GIGABYTE Advances Local AI Infrastructure with WRX90 AI TOP Motherboard

GIGABYTE announced the WRX90 AI TOP workstation motherboard, supporting AMD Ryzen™ Threadripper™ PRO series processors. Engineered specifically for local AI workloads and intensive computing, the motherboard features multi-GPU PCIe 5.0 expansion slots, high-capacity ECC memory support, and enterprise-grade BMC remote management. It provides a robust, highly extensible hardware foundation for professionals to deploy dedicated local AI platforms and manage model development on-premises.

Scale Up for AI Development

To support evolving AI applications, the WRX90 AI TOP allows organizations to configure and expand their hardware as computational workloads grow. Equipped with seven PCIe 5.0 x16 slots and two MCIO 8i connectors, the motherboard offers extensive bandwidth capable of supporting up to nine GPU devices through compatible expansion hardware. Complementing this expansion capacity, eight-channel DDR5 memory support accommodates up to 2 TB of ECC RDIMM for large datasets and memory-intensive applications. The inclusion of Error-Correcting Code (ECC) memory automatically detects and corrects data corruption, ensuring data integrity during prolonged AI computing sessions.

Engineered for Reliability

Sustained performance requires a foundation engineered for heavy continuous loads. The WRX90 AI TOP features a 32+3+3+3-phase digital power design paired with a 14-layer server-grade PCB and a full-metal backplate, ensuring stable power delivery, signal integrity, and structural rigidity. Thermal engineering plays a vital role in preventing throttling under sustained operation. To tackle heat buildup around closely spaced memory modules, the included DDR Wind Blade XTREME Duo fans improve airflow to reduce average memory temperatures by up to 43°C compared to unassisted setups. Dedicated SYS Wind Blade fans and extensive heatsinks cover the power delivery and chipset areas, maintaining optimal thermal control during heavy processing runs.

Manage Smarter. Connect Faster.

For streamlined operational management, the WRX90 AI TOP integrates an onboard BMC and a dedicated IPMI LAN port. This enables IT administrators to perform remote hardware monitoring, power control, and troubleshooting at any time. These capabilities significantly simplify maintenance workflows for IT teams while ensuring long-term operational stability for local computing environments. Additionally, dual server-grade 10GbE LAN ports enable rapid data transfers to centralized storage arrays without the need for an add-in network card, while dual USB4® Type-C ports offer high-speed connectivity for external storage and peripherals.

The WRX90 AI TOP expands GIGABYTE’s AI TOP portfolio with a workstation motherboard that bridges extreme computing power with practical management tools. Its combination of expandable architecture and durable hardware design empowers enterprise users and developers to build systems optimized for long-term AI development.

Read More

Cadence Unveils Maestro Domain-Specific AI Models to Drive Recursive, Agentic Chip Design

Cadence announced the Maestro™ AI Model Portfolio, which powers Cadence® AI Super Agents across semiconductor and system design workflows. By integrating these models with reinforcement learning (RL) loops and its proven electronic design automation (EDA) and system design automation (SDA) tools, Cadence delivers a deeply integrated agentic engineering platform designed to improve results, accelerate time to market and continuously enhance performance through recursive self-improvement.

The Maestro technology offers customers a secure, Cadence-managed platform that combines exceptional domain-specialized intelligence and expertise with the control, choice, and flexibility they need to innovate with confidence. The Maestro portfolio spans multiple model sizes and architectures, combining domain-specialized post-trained large language models (LLMs), trained decision models and self-improving agent skills. Integrated with Cadence AI Super Agents—including AuraStack™, ChipStack™, InnoStack™, and ViraStack™ AI Super Agents—as well as its proven EDA and SDA tools and reinforcement learning, the Maestro portfolio brings specialized intelligence to complex engineering workflows through an optimal blend of frontier, open-weight and domain-specialized models across cloud and on-premises environments.

“Revolutionizing complex engineering tasks requires deep domain-specialized intelligence,” said Anirudh Devgan, president and CEO at Cadence. “Combining our Maestro portfolio of domain-specialized models with leading frontier and open-weight models enables our AI Super Agents to drive Cadence EDA and SDA tools with exceptional skill and efficiency to transform chip and system design.”

