete graphics card VRAM.

Microsoft also addresses the gaming ecosystem on Arm through compatibility updates. Anti-cheat support has been enabled for titles like League of Legends and Valorant from Riot Games, as well as Fortnite from Epic Games. The broader hardware ecosystem includes alternative Surface laptop models in 13-inch and 15-inch sizes alongside the Surface Laptop Ultra.

source https://www.youtube.com/watch?v=gqWEqLem4iY

Nvidia RTX Spark Unreal Engine demonstrated the laptop form factor of its GB10 Blackwell system architecture at Computex 2026. This prototype houses the same silicon architecture previously deployed in desktop small box configurations, integrating a 20-core ARM CPU and 128 GB of unified memory. The design adapts the high-performance processor architecture into a mobile envelope, addressing the thermal management and power delivery constraints inherent in portable hardware.

—
HDMI® Technology is the foundation for the worldwide ecosystem of HDMI-connected devices; integrated with displays, set-top boxes, laptops, audio video receivers and other product types. Because of this global usage, manufacturers, resellers, integrators and consumers must be assured that their HDMI® products work seamlessly together and deliver the best possible performance by sourcing products from licensed HDMI Adopters or authorized resellers. For HDMI Cables, consumers can look for the official HDMI® Cable Certification Labels on packaging. Innovation continues with the latest HDMI 2.2 Specification that supports higher 96Gbps bandwidth and next-gen HDMI Fixed Rate Link technology to provide optimal audio and video for a wide range of device applications. Higher resolutions and refresh rates are supported, including up to 12K@120 and 16K@60. Additionally, more high-quality options are supported, including uncompressed full chroma formats such as 8K@60/4:4:4 and 4K@240/4:4:4 at 10-bit and 12-bit color.
—

The laptop hardware runs the Windows operating system, marking a transition from the Ubuntu-based DGX OS environment used in previous small-form-factor iterations. This OS transition enables developers and creators to access the broader Windows application ecosystem while utilizing the unified memory architecture. The platform supports dynamic power profile adjustments, allowing users to configure performance modes through the operating system to balance thermal output, battery life, and processing throughput.

During the technical demonstration, the system ran the Unreal Engine Editor to showcase its performance in real-time 3D asset rendering and game development workflows. Historically, mobile workstations have been constrained by separate GPU and CPU memory pools, often limiting dedicated graphics memory to 16 GB or 24 GB. The 128 GB unified memory pool in this prototype allows the GPU to access the entire memory allocation dynamically, eliminating the need to optimize scenes or reduce texture resolutions to fit within narrow hardware limits.

To highlight this capability, the system loaded a 3D city scene totaling approximately 70 GB directly into the unified memory. This setup allows developers to navigate large environments smoothly without experiencing loading screens or latency from transferring assets between system memory and GPU memory. The unified memory architecture enables real-time editing and viewport navigation of complex geometry and high-resolution textures at scale.

The integration of unified memory and Windows OS also opens up possibilities for local developer workflows involving agentic AI applications. Developers can run agentic frameworks to interact with local Windows applications directly on the hardware, bypassing the OS-level application limitations of previous Linux-only setups. This configuration supports tasks ranging from automated gameplay testing in QA environments to complex game asset generation and content creation pipelines.

source https://www.youtube.com/watch?v=Ps61Suj5mik

E Ink demonstrated its latest e-paper display technologies at Computex 2026, showcasing the versatile applications of its digital paper platforms. Among the primary exhibits was the Prism e-paper platform, which uses tiled segmented displays to create custom pixelated patterns and images. These tiled segments allow for highly configurable layouts without size limitations, enabling public installations similar to those deployed at the Taipei Zoo.

