M5Stack Unit PoE-P4 Pairs RISC-V ESP32-P4 and 802.3at PoE in 64mm Module
M5Stack unveiled the Unit PoE‑P4, a 64 × 24 mm module that combines Espressif’s ESP32‑P4 RISC‑V SoC with IEEE 802.3at Power over Ethernet. The board packs 16 MB Flash, 32 MB PSRAM, a 10/100 Mbps Ethernet PHY, dual MIPI DSI/CSI interfaces, and dual USB‑C ports, delivering up to 6 W of PoE power. Development is supported through Arduino and ESP‑IDF, with full schematics and firmware available online. The unit is priced at $21.50, positioning it as a low‑cost solution for edge AI and IoT deployments.
Xanadu and Mitsubishi Chemical Detail Quantum Algorithms for EUV Semiconductor Research
Xanadu Quantum Technologies and Mitsubishi Chemical have published a pre‑print detailing quantum simulation algorithms aimed at improving extreme ultraviolet (EUV) lithography, a key step in advanced semiconductor manufacturing. The approach models the quantum‑driven radiation‑induced blur that limits pattern fidelity, using...
AMD’s RDNA 5 “AT0”: Enthusiast GPU with the Handbrake On?
AMD is rumored to be developing a high‑end RDNA 5 GPU codenamed AT0, featuring up to 96 compute units, a 512‑bit memory interface, and possibly 32 GB of VRAM. The chip would mark a return to the $1,500‑plus enthusiast segment that AMD...
Galaxy S26 in the SoC Duel: Exynos 2600 versus Snapdragon 8 Elite Gen 5 with Clear CPU Leadership for Qualcomm
Samsung’s Galaxy S26 ships with two flagship SoCs: the new Exynos 2600, its first 2‑nanometer GAA chip, and Qualcomm’s Snapdragon 8 Elite Gen 5. Early Geekbench 6 results show Snapdragon leading CPU performance by roughly 18 % in single‑core and modestly ahead in multi‑core scores. Samsung’s Xclipse 960 GPU,...
NVIDIA Withdraws GeForce Driver 595.59; Black Screens, Crashes, and Fan Bugs Cause Damage to Trust
NVIDIA pulled the GeForce Game Ready Driver 595.59 after users reported black screens, system crashes, unstable clock speeds, and fan‑control failures, especially on RTX 50 series GPUs. The driver was intended to optimize Resident Evil Requiem and embed DLSS 4 MFG and Ray...
NVIDIA Positions Groq as It Once Did Mellanox, with LPU Set to Become the Latency Weapon in the Inference War
NVIDIA announced a non‑exclusive agreement with Groq valued at up to $20 billion, positioning the startup’s low‑latency processing units (LPUs) as a core component of its inference strategy. CEO Jensen Huang likened the integration to the earlier Mellanox acquisition, suggesting LPUs...
LXD 6.7 Released With AMD GPU Passthrough Support
Canonical released LXD 6.7, the latest update to its container and virtual‑machine manager for Ubuntu. The release introduces AMD GPU passthrough support using the new AMD CDI interface and a gpu_cdi_amd extension. It also upgrades VM GPU passthrough with newer QEMU...
Liquid Metal Droplets Fuse Themselves Into Stretchable Circuits
Researchers at Qingdao University and China University of Petroleum discovered that liquid‑metal droplets can self‑sinter during ordinary solvent evaporation, using a Marangoni‑driven surface‑tension gradient between ethanol and toluene. The process creates a Janus film with a conductive liquid‑metal‑rich layer and...
AI Drives Strong Semiconductor Market in 2025-2026
The global semiconductor market surged to $792 billion in 2025, a 25.6% year‑over‑year increase—the strongest since the COVID‑recovery spike in 2021. AI demand propelled Nvidia’s revenue up 65% and drove a collective 29% revenue rise among major memory makers. While most...
ORNL Launches the Next-Generation Data Centers Institute
Oak Ridge National Laboratory has launched the Next Generation Data Centers Institute (NGDCI) to tackle the exploding electricity demand of AI‑driven data centers. The institute will combine ORNL’s strengths in energy science, high‑performance computing, cybersecurity and grid research to make...
Broadcom Ships 3.5D Face-to-Face Compute SoC
Broadcom announced the start of shipments for the industry’s first 2nm custom compute SoC built on its 3.5D eXtreme Dimension System in Package (XDSiP) platform. The chip leverages Face‑to‑Face (F2F) 3D integration to combine 2.5D and 3D‑IC techniques, delivering unprecedented...
CIQ Launches RLC Pro Enterprise Linux Subscription for Production AI and HPC Environments
CIQ announced Rocky Linux from CIQ Pro (RLC Pro), a commercially supported Enterprise Linux subscription that bundles long‑term support, FIPS 140‑3 validation, indemnification and direct bug‑fix services. The offering targets AI, high‑performance computing and regulated industries, promising 3‑5 year lifecycle stability without forced...