“Agentic chip design requires AI that can generate design content, evaluate tradeoffs and learn from verified engineering results,” said Tim Costa, vice president and general manager of industrial and computational engineering at NVIDIA. “Our collaboration with Cadence brings NVIDIA Nemotron, OpenShell and deep engineering expertise to Maestro’s domain-specialized models. By grounding reinforcement learning in feedback from EDA tools and deterministic solvers, we can teach agents to make better design decisions based on how well their work meets engineering objectives and constraints.”

“We are deploying Cadence’s InnoStack AI Super Agent into our production design environment,” said Hisato Yoshida, COO of Socionext. “By combining agentic AI with advanced EDA tools for physical design, we expect up to 5X overall productivity gains while maintaining design quality. We see the Maestro portfolio as the next step in driving even greater efficiency across our design process.”

“Ultra-low power devices require innovation across disciplines. Agentic AI unlocks new power opportunities,” said Garima Sharda, vice president of engineering at Silicon Labs. “By leveraging the Cadence Verisium Debug as part of ChipStack AI Super Agent, our engineering teams have achieved acceleration in root-cause analysis and end-to-end debug during pre-silicon low power verification. We look forward to realizing further benefits through Cadence’s Maestro domain-specialized models.”

A Self-Improving Learning Environment for Engineering AI

Initial results show up to 30% improvement in quality of results and up to 5X greater token efficiency versus a general-purpose foundation model. These gains come from a vertically integrated learning environment combining Cadence domain knowledge, curated engineering datasets, AI Super Agents, EDA and SDA tools, and evaluation harnesses. Cadence’s principled algorithms and deterministic solvers evaluate outputs against engineering objectives, constraints and ground truth, creating closed and verified feedback loops for reinforcement learning and recursive improvement.

Domain Intelligence with Intelligent Model Choice

The Cadence learning environment unifies complementary AI capabilities in a single workflow. Frontier and open-weight models provide broad language, reasoning and multimodal capabilities; the Maestro AI Model Portfolio provides domain expertise for complex chip and system design. Given a high-level objective, Cadence AI Super Agents interpret design intent, decompose tasks, orchestrate the Maestro portfolio with other leading frontier and open-weight models, and drive Cadence EDA and SDA tools. The combined solution interprets specifications, generates design content, analyzes results, evaluates tradeoffs and guides decisions based on engineering requirements, cost and security. The Maestro portfolio, including NVIDIA Nemotron and other models, can be further post-trained on a customer’s proprietary data, designs, and workflows and, by balancing EDA and SDA tool calls with LLM invocations, it significantly reduces token usage while maintaining execution efficiency.

The complete Cadence agentic environment—including AI Super Agents, Maestro AI Model Portfolio and EDA tools—can run on Cadence’s secure cloud platform, secure public clouds, or in a customer-controlled on-premises environment to meet data privacy, security and deployment requirements. Cadence and NVIDIA are also working together on agent security through integration with the NVIDIA OpenShell Reference Architecture

Availability

The first Maestro models are available to select early access customers in the fourth quarter of 2026.

Read More

KLEVV Reimagines Award-Winning Flagship Heatsink in Black-and-Gold CRAS V RGB PRIME DDR5

KLEVV has debuted the CRAS V RGB PRIME DDR5 series, reinterpreting its Red Dot and iF Design Award-winning aluminum heatsink with a striking black-and-gold finish. The chassis utilizes hard-anodized treatment and precision surface finishing to produce a piano-black luster paired with metallic gold accents, creating an eye-catching aesthetic centerpiece for bespoke enthusiast PC builds. With broad QVL validation across major motherboard vendors and full synchronization support for ASUS, GIGABYTE, MSI, and ASRock RGB ecosystems, the new flagship memory combines premium thermal dissipation with refined visual luxury ahead of its November 2026 commercial rollout.

Refined Flagship Design and Craftsmanship

Building on the aluminum heat spreader design of the flagship CRAS V RGB, winner of both the Red Dot and iF Design Awards, the CRAS V RGB PRIME introduces a striking new black-and-gold interpretation of the series' signature look.