—
HDMI® Technology is the foundation for the worldwide ecosystem of HDMI-connected devices; integrated with displays, set-top boxes, laptops, audio video receivers and other product types. Because of this global usage, manufacturers, resellers, integrators and consumers must be assured that their HDMI® products work seamlessly together and deliver the best possible performance by sourcing products from licensed HDMI Adopters or authorized resellers. For HDMI Cables, consumers can look for the official HDMI® Cable Certification Labels on packaging. Innovation continues with the latest HDMI 2.2 Specification that supports higher 96Gbps bandwidth and next-gen HDMI Fixed Rate Link technology to provide optimal audio and video for a wide range of device applications. Higher resolutions and refresh rates are supported, including up to 12K@120 and 16K@60. Additionally, more high-quality options are supported, including uncompressed full chroma formats such as 8K@60/4:4:4 and 4K@240/4:4:4 at 10-bit and 12-bit color.
—

The company detailed the technical specifications of its Marquee platform, a robust color e-paper solution designed specifically for outdoor signage. Built on an architecture similar to E Ink’s black-and-white displays, the Marquee platform incorporates four color particles within its microcapsule system. By manipulating these particles using physics and chemistry, the display achieves a wide color gamut. Because of the multi-particle mixture, update times are slower than traditional monochrome displays, making the platform ideal for static outdoor signage rather than dynamic consumer e-readers or e-notes.

For indoor signage and consumer devices, E Ink presented the Spectra series, which delivers higher color contrast and a wider color gamut compared to the outdoor-oriented Marquee platform. The Spectra displays are optimized for indoor environments where durability requirements are lower but visual pop is critical. Displays on showcase included 40.5-inch and 75-inch formats, which are targeted at digital picture frame manufacturers and commercial indoor poster applications. E Ink collaborates with partners such as PocketBook to integrate e-paper displays into consumer portfolios.

Automotive integration remains a key development area for E Ink, highlighted by its ongoing collaboration with BMW. The booth featured a conformable, color-changing vehicle prototype representing fourth-generation e-paper technology. While the conformable display remains in the prototype stage, a production-ready black-and-white hood design has been integrated into the BMW iX Flow. This hood, which displays a Shanghai skyline pattern first shown at Auto China, uses a seamless, wrap-like application over the vehicle panels and wheels, with pre-orders scheduled to open later this year.

E Ink also displayed its Kaleido series, featuring a color filter printed directly on top of the black-and-white particle platform. The Kaleido series supports faster refresh rates, achieving up to 11 frames per second to enable dynamic video playback. This allows signage to combine static high-resolution content on the upper portion with localized, motion-based updates on the lower section. The technology integrates active matrix and segmented display designs, allowing systems to scale from individual 40.5-inch panels up to multi-tile outdoor billboards.

source https://www.youtube.com/watch?v=tvsXZqPp2fs

Nvidia RTX Spark Gaming demonstrated the RTX Spark processor integrated into a Microsoft Surface Laptop Ultra at Computex 2026. This platform showcases a playable AAA gaming demo running on ARM-based silicon. The demonstration highlights seamless execution of gaming workloads, utilizing X86-to-ARM emulation and translation layers to ensure day-one compatibility for major game titles. The engineering focus centers on delivering smooth gameplay experiences on portable form factors.

—
HDMI® Technology is the foundation for the worldwide ecosystem of HDMI-connected devices; integrated with displays, set-top boxes, laptops, audio video receivers and other product types. Because of this global usage, manufacturers, resellers, integrators and consumers must be assured that their HDMI® products work seamlessly together and deliver the best possible performance by sourcing products from licensed HDMI Adopters or authorized resellers. For HDMI Cables, consumers can look for the official HDMI® Cable Certification Labels on packaging. Innovation continues with the latest HDMI 2.2 Specification that supports higher 96Gbps bandwidth and next-gen HDMI Fixed Rate Link technology to provide optimal audio and video for a wide range of device applications. Higher resolutions and refresh rates are supported, including up to 12K@120 and 16K@60. Additionally, more high-quality options are supported, including uncompressed full chroma formats such as 8K@60/4:4:4 and 4K@240/4:4:4 at 10-bit and 12-bit color.
—

The core silicon powering the RTX Spark is the GB10 chip, sharing the same underlying architecture as the enterprise-grade DGX systems but adapted for consumer laptop form factors. The system-on-chip design features the CPU and GPU side-by-side on a single die, linked by a high-speed interconnect. This close integration allows for a shared memory pool, which is demonstrated through a real-time Unreal Engine city creation demo where users can zoom in and out of complex environments without memory bottlenecks.

