Today's Hardware Pulse
Nvidia driver hints at next‑gen DLSS 5 work
Nvidia’s GeForce Game Ready driver 610.47 adds three new profile options—Enable DLSS NR override, Enable DLSS NR SL override, and Override DLSS NR presets—suggesting continued development of DLSS 5 neural‑rendering technology. The earlier DLSS 5 reveal had sparked debate over AI‑driven image alteration, and the new driver hints at a quiet but ongoing effort.
Also developing:
By the numbers: Cyient Semiconductors raises $30M Series A
AMD Prototyping AMDGPU SVM Atop DRM_GPUSVM Framework
AMD engineers have released a proof‑of‑concept Shared Virtual Memory (SVM) implementation built on the new DRM_GPUSVM framework. The prototype, though basic, successfully passes the ROCR suite and most AMDKFD and HIP validation tests. It currently lacks advanced features such as multi‑GPU coordination and vRAM‑to‑vRAM migration. AMD is using this effort to evaluate the framework’s design and solicit community feedback before any production adoption.
NVIDIA and ST Present New Delivery Boards for 800VDC Architectures
NVIDIA and STMicroelectronics unveiled two 800 VDC delivery boards aimed at high‑efficiency power conversion. The 6 kW, 850 kHz LLC converter uses a 700 V GaN primary and a 40 V MOSFET secondary, delivering 12 V at 97.5% efficiency and 2500 W/in³ power density. The 20 kW, 650 kHz...
Infineon Introduces CoolGaN-Based High-Voltage Intermediate Bus Converter Reference Designs
Infineon unveiled two high‑voltage intermediate‑bus converter (HV IBC) reference designs powered by its 650 V CoolGaN switches, aimed at AI server platforms operating on ±400 V and 800 V DC. The 800 V‑to‑50 V module delivers over 98 % efficiency, 2.5 kW/in³ density in a 60 × 60 × 11 mm package, while...
Micron’s Heavy Factory Spending Overshadows Booming Memory Sales
Micron Technology announced it will exceed $25 billion in capital spending this fiscal year, surpassing analyst expectations of $22.4 billion. The surge in AI‑driven memory demand is prompting the company to invest heavily in new fabs and advanced packaging. Micron also projected...
MacBook Neo Outperforms Air M1 in Key Benchmarks
How does the MacBook Neo compare to the MacBook Air M1? Here’s a quick benchmark comparison. The differences are interesting
Comcast & NVIDIA’s Killer AI Cocktail: Edge, SLMs, and 15ms Latency
Comcast and NVIDIA announced a joint deployment that places GPU accelerators at the network edge to run stateful small language models (SLMs) within 15 ms of the user. By processing tokens locally, the solution eliminates the round‑trip latency inherent in centralized...
New Intel Arc Drivers Promise Faster Game Loads with Shader Distribution
Intel has released a new generic graphics driver that introduces the Graphics Shader Distribution Service, promising up to a 2× reduction in first‑load times for games on Arc B‑Series and Core Ultra GPUs. The update also delivers an average 9%...
AI Software for Smart Glasses Wins £1m Prize for Technology to Help People with Dementia
CrossSense Ltd’s AI‑powered smart glasses, featuring the Wispy conversational assistant, have won the £1 million Longitude prize for dementia technology. The glasses combine a camera, microphone and speakers to deliver real‑time verbal cues and floating text that help wearers complete daily...
Sandia Labs Tests Reconfigurable Accelerators for Adaptive Supercomputing
@SandiaLabs is testing supercomputers with reconfigurable accelerators, similar to FPGAs, that optimize their hardware for specific computations that are being run. https://spectrum.ieee.org/reconfigurable-supercomputer
Telit Cinterion Pushes 5G NR Release 18 to the Edge – and Onto the Rails
Telit Cinterion has unveiled two 5G NR Release 18 sub‑6 modules, the FE990D50 and FE990D60, plus a rail‑specific FE990D60‑FR variant that adds FRMCS band support. The modules feature eight‑receive antenna capability, LTE Category 20 fallback, a 2.2 GHz quad‑core Cortex‑A55 processor and native...
Mentor Zoom Inspection Camera Enables Enhanced Industrial Environment Inspections
Waygate Technologies, a Baker Hughes business, launched Mentor Zoom, a new pan‑tilt‑zoom camera system for industrial inspections. The solution includes two HD camera heads—10× and 30× optical zoom—with a 4,000‑lumen light source and a removable touchscreen‑joystick controller that works with gloved...