An advanced hard-anodizing process, combined with a specialized surface treatment, gives the modules a smooth, piano-lacquer-like sheen that adds depth and sophistication from every angle. The result is an understated yet luxurious aesthetic that offers users a fresh, premium design choice while preserving the refined craftsmanship of the original CRAS V RGB.

High-Speed Performance with Lower Latency

The CRAS V RGB PRIME further advances KLEVV's DDR5 overclocking memory lineup with tested speeds of 6000, 6400, and 8000 MT/s and CAS latencies of CL26, CL28, and CL38, respectively.

The 6000 MT/s configuration features tested timings of 26-36-36-76, while the 6400 MT/s configuration delivers timings of 28-38-38-78. For users seeking extreme frequencies, the 8000 MT/s models operate at tested timings of 38-48-48-128. All configurations are tested at 1.4V.

The 6000 and 6400 MT/s configurations are available in dual-channel kits ranging from 32GB (16GBx2), 64GB (32GBx2), to 128GB (64GBx2), while the 8000 MT/s models are offered in both 32GB (16GBx2) and 48GB (24GBx2) kits. This diverse range gives users greater flexibility to select the ideal balance of speed, latency, and capacity for their systems.

 Next-Generation AMD EXPO™ Technology: Featuring Ultra Low Latency

“AMD EXPO™ Technology: Featuring Ultra Low Latency” is the latest evolution of AMD's established memory overclocking technology. The new AMD EXPO ULL profile incorporates optimized memory sub-timing settings designed to reduce latency and improve overall performance on compatible AMD platforms. As the first product in KLEVV's overclocking memory lineup to adopt this technology, the CRAS V RGB PRIME delivers a more responsive gaming and overclocking experience.

The 6000 MT/s and 6400 MT/s 32GB (16GBx2) configurations are certified for AMD EXPO Ultra Low Latency. According to AMD, certified Ultra Low Latency memory can deliver an average 4% improvement in FPS compared with non-ULL AMD EXPO memory on compatible systems.*

Broad Intel and AMD Platform Support

Offering broad support across the latest desktop platforms, the CRAS V RGB PRIME is compatible with AMD Ryzen™ processors on the AM5 platform and Intel® Core™ Ultra 200S Plus processors.

Each kit comes preloaded with both Intel® Extreme Memory Profile and AMD EXPO™ Technology overclocking profiles, allowing users to unlock high-speed performance through a simple BIOS setting on compatible motherboards with supported BIOS versions (subject to supported BIOS versions). With both profiles included, users can confidently choose their preferred speed and capacity without the hassle of selecting separate Intel- or AMD-specific models.

Premium Components for Uncompromising Stability

The CRAS V RGB PRIME features rigorously screened, high-quality memory ICs from a leading Korean manufacturer, ensuring consistent high-speed performance and exceptional stability under demanding workloads.

Paired with a high-density 10-layer PCB, the modules deliver enhanced signal integrity and greater resistance to interference during high-speed operation. Integrated PMIC and on-die ECC technologies further support stable power management and long-term reliability, creating a robust foundation for gaming and overclocking performance.

To ensure broad compatibility, the CRAS V RGB PRIME has also been validated on the QVLs (Qualified Vendor Lists) of four major motherboard manufacturers. Its RGB lighting supports ASUS Aura Sync, GIGABYTE RGB Fusion, MSI Mystic Light Sync, and ASRock Polychrome Sync, allowing users to seamlessly coordinate lighting across compatible components and systems.

A Complete DDR5 Gaming/OC Memory Lineup

The introduction of the CRAS V RGB PRIME further expands KLEVV's comprehensive DDR5 Gaming/OC memory portfolio.

With the CRAS, URBANE, BOLT, and FIT series available across a wide range of designs and specifications, KLEVV continues to provide flexible memory choices for different performance requirements, system configurations, and aesthetic preferences.

Warranty and Availability

The CRAS V RGB PRIME DDR5 comes with a limited lifetime peace of mind warranty and will be available from November 2026. Please visit the KLEVV website to check product availability across online and offline retail channels. 