Operating within a laptop form factor requires careful management of thermal limits, battery capacity, and power delivery. The GB10 platform implements dynamic scaling of power usage, featuring a lower power mode for battery preservation and a boost mode to elevate performance during demanding tasks. The side-by-side die configuration optimizes electrical efficiency, enabling partner manufacturers to design thinner, faster laptops without compromising gaming performance.

At launch, the RTX Spark platform officially supports the Windows on ARM operating system. While developer-focused DGX platforms run Linux, this consumer-oriented rollout focuses on optimizing Windows applications and games. Nvidia RTX Spark Gaming engineers are working to refine the translation and emulation layers, ensuring that X86 legacy gaming titles execute reliably on the ARM desktop and laptop architectures.

The broader ecosystem for this silicon platform includes multiple hardware partners preparing a variety of form factors, including mini PCs and laptops. These partner systems are scheduled to launch in the fall of 2026. The co-developed silicon architecture, leveraging a unified GPU and CPU design, aims to establish a new performance tier for ARM-based personal computing devices.

source https://www.youtube.com/watch?v=x1n-FPqMeWg

Sunix presented its range of industrial connectivity, IoT, and edge AI hardware at Computex 2026. Established in 1986 and based in Taiwan, the company designs custom integrated circuits for serial and CAN bus communication, utilizing foundry resources from TSMC and UMC. Sunix emphasizes hardware validation and reliability for industrial PC applications, transitioning over nearly four decades from original PCI I/O controllers to PCIe and M.2 formats.

—
HDMI® Technology is the foundation for the worldwide ecosystem of HDMI-connected devices; integrated with displays, set-top boxes, laptops, audio video receivers and other product types. Because of this global usage, manufacturers, resellers, integrators and consumers must be assured that their HDMI® products work seamlessly together and deliver the best possible performance by sourcing products from licensed HDMI Adopters or authorized resellers. For HDMI Cables, consumers can look for the official HDMI® Cable Certification Labels on packaging. Innovation continues with the latest HDMI 2.2 Specification that supports higher 96Gbps bandwidth and next-gen HDMI Fixed Rate Link technology to provide optimal audio and video for a wide range of device applications. Higher resolutions and refresh rates are supported, including up to 12K@120 and 16K@60. Additionally, more high-quality options are supported, including uncompressed full chroma formats such as 8K@60/4:4:4 and 4K@240/4:4:4 at 10-bit and 12-bit color.
—

The company’s edge AI acceleration hardware includes PCIe cards and M.2 modules developed in collaboration with Hailo. These solutions are offered in configurations utilizing one, two, or four Hailo AI processors. Designed for low power consumption, the PCIe accelerators run directly off motherboard slot power without requiring auxiliary power cables. The modules deliver up to 40 TOPS of AI performance, enabling real-time local inference and support for large language models, including 3.8 billion and 7 billion parameter Llama models.

For software deployment, Sunix hardware interfaces with Hailo model conversion utilities to compile neural networks for the Hailo AI system-on-chip. At the exhibition, Sunix demonstrated image analysis and automatic detection capabilities using a video management system powered by Network Optix software. This architecture enables real-time edge-based inferencing on customer devices without relying on cloud resources.

Sunix also showed IoT and computing module solutions, including a System on Module (SOM) based on the MediaTek Genio 720 SoC. This smart module features an integrated neural processing unit delivering up to 10 TOPS, targeting edge AI applications such as digital signage players with HDMI output when integrated into ODM partner systems. Other remote I/O products gather data from serial sensors, AI sensors, and input devices over the internet, utilizing M.2 to digital I/O interfaces.

For high-performance connectivity, Sunix exhibited a USB4 host card certified by the USB4 community. The card supports up to 100W power output, enabling single-cable connectivity for power and display signals to devices such as medical-grade monitors. Additionally, the company provides KFD CAN bus communication cards for integration in specialized systems like autonomous driving buses, where they serve as the central interface for collecting telemetry and operational data.

source https://www.youtube.com/watch?v=xGl6bEEXDrs

Gigabyte AI Top 500 showcased local artificial intelligence training and inference hardware at Computex 2026, centering on the GRX50 motherboard. The platform features eight memory slots supporting up to 2 TB of total memory capacity. In the demonstrated configuration, the board was equipped with 768 GB of system memory utilizing 96 GB modules, paired with NVIDIA RTX 5090 graphics cards to handle demanding AI workloads directly at the desktop.