How to Securely Erase an Old Hard Drive on macOS Tahoe
Apple’s macOS 26 Tahoe no longer offers the historic Security Options button in Disk Utility, effectively dropping the GUI‑based secure erase feature for spinning‑disk hard drives. The official user guide still references the option, creating confusion for users who need to meet...
Microsoft Updates DirectX Shader Compiler With Improved Vulkan Driver Interoperability
Microsoft released DirectX Shader Compiler version 1.9, adding production support for Shader Model 6.9. The update focuses on substantial SPIR‑V backend enhancements, including improved layout, ABI correctness, expanded type‑system support, and more reliable code generation. These changes tighten interoperability with Vulkan...
Walmart to Begin Selling Onn Smart TVs Running Google TV for the First Time
Walmart is extending its low‑cost Onn brand into smart‑TV territory by introducing Google TV‑powered models for the first time. A Bluetooth certification filing reveals three series—SGQ, MQP and SQ—offering 55‑, 65‑ and 75‑inch 4K screens. Pricing is expected to stay...
Pawsey Expands High-Memory Resources on Setonix for Data-Intensive Workloads
Pawsey Supercomputing Research Centre has doubled the high‑memory CPU nodes on its Setonix system from eight to sixteen, each equipped with dual 64‑core AMD EPYC 7763 processors and 1 TB of shared RAM. The expansion responds to a user‑survey‑identified demand for...
Lenovo Legion Go Fold Concept Is a Portable Gaming PC with a Foldable Display
Lenovo unveiled a Legion Go Fold concept at Mobile World Congress, featuring a foldable pOLED screen that measures 7.7 inches when closed and expands to 11.6 inches unfolded. The device runs on an Intel Core Ultra 7 258V processor with 32 GB LPDDR5x memory and integrated Arc 140V...
Berkeley Lab Explores Thermodynamic Computing for AI
Lawrence Berkeley National Laboratory researchers unveiled a thermodynamic computing prototype that harnesses thermal noise to generate simple images, mimicking generative AI behavior. By exploiting fluctuations instead of suppressing them, the system flips conventional digital logic and operates via Langevin dynamics....
Asus and Dell Introduce Windows 365 Cloud PCs (Basically Mini PCs that Are Dumb Terminals)
Microsoft’s Windows 365 cloud PC platform is expanding as Asus and Dell unveil dedicated mini‑PC terminals designed specifically for the service. The Asus NUC 16 for Windows 365 packs a compact 0.7 L chassis, DDR5 memory, and a 20 Gbps USB‑C port, while the Dell...
How Customized Foundation IP Is Redefining Power Efficiency and Semiconductor ROI
Synopsys is leveraging its Foundation IP platform to let semiconductor designers tailor standard cells, memory compilers, and EDA flows for specific workloads. In hyperscale AI servers, customized IP and optimized flows delivered up to 51% power reduction and a 5%...
Eyeing the Massive HPC Opportunity for Unlocking Agentic AI with Silicon Photonics
Silicon photonics is emerging as a viable alternative to copper interconnects for high‑performance computing (HPC) systems that power agentic AI workloads. Nvidia’s NVL72 uses copper‑based NVLink, but its bandwidth and latency are straining under next‑generation inference demands. French startup Scintil...
Supermicro Debuts High-Density MicroBlade Server Featuring AMD EPYC 4005
Supermicro unveiled its MicroBlade platform, the industry’s first blade server built around AMD’s EPYC 4005 series and delivering the highest compute density on the market. The 6U chassis can host up to 40 nodes, while a standard 48U rack can...
Akeana Partners with Axiomise for Formal Verification of Its Super-Scalar RISC-V Cores
Akeana Inc. has teamed with formal‑verification specialist Axiomise to exhaustively verify its 4 nm super‑scalar RISC‑V test chip, Alpine. Using Axiomise’s CoreProve® framework, the collaboration identified functional bugs and redundant logic within months, enabling RTL optimizations before tape‑out. The formal approach...
An AI-Native Architecture That Eliminates GPU Inefficiencies
A recent MIT Technology Review analysis highlighted the massive energy footprint of generative‑AI inference, prompting a search for more efficient silicon. VSORA responded with a purpose‑built Matrix Processing Unit (MPU) that abandons the GPU‑centric SIMT model in favor of tensor‑level...
HONOR MagicPad 4 Specifications Revealed with Snapdragon 8 Gen 5, 165Hz OLED Display and 4.8mm Thin Design
HONOR unveiled the MagicPad 4, its first tablet to ship with Qualcomm's Snapdragon 8 Gen 5 mobile platform, positioning it against the iPad Pro and Galaxy Tab S series. The 12.3‑inch OLED screen offers a 165 Hz refresh rate, 2,400‑nit peak brightness, and a 4.8 mm thin chassis...