TIA Advances DCE 9000 Initiative
The Telecommunications Industry Association (TIA) announced continued momentum for its Data Center Excellence (DCE 9000) initiative, the first quality‑management‑system standard tailored to modern data‑center physical infrastructure. Driven by rapid AI, cloud, and digital‑infrastructure investment, the effort targets supply‑chain resilience, operational reliability,...
SEEQC Reports 1st Quantum Computer with Integrated Qubit Control on a Chip at Millikelvin Temperatures
SEEQC announced the first full‑stack quantum computer that integrates superconducting digital control circuitry directly on a chip operating at 10 millikelvin. The five‑qubit processor, paired with a separate SFQ control chip, achieved single‑qubit gate fidelities above 99.5% and demonstrated nanowatt‑scale power...
Storage Vendors Orbit the Nvidia Sun at GTC
At Nvidia’s GTC 2026, storage leaders Hitachi Vantara, IBM, Nutanix and Seagate unveiled new integrations with Nvidia’s Blackwell GPUs and AI software stacks. Hitachi iQ now supports Blackwell and RTX PRO GPUs and the STX reference architecture, while IBM demonstrated a...
Storage Vendors Orbit the Nvidia Sun at GTC
At GTC 2026 Hitachi Vantara, IBM, Nutanix and Seagate unveiled new integrations with Nvidia’s Blackwell GPUs and AI software, tightening the bond between storage and accelerated compute. Hitachi iQ added support for Blackwell and RTX PRO GPUs and AI blueprints; IBM demonstrated an 83%...
Early Rubin GPU Access Confirms Nebius as True Partner
$NBIS priority allocation from Vera Rubin clearly signals that Nebius is being treated as a real partner That matters because in AI infrastructure, getting the newest GPUs early directly translates into attracting top customers and higher-margin workloads. This validates $NBIS strategy...
Photonic Chip Company Q.ANT Deploys Second-Gen Photonic Processors at Germany’s LRZ
Q.ANT has installed its second‑generation photonic Native Processing Units at Germany’s Leibniz Supercomputing Center, expanding on a 2025 pilot. The new NPUs deliver over 50× higher matrix‑multiplication throughput and six times lower energy use while connecting via standard PCIe alongside...
How Piezoelectric Energy Harvesting Is Solving the Battery Waste Crisis in Industrial IoT
Industrial IoT deployments rely on millions of short‑life batteries, creating a looming waste problem that could reach 1.4 million metric tons by 2030. High‑temperature piezoelectric energy harvesting converts machine vibration into electricity, tolerating up to 350 °C and eliminating the need for...
Act Fast: Kid-Friendly Volume-Limiting ANC Headphones From iClever Slashed to $34.99 (30% Off)
iClever has slashed its BTH20 kids‑friendly Bluetooth headphones to $34.99, a 30% discount. The model combines an 80 dBA volume limiter, active noise‑cancelling that cuts up to ~40 dB, and a rugged BPA‑free build designed for ages 3 and up. Battery life...
New at Mouser: Digi Connect Sensor XRT-M for IoT, Industrial, Sensor, and Gateway Applications
Mouser Electronics has begun shipping Digi International’s new Digi Connect Sensor XRT‑M, a rugged LTE‑M cellular gateway designed for zero‑infrastructure IoT monitoring. The battery‑or‑solar powered device offers edge processing, cloud connectivity via Digi Axess, and IP68 protection for harsh environments. It...
Ohio Citizens Tell Hyperscalers to Take Their Supersized Datacenters Elsewhere
Ohio residents have filed a petition to amend the state constitution, banning datacenters larger than 25 MW. The initiative, led by citizens in Adams, Brown and Clermont counties, collected roughly 1,800 signatures, surpassing the 1,000‑signature threshold to trigger a ballot measure....
Accelerating Computational Lithography Using Massively Parallel GPU Rasterizer
Siemens EDA unveiled a GPU‑accelerated rasterization algorithm that transforms computational lithography workflows. By decomposing layouts into tiles and processing them on NVIDIA H100 GPUs, the method attains speedups of up to 290× for Manhattan geometries and 45× for curvilinear designs....
AMD Keeping Samsung Close Amid Memory Shortage
AMD has entered a strategic agreement with Samsung to secure next‑generation HBM4 memory amid a broader industry shortage. The partnership aims to lock in supply for AMD's upcoming AI‑focused GPUs, addressing concerns raised after the company warned of potential shortages...