—
HDMI® Technology is the foundation for the worldwide ecosystem of HDMI-connected devices; integrated with displays, set-top boxes, laptops, audio video receivers and other product types. Because of this global usage, manufacturers, resellers, integrators and consumers must be assured that their HDMI® products work seamlessly together and deliver the best possible performance by sourcing products from licensed HDMI Adopters or authorized resellers. For HDMI Cables, consumers can look for the official HDMI® Cable Certification Labels on packaging. Innovation continues with the latest HDMI 2.2 Specification that supports higher 96Gbps bandwidth and next-gen HDMI Fixed Rate Link technology to provide optimal audio and video for a wide range of device applications. Higher resolutions and refresh rates are supported, including up to 12K@120 and 16K@60. Additionally, more high-quality options are supported, including uncompressed full chroma formats such as 8K@60/4:4:4 and 4K@240/4:4:4 at 10-bit and 12-bit color.
—

The system enables running massive artificial intelligence models locally without requiring specialized industrial power infrastructure. Operating on standard household or office voltage, the baseline configuration consumes approximately 1,600 watts of power. For scaling performance, the platform supports dual power supplies and multiple RTX 5090 GPUs, raising the total power budget to 3,200 watts to handle complex, multi-GPU configurations.

A key hardware integration is the Phison AI TOP 100 SSD, which acts as a memory extension to enable execution of models far exceeding physical GPU VRAM limits. While a standard RTX 5090 setup can run a 20 billion parameter model using its onboard VRAM, combining it with 768 GB of system memory expands capacity to a 400 billion parameter model. Incorporating the specialized Phison AI TOP 100 SSD allows the system to process models with up to 600 billion parameters.

The software environment supporting the hardware allows users to run, evaluate, and test a wide variety of open-source models, including Llama 3.2 3B, DeepSeek 671B, Gemma, and Qwen. It supports multiple precision levels such as FP4, FP8, FP16, and FP32, providing a benchmarking framework to help operators determine the optimal balance of speed and accuracy for their specific workflows. In a practical enterprise demonstration, the system processed 80 concurrent insurance consultation queries using Llama 3.2 3B in 22 seconds.

For mainstream applications, mid-range solutions based on AMD and Intel architectures are also available. The Intel mainstream platform features the Z890 chipset and supports dual RTX 5090 graphics cards, running models up to 400 billion parameters for budget-conscious setups. To ensure stability during continuous 24/7 training runs, these systems incorporate thermal enhancements including optimized airflow venting and integrated dust defense mechanisms.

source https://www.youtube.com/watch?v=37vSBaiJKcI

DJI 20260603031949 0009 D CHAR withad showcased its latest AI hardware solutions at Computex 2026, highlighting the flagship ARTOP 500 system designed to run large language models locally on standard office power. The ARTOP 500 is built on the GRX50 motherboard, which features memory slots supporting up to 2TB of memory. In the demonstrated configuration, the motherboard was equipped with 768GB of memory consisting of eight 96GB modules. The system integrates NVIDIA RTX 5090 GPUs, supporting up to two of these graphics cards to handle demanding computational workloads. For alternative configurations, the platform is also compatible with AMD hardware, including the AMD Ryzen 9700X processor.