HONOR MagicPad 4 Specifications Revealed with Snapdragon 8 Gen 5, 165Hz OLED Display and 4.8mm Thin Design
HONOR unveiled the MagicPad 4, its first tablet powered by Qualcomm’s Snapdragon 8 Gen 5 mobile platform, and slated for a global launch on 2 March 2026. The 12.3‑inch OLED display delivers a 165 Hz refresh rate, 2,400 nits peak brightness, and a 93 % screen‑to‑body ratio while the...
Exaviz Cruiser Brings 8-Port PoE+, 2.5GbE, and NVMe to Raspberry Pi CM5
Exaviz has introduced the Cruiser carrier board for the Raspberry Pi Compute Module 5, delivering an 8‑port PoE+ switch, 2.5 GbE WAN, NVMe storage, dual HDMI, USB connectivity, and ESP32‑C6 wireless. The lineup includes three variants: the full‑feature Cruiser with eight...
NVIDIA’s RTX 5080 Takes the Lead at Mindfactory, AMD Remains Ahead Overall
Mindfactory’s latest sales data shows NVIDIA’s GeForce RTX 5080 has become the top‑selling card within the RTX 50 lineup, outpacing the RTX 5070 Ti and other 16 GB variants. AMD retains overall market leadership at the retailer, with the Radeon RX 9070 XT and RX 9060 XT leading the...
NCC Croatia Brings CFD and HPC to Superyacht Engineering Workflows
NCC Croatia partnered with Lürssen Design Centre Kvarner to embed computational fluid dynamics (CFD) and high‑performance computing (HPC) into superyacht ventilation design. Engineers received hands‑on training and used OpenFOAM on NCC’s supercomputer to model a real‑world duct network. The simulations...
Ciena Unveils New Pluggable Optical Engine to Meet Data Center AI Demands
Ciena introduced the Vesta 200 6.4T CPX, a pluggable co‑packaged optical (CPO) engine that delivers the industry’s highest density and up to 70% lower power consumption. The module supports 200 Gb/s per lane, retimer‑free linear‑drive operation and a 20 dB electrical loss budget, enabling...
Simon Fraser University Signs MOUs to Advance Sustainability at Cedar Supercomputing Centre
Simon Fraser University has signed memoranda of understanding with Cerio, Corix and Moment Energy to deepen the sustainability of its Cedar Supercomputing Centre, Canada’s most powerful academic HPC system. The centre already runs on clean hydroelectric power and boasts a...
LLVM Clang 22 Compiler Performance Largely Unchanged Over Clang 21 On AMD Zen 5
The LLVM Clang 22 compiler was benchmarked on an AMD EPYC 9655P Zen 5 server and showed performance largely identical to Clang 21 across a suite of open‑source C/C++ workloads. Tests also confirmed that Clang 22 remains neck‑and‑neck with GCC on the same hardware, with only modest,...
Caspia Technologies Unveils A Breakthrough in RTL Security Verification Paving the Way for Agentic Silicon Security
Caspia Technologies announced the general availability of CODAx V2026.1, an AI‑enhanced RTL security analyzer that checks over 150 insecure coding practices against more than 1,000 hardware vulnerability references. The tool can scan half‑million lines of RTL in under an hour,...
Designing the Future: AI-Driven Multi-Die Innovation in the Era of Agentic Engineering
At the 2026 Chiplet Summit, Synopsys unveiled an AI‑driven, agentic approach to multi‑die design that promises to automate and accelerate the entire workflow. The semiconductor industry is shifting toward heterogeneous chiplet architectures and 3‑D stacking, creating unprecedented design complexity. AI...
AMD Announces The EPYC 8005 "Sorano" Series
AMD announced the EPYC 8005 “Sorano” series, a Zen 5‑based, single‑socket processor line succeeding the 8004 “Siena” family. Designed for 1P servers, the chips target telco and radio‑access‑network workloads with a focus on performance‑per‑dollar and performance‑per‑watt. The series promises high core counts,...
An Agentic Formal Verifier. Innovation in Verification
Infineon’s recent paper introduces Saarthi, an autonomous AI formal verification engineer that leverages multi‑agent LLM frameworks to generate, prove, and analyze RTL properties. The system creates English‑language verification plans, refines SystemVerilog assertions, and runs Cadence’s Jasper model‑checker in an iterative...
Arm & Linaro Launch New "CoreCollective" Consortium - With Backing From AMD & Others
Arm and Linaro have unveiled the CoreCollective consortium, an open‑source‑focused alliance aimed at accelerating the Arm software ecosystem. The group is financially backed by Arm and invites any vendor to join for free. Founding members include AMD, Ampere, Canonical, Google,...