$42 3D Printer Trades Features for Price in “Race to the Bottom”
A $42 Frequency Division Multiplexing 3D printer has hit the market, slashing desktop printer prices to unprecedented lows. The machine offers a 100 × 100 × 100 mm build volume, 230 °C hot end limited to PLA, and print speeds of 10‑40 mm/s. It lacks automated leveling,...
Samsung Pulls Trifold Phone From Market After only Three Months
Samsung announced a premium trifold smartphone in early 2024, pricing it around $1,500 and touting a tablet‑sized display that folds into a pocket‑friendly form factor. Within three months, the device failed to meet sales expectations, prompting the company to pull...
LRZ Decommissions CoolMUC-3 Supercomputer After Almost Decade of Use
Germany’s Leibniz Supercomputing Centre (LRZ) has retired its CoolMUC‑3 supercomputer after almost a decade of service. Deployed in 2017, the nine‑rack cluster housed Intel Xeon Phi processors and delivered roughly 400 teraflops for fluid‑dynamic research at Bavarian universities. It pioneered Megaware’s ColdCon...
Memory Chip Cycle Revives; Micron Still Cheap
Micron reports earnings tonight. The stock is up 350% in a year. And somehow, at 10.7x forward earnings, it might still be cheap. Here's why the memory chip industry is experiencing something that hasn't happened in four decades — and...
Data Centre Capacity in India Crosses 1,500 MW: Govt
India’s data centre capacity surged to over 1,500 MW in 2025, up from 375 MW in 2020, reflecting rapid digital adoption. The bulk of this capacity is clustered in Mumbai and Navi Mumbai (790 MW), with Chennai, Bengaluru, Hyderabad and Delhi‑NCR providing the remainder....
Atom Computing Integrates NVQLink Into Quantum Systems
Atom Computing announced that it has successfully integrated NVIDIA’s NVQLink, a low‑latency, high‑bandwidth communication interface, into its proprietary quantum control‑systems stack. The integration delivers ultra‑low latency pathways that accelerate logical‑qubit cycle speeds and support scaling to thousands of qubits. Validation...
StorMagic and HiveRadar Partner to Deliver Mobile Edge Infrastructure with Built-In Resilience
StorMagic and HiveRadar have teamed up to launch a combined edge‑computing offering that pairs StorMagic’s SvHCI virtualization software with HiveRadar’s Portable Edge Data Center hardware. The solution delivers enterprise‑grade virtual machines, encrypted storage and real‑time telemetry in a rugged, off‑grid...
Tait Announces Shared-Channel Version of DMR Tier 3
Tait Communications has unveiled the OpenTrunk shared‑control‑channel solution, a DMR Tier 3‑based trunking system that operates on shared spectrum instead of a dedicated control channel. The offering addresses the scarcity and high cost of licensed spectrum in major U.S. metros such...
Sony Releases New PlayStation 3 Update Nearly 20 Years After Launch
Sony has issued a new system software update (version 4.93) for the PlayStation 3, nearly two decades after the console’s launch. The update’s notes repeat the generic claim of "improving system performance," but its real purpose is to refresh the Blu‑ray...
Best DLSS 5 Review yet, yet Presentation Falters
This is the best take on DLSS 5 I’ve seen yet. There IS cool tech here but the presentation was Xbox One launch levels of out of touch…
China’s New Horse-Inspired Robot Can Haul Gear, Climb Stairs
Hangzhou‑based DEEP Robotics unveiled a limited‑edition robot horse to mark the 2026 Year of the Horse. Weighing 66 lb and standing half a meter tall, the bionic quadruped can haul up to 110 lb, climb 25 cm stairs and tackle 45‑degree slopes while...
The Retroid Pocket G2 Is the Latest Victim of the RAM Crisis, with the Android Gaming Brand Issuing a Worrying...
The ongoing global RAM shortage has forced Retroid to temporarily discontinue its Pocket G2 handheld, citing volatile memory pricing. The company also raised the price of its Pocket Classic from $129 to $149, reflecting higher component costs. Retroid recommends the...
Teradyne Launches Photon 100 Opto-Electric Automated Test Platform
Teradyne announced the Photon 100, an opto‑electric automated test platform designed for high‑volume silicon photonics (SiPh) and co‑packaged optics (CPO) manufacturing. The system merges optical and electrical instrumentation on the UltraFLEXplus chassis, covering wafer, optical‑engine and module‑insertion testing. By delivering scalable,...