—
HDMI® Technology is the foundation for the worldwide ecosystem of HDMI-connected devices; integrated with displays, set-top boxes, laptops, audio video receivers and other product types. Because of this global usage, manufacturers, resellers, integrators and consumers must be assured that their HDMI® products work seamlessly together and deliver the best possible performance by sourcing products from licensed HDMI Adopters or authorized resellers. For HDMI Cables, consumers can look for the official HDMI® Cable Certification Labels on packaging. Innovation continues with the latest HDMI 2.2 Specification that supports higher 96Gbps bandwidth and next-gen HDMI Fixed Rate Link technology to provide optimal audio and video for a wide range of device applications. Higher resolutions and refresh rates are supported, including up to 12K@120 and 16K@60. Additionally, more high-quality options are supported, including uncompressed full chroma formats such as 8K@60/4:4:4 and 4K@240/4:4:4 at 10-bit and 12-bit color.
—

The ARTOP 500 system enables local deployment of extremely large AI models, supporting parameter sizes up to 600 billion. When relying solely on the VRAM of the NVIDIA RTX 5090 GPU, the system can run a 20 billion parameter model. By utilizing the 768GB of system memory, the capacity expands to handle models up to 400 billion parameters. To achieve the maximum 600 billion parameter capability, the hardware is bundled with an ARTOP 100 1TB SSD, leveraging storage to run larger language models that would otherwise require data center infrastructure.

A key design aspect of the ARTOP 500 is its power efficiency and deployment simplicity. The hardware operates on standard household and office voltage without the need for specialized power converters. The baseline configuration draws 1600W of power, while adding a second graphics card utilizes an additional power supply unit to bring the total capacity to 3200W. This standard voltage compatibility allows enterprises to deploy large language models directly on office desks or within standard workspace environments, avoiding the need for dedicated server room infrastructure or high-voltage power installations.

To demonstrate real-world performance, the platform was showcased running a demo kit tailored for an insurance industry use case. The test simulated 80 simultaneous users submitting unique queries to the Llama 3.2 3B model. The ARTOP 500 processed all 80 concurrent chat requests and returned responses in 22 seconds. The system has validated over 100 language models, including DeepSeek V1 671B, Llama, Gemma, and Kimi K2.5. The testing suite also allows users to evaluate model speeds across multiple precision levels, including FP4, FP8, FP16, and FP32, to optimize performance.

For mainstream and mid-tier deployments, the product lineup includes the ARTOP 400s, which features compatibility with both AMD and Intel chipsets, including the Intel 890 chipset. The ARTOP 400s supports up to two NVIDIA RTX 5090 GPUs and runs models up to 400 million parameters. To ensure reliability during continuous 24/7 AI model training and inference workloads, the chassis design features a specialized ventilation door to optimize airflow across key chipsets, along with integrated dust filters to prevent contamination and maintain thermal stability.

source https://www.youtube.com/watch?v=O3AzI64yPq4

Gigabyte AI Top Atom is a compact personal AI supercomputer designed to enable local AI model training, fine-tuning, and inference. Built around the NVIDIA GB10 Grace Blackwell Superchip, each individual hardware unit delivers up to one petaflop of FP4 computing power and features 128GB of coherent unified LPDDR5X memory. At Computex 2026, the platform demonstrated its clustering capabilities by connecting four units together to form a unified four-petaflop workstation system.

—
HDMI® Technology is the foundation for the worldwide ecosystem of HDMI-connected devices; integrated with displays, set-top boxes, laptops, audio video receivers and other product types. Because of this global usage, manufacturers, resellers, integrators and consumers must be assured that their HDMI® products work seamlessly together and deliver the best possible performance by sourcing products from licensed HDMI Adopters or authorized resellers. For HDMI Cables, consumers can look for the official HDMI® Cable Certification Labels on packaging. Innovation continues with the latest HDMI 2.2 Specification that supports higher 96Gbps bandwidth and next-gen HDMI Fixed Rate Link technology to provide optimal audio and video for a wide range of device applications. Higher resolutions and refresh rates are supported, including up to 12K@120 and 16K@60. Additionally, more high-quality options are supported, including uncompressed full chroma formats such as 8K@60/4:4:4 and 4K@240/4:4:4 at 10-bit and 12-bit color.
—

To enable high-speed clustering, the hardware uses an NVIDIA ConnectX-7 network interface card connected to a high-bandwidth switch. This configuration supports 200 GbE bandwidth using fiber ethernet cables terminated with QSFP connectors to ensure low latency and high throughput. Although GIGABYTE officially announces cluster configurations of up to four units, the networking infrastructure allows scaling to eight units if an eight-port switch is deployed, enabling even larger model footprints.