Astera Labs: Fiscal 4Q25 and FY25 Financial Results
Astera Labs reported record FY2025 revenue of $852.5 million, a 115% year‑over‑year increase, and a Q4 revenue of $270.6 million, up 17% sequentially. GAAP operating margin reached 20.3% for the year, while non‑GAAP margin was 39.2%, reflecting strong profitability. The company began...
Nimbus Data Unveils FlashMax Data Platform for Mission-Critical Enterprises and Cloud Providers
Nimbus Data introduced FlashMax, a next‑generation multiprotocol all‑flash platform that consolidates block, file and object storage into a single namespace. The system adds DirectLink PCIe expansion, hardware‑accelerated deduplication and compression, and rack‑level resiliency, supporting up to 20 PB raw (100 PB effective)...
IM Launches New-Gen of eMMC for Industrial Applications
Intelligent Memory Ltd. (IM) has launched a new low‑density eMMC product line, offering 8 GB and 16 GB capacities in a 153‑ball BGA package for industrial applications. The devices target environments requiring stability, longevity and right‑sized storage, such as factory automation, medical...
NXP TJA1410 and TJF1410 PMD Transceivers Enable “CAN-Like” Single Pair Ethernet (SPE) Connectivity
NXP has introduced two 10BASE‑T1S Physical Medium Dependent transceivers – the automotive‑grade TJA1410 and the industrial‑grade TJF1410 – that offload only the analog front‑end while the digital PHY resides in the host MCU or switch. Both devices use a three‑pin...
MWC 2026 | Ericsson to Showcase 6G Capabilities with Apple, MediaTek
Ericsson announced at MWC 2026 that it will showcase live 6G demonstrations in partnership with Apple and MediaTek. The Apple demo will feature Multi‑RAT Spectrum Sharing (MRSS) to illustrate seamless 5G‑to‑6G coexistence, while MediaTek will present a 6G centimeter‑wave data‑call prototype...
AMD Bets on Billion-Dollar Offensive in AI Infrastructure Market with Meta Mega Deal
AMD announced a multi‑year AI infrastructure partnership with Meta, delivering next‑generation Instinct GPUs, EPYC processors, and rack‑scale systems tailored for Meta’s data centers. The deal includes up to 6 GW of GPU capacity and a performance‑based warrant for up to 160 million...
OptiScaler Activates FSR 4 Under Vulkan Before AMD Itself Delivers
OptiScaler’s test build 0.9.0‑pre10 adds FSR 4 support to Vulkan games by routing the upscaler through a DirectX 12 bridge, effectively bypassing AMD’s official limitation to RDNA 4 GPUs and DX12 titles. The community‑driven implementation works in titles such as Doom: The Dark...
Dual Screen Laptops Stopped Being a Gimmick
ASUS unveiled the 2026 ROG Zephyrus Duo, the first dual‑screen gaming laptop designed for practical use rather than a novelty. It features two 16‑inch 3K OLED panels with 120 Hz refresh, an RTX 5090 GPU running at 135 W, and a five‑mode hinge...
Marvell to Showcase PCIe 8.0 SerDes Demonstration at DesignCon 2026
Marvell Technology announced it will showcase a PCIe 8.0 SerDes operating at 256 GT/s during DesignCon 2026 in Santa Clara. The demo, presented at booth #904, uses the TE Connectivity AdrenaLINE Catapult connector and the Alaska P PCIe 6.0 retimer platform. PCIe 8.0, slated for finalization...
QEC4QEA to Support Development of Quantum-Enhanced Applications Across Europe
QEC4QEA, a new Quantum Excellence Centre funded by EuroHPC JU and Horizon Europe, receives roughly €4.9 million to accelerate quantum‑enhanced application development across Europe. The centre links end‑users with quantum developers, experts and computing providers, offering guidance, tools, training and hybrid quantum‑HPC resources....
It’s Time to Kill Slow Fingerprint Scanners on All Phones over $300
Optical fingerprint scanners remain common on mid‑range and premium Android phones despite slower performance. Ultrasonic sensors cost only $6‑8 more per unit, yet manufacturers like Samsung and Google continue to use cheaper optical units in $400‑700 devices. Xiaomi demonstrates feasibility...
Reimagining Compute in the Age of Dispersed Intelligence
At the 2025 RISC‑V Summit, DeepComputing CEO Yuning Liang argued that AI will shift from cloud‑centric models to on‑device intelligence, leveraging aggressive quantization to run multiple models on a smartphone. He envisions a lightweight AI operating system that orchestrates models,...
Prusa Details XL Roadmap, Prices, And Toolheads
Prusa Research announced a $200 price reduction for the assembled Original Prusa XL, positioning it as the most affordable multi‑material tool‑changer in its lineup. The XL retains its active tool‑changing architecture, supporting up to five full heads, while the newer INDX...