Read More

Meta Integrates NVIDIA Vera Rubin, BlueField-4, and OpenShell to Scale Muse Agent Workloads

Deploying autonomous AI agents requires fundamentally different infrastructure than traditional batch inference or autoregressive text generation. Autonomous agents run interactive, multi-step tool-use loops, query external databases, and handle persistent memory states. To support its next-generation Muse family of models and agents, Meta has entered an expanded co-engineering partnership with NVIDIA to build out an end-to-end silicon, networking, and software stack optimized for performance-per-watt and total cost of ownership (TCO).

Compute Fabric: Vera CPUs and Rubin Architecture

At the compute layer, Meta is expanding deployments of NVIDIA's upcoming Vera Rubin platform:

• Vera CPUs: Designed for high-efficiency general compute and host orchestration, offloading control flow, data serialization, and agent prompt dispatching from the primary accelerators.

• Rubin GPUs: Deliver next-generation tensor compute capabilities tailored for complex reasoning, multimodal input tokenization, and dynamic agentic branching.

• Workflow Optimization: Joint engineering teams will optimize the compute stack across both model training runs and high-concurrency inference pipelines, balancing compute density against energy consumption.

Network Topology: Spectrum-X and BlueField-4 DPUs

Agent workflows generate bursty, non-deterministic inter-node traffic as distributed agents query disparate memory stores and reasoning engines. To prevent tail latency from degrading user response times, Meta is integrating:

• NVIDIA Spectrum-X™ Ethernet: Delivers lossless, high-throughput, adaptive-routing Ethernet infrastructure specifically tailored for generative AI clusters.

• NVIDIA BlueField-4® DPUs: Offload network virtualization, in-transit telemetry, and cryptographic overhead directly from the host system, maximizing effective network bandwidth and ensuring deterministic packet delivery across massive agent fabrics.

Agency Security: NVIDIA OpenShell and Sentry

Beyond pure hardware throughput, agent safety poses unique execution risks because agents execute code, call APIs, and parse untrusted inputs. The collaboration addresses this at the systems level through:

• NVIDIA OpenShell Reference Architecture: A framework establishing sandboxed runtime environments, ensuring AI agents operate within isolated, strictly enforced privilege boundaries.

• NVIDIA Sentry: Real-time runtime monitoring that inspects structured data, unstructured streams, and knowledge graph queries to detect rogue agent behaviors, policy drift, or adversarial prompt injections before execution.

Read More

ASUS Launches ProArt P14 and P16 with NVIDIA RTX Spark in Malaysia, Starting at RM18,999

ASUS Malaysia has officially commenced sales for the ProArt P14 (H7407) and ProArt P16 (H7607), bringing NVIDIA's new ARM-based RTX Spark N1X computing platform to local creators, developers, and power users.

Following an exclusive pre-order run for the maxed-out 128GB Unified Memory configuration earlier this month, ASUS has added a 48GB unified memory tier, bringing the starting price down to RM18,999 for both the 14-inch and 16-inch form factors.

Key Specifications & Malaysian Pricing:

• ProArt P14 / ProArt P16 (48GB LPDDR5X + 1TB SSD): RM18,999

• ProArt P14 / ProArt P16 (128GB LPDDR5X + 1TB SSD): RM29,999

• Colors: Nano Black, Neo White (Anti-smudge coating)

• Display:

  • P14: 14-inch 3K Lumina Pro OLED (2880 × 1800), 120Hz, 1,100 nits peak HDR.

  • P16: 16-inch 4K UHD+ Lumina Pro OLED (3840 × 2400), 120Hz, G-Sync, 1,600 nits peak HDR.

• Chassis: 1.48kg / 13.9mm thin (P14); 1.77kg / 12.9mm thin (P16).

Availability & Launch Bundles: The 48GB configuration is available immediately across authorized ASUS retail dealers in Malaysia and the ASUS Official e-Store, while the 128GB version remains exclusive to the e-Store. Early purchases made before November 30, 2026 include a complimentary bundle featuring a ProArt monitor (PA279CRV for the 48GB edition; PA279CDV OLED for the 128GB edition), a ProArt Mouse MD301, and bundled software licenses covering Microsoft Office Home 2024 and 3 months of Adobe Creative Cloud.

Read More