The system is designed to run large language models and other complex computational tasks locally. A single Gigabyte AI Top Atom box can run models with up to 120 billion parameters, while a clustered four-unit setup scales the capacity to handle models containing up to 480 billion parameters. Users can choose to cluster the units together to operate as a single supercomputer or partition them to run separate AI agent workloads concurrently.

A key application demonstrated at the event is the autonomous orchestration of thermal simulations using the open-source OpenClaw agent framework. In this workflow, one Gigabyte AI Top Atom acts as the commander agent running OpenClaw, which manages and automates the simulation tasks. The commander distributes workload instructions to three subordinate units running GROMACS molecular dynamics simulation software to compute thermal interface material properties for 1000W Blackwell processors.

By automating over 100 iterative simulation runs, the OpenClaw agent optimizes the material formulas without manual intervention. This distributed workflow reduces simulation times from a full week on a single unit to just two days across three units, representing at least a threefold speedup. The resulting recipe optimized the heat dissipation channels, converting them into straight lines that allow heat to escape the chip efficiently and improving thermal performance from 22 to 4.2.

source https://www.youtube.com/watch?v=A5UYT5evi9w

Gigabyte presented the WRX90 AI Top motherboard at Computex 2026, designed for local workstation-class AI training and inference. The board features the AMD WRX90 chipset and supports AMD Ryzen Threadripper PRO 9000 and 7000 WX-Series processors. It acts as a foundation for multi-GPU configurations, facilitating memory expansion and high-speed data transmission for large language models and other compute-intensive workloads.

—
HDMI® Technology is the foundation for the worldwide ecosystem of HDMI-connected devices; integrated with displays, set-top boxes, laptops, audio video receivers and other product types. Because of this global usage, manufacturers, resellers, integrators and consumers must be assured that their HDMI® products work seamlessly together and deliver the best possible performance by sourcing products from licensed HDMI Adopters or authorized resellers. For HDMI Cables, consumers can look for the official HDMI® Cable Certification Labels on packaging. Innovation continues with the latest HDMI 2.2 Specification that supports higher 96Gbps bandwidth and next-gen HDMI Fixed Rate Link technology to provide optimal audio and video for a wide range of device applications. Higher resolutions and refresh rates are supported, including up to 12K@120 and 16K@60. Additionally, more high-quality options are supported, including uncompressed full chroma formats such as 8K@60/4:4:4 and 4K@240/4:4:4 at 10-bit and 12-bit color.
—

The motherboard features seven native PCIe Gen 5 slots on the board. Six of these slots operate at full x16 bandwidth, while one slot operates at x8 bandwidth. An integrated MCIO 8i connector allows expansion for two additional GPUs, bringing the total supported GPUs to nine, or can be used for extra M.2 storage expansion. This storage architecture supports one PCIe Gen 5 M.2 SSD and two PCIe Gen 4 M.2 SSDs, which can be configured in RAID.

To manage the thermal output of high-density hardware, the board utilizes an active cooling system. This includes a cooling tower with two active fans to cool up to eight ECC RDIMM slots, supporting up to 2 TB of memory. The VRM features a 32+3+3 power phase design with dedicated active cooling, ensuring stable power delivery to the CPU under sustained loads. The AMD WRX90 chipset and primary M.2 slots are also covered by active thermal cooling.

For network connectivity and system management, the motherboard is equipped with dual 10GbE LAN ports. System administrators can utilize the integrated ASPEED AST2600 BMC controller for server-grade remote management via IPMI. The platform also supports USB4 connectivity and DisplayPort input/output configurations.

Power delivery is managed through support for dual power supply units, allowing users to connect multiple PSUs to satisfy the power demands of multiple graphics cards. A single 1600W power supply unit with native 16-pin PCIe 5.0 power connectors can support up to four GPUs, while adding a second unit enables full multi-GPU deployments. The physical layout of the motherboard is optimized for integration into dedicated workstation cases or 5U rackmount servers.

source https://www.youtube.com/watch?v=MKz4Qfwc994