NVIDIA SoC Detects Faces Under 1 Ms, Ultra‑Low Power
Curious about how NVIDIA's latest SoC achieves face detection in less than a millisecond while conserving power? https://spectrum.ieee.org/face-recognition-nvidia-chip-soc?share_id=9270633
AXTI Faces Supply Risk as Competitor Secures Long-Term Deal
The comments sure seemed negative on $AXTI This will most definitely become an issue for $AXTI - One major customer saying denied the existence of an InP bottleneck while another secured a 7-year supply agreement with AXTI's main competitor, Sumitomo https://t.co/ooUWdVkK2o
Nvidia’s Always-On Chip Detects Faces in Less Than a Millisecond
Nvidia researchers unveiled an always‑on vision system that can detect human faces in under one millisecond while consuming less than 5 mW of power. The chip, called Alpha‑Vision, operates at 60 fps and is active only 5 % of each frame cycle, achieving...
Agentic AI Demands Diverse Compute Engines: CPUs, GPUs
Agentic AI will require all kinds of compute engines, including CPUs, GPUs, DPUs, and now LPUs. $NVDA $GOOG $AMD $INTC https://t.co/WVnnc36X2U
Dell UltraSharp 52 Makes All Other Screens Feel Lacking
The only downside of the @DellTech UltraSharp 52 is that when I move away from it I hate every screen I use
Flash Isn?t Just Expensive?It?s a Supply-Chain Risk
Flash memory prices are climbing even though manufacturers report no major supply disruptions, driven by soaring AI demand and a production base heavily concentrated in a few regions. This concentration turns modest demand shifts into sharp price spikes and longer...
Fortinet Bets on AI, Silicon, and Edge Data Centers
.@Fortinet’s sees advantage in AI, chip, data center investments https://t.co/l4n1P1c5O0 Fortinet is betting that long-term investments in its own AI data centers and silicon, being early to quantum-safe offerings and AI agents moving to the edge will drive its results...
GTC Hype Mirrors
GTC this year giving 2017 ICO mania vibes - people adding “Nvidia powered / partner” when they are just customers compute economy local top in 12 - 18 months?
AMD Prepares Ryzen 7 9750X and 9650X Zen 5 Refresh CPUs
AMD is set to launch two Zen 5 refresh CPUs – the Ryzen 7 9750X and Ryzen 5 9650X – that boost base clocks and raise default power limits without changing the underlying architecture. The 9750X moves to an 8‑core/16‑thread design with a 4.20 GHz base...
New Tools Boost Multi‑Die Assembly Yield and Speed
How new equipment and methodologies are improving reliability, yield, and time-to-market for multi-die assemblies. https://t.co/aO5qUKLpbY #semiconductor #advancedpackaging #metrology @cohu_inc @oontoinnovation @yieldWerx #ASE #DataAnalytics #metrology https://t.co/draBrAqta5
Chiplet Architectures Could Overcome Miniaturization Limits
Are Chiplet Architectures a Real Answer to Chip Miniaturization Limits? by @antgrasso #EmergingTech #Innovation #Tech #Technology https://t.co/AZQUeZYhHw
16TB M.2 NVMe SSD Listed at Nearly $16,000 Online
Exascend has listed a 16 TB PCIe 4.0 NVMe SSD in the standard M.2 2280 form factor for $15,935. The drive prioritises storage density over raw speed, delivering up to 3,270 MB/s reads and 2,980 MB/s writes while consuming under 1.3 W idle and 7.2 W active....
AMD Condemns Chuwi Ryzen Laptop Processor Mislabeling, Reserves Legal Action
AMD issued a formal statement condemning Chinese OEM Chuwi for falsely branding laptops with a Ryzen 5 5500U as the newer Ryzen 5 7430U. The mislabeling involved BIOS string changes and external 7000‑series branding, violating AMD’s OEM agreements. AMD emphasized that the practice was...
Anschütz Clears CDR for Hunter-Class Frigate Integrated Navigation Systems
Anschütz has completed the Critical Design Review for its Warship Integrated Navigation and Bridge System (WINBS) destined for the Royal Australian Navy’s Hunter‑class frigates. The CDR confirms the design meets technical and operational requirements and clears the way for production...