Hardware & Chips — Posts

58 posts on "Hardware & Chips" by Shlomo Strauss — bilingual writing on technology, science, and business.

• The academic roots behind China's new AI models — Despite hardware bans and Western skepticism, MoonShot and Z AI achieve impressive technological feats through advanced math and academic research.
• The Shared Logic of 19th-Century Elevators and Modern AI — While modern AI servers rely on microscopic transistors instead of moving parts, they operate on the exact same logic gates used in 1880s elevators.
• Google hardwires Transformer architecture into new Gemini chips — Google is tackling massive AI compute costs by developing custom ASIC chips with the Transformer architecture hardwired directly into the silicon.
• The impending collapse of GoPro and its technological legacy — Facing financial struggles and fierce competition, GoPro nears the end of its path while leaving a lasting legacy in digital image stabilization.
• NASA and Microchip to build space processors 100 times faster — NASA and Microchip are developing a radiation-resistant space chip 100 times faster than current models, with future uses for critical systems on Earth.
• How Cerebras Built a Giant AI Chip and Why It Is Going Public — While Cerebras boasts revolutionary wafer-scale AI chips, its upcoming multi-billion dollar IPO faces risks from extreme customer concentration.
• How USB-C Laptop Charging Became Safe Enough for the EU Mandate — A new EU law mandates USB-C charging for laptops, made possible by hardware solutions that prevent the short circuits that once burned motherboards.
• Nvidia's $9 Million Server Threatens Traditional Hardware Makers — Nvidia's upcoming NVL144 server highlights the company's push to control the entire AI supply chain, threatening traditional hardware manufacturers.
• How Tesla solved the battery and heat problems that held back — From cylindrical cells with interleaved cooling tubes to a structural battery floor, a look at the engineering breakthroughs behind Tesla's thermal and…
• The engineering inside Tesla's motors that most drivers never think — Dual motors, carbon-fiber sleeves, and silicon carbide inverters — a clear look at the engineering solutions Tesla built into its electric drivetrains.
• Why old HDD drives are suddenly back in high demand — EU AI data-retention rules, rising SSD costs, and new laser and microwave writing technologies are driving an unexpected HDD revival with capacities up to
• A QR code smaller than a bacterium just broke a world record — Researchers from TU Wien and startup Cerabyte etched a 1.98 sq-micron QR code onto ceramic, demonstrating a storage method that could hold 2TB per A4…
• Why Nokia got Asus and Acer banned from selling PCs in Germany — A Nokia patent lawsuit over HEVC video decoding reveals the licensing chaos behind modern video codecs and why the industry is moving to AV1.
• Intel's ZAM memory aims to pack 512 GB where HBM tops out at 64 — Intel and SoftBank's ZAM technology uses diagonal drilling between memory layers to promise 512 GB per package — eight times the capacity of today's best…
• Tower Semiconductor and Nvidia bet on silicon photonics for AI — A new deal between Tower Semiconductor and Nvidia will bring silicon-nitride optical interconnects inside AI servers, promising up to 70% energy savings…
• How a Japanese toilet maker became a key supplier to the chip — Toto's mastery of ceramic shrinkage, developed for luxury toilets, led to an electrostatic wafer-holding technology now in high demand as AI chip…
• Why Nvidia's $20 billion Groq acquisition is a bet on the inference — As the AI industry shifts from model training to cost-efficient inference, Nvidia's Groq acquisition signals a strategic pivot to compete with Google's…
• The glass fabric holding the chip industry together — and Intel's — Why boron-free T-Glass became a critical bottleneck for AI chip production, and how Intel is developing a pure glass substrate as a future alternative.
• Why Nvidia skipped new GPU launches and what DLSS 4.5 signals instead — VRAM shortages driven by AI server demand blocked new GPU launches, so Nvidia responded with a Transformer-based DLSS 4.5 upgrade that generates graphics…
• A single server ran for four months to break the pi calculation — StorageReview set a new pi digit record using a single Dell server equipped with 384 CPU cores, 1.5 TB of RAM, and 2.5 PB of Micron SSDs serving as swap…
• The technology powering the Sphere's massive LED exterior — A look at how 1,500 Nvidia A6000 GPUs, 1.2 million LED tiles, and custom sync software bring the Sphere's exterior screen to life.
• Breaking a CPU overclocking record with liquid helium — From quartz oscillators to liquid-helium cooling at near absolute zero, this post explains the physics behind pushing a CPU past 9 GHz to set a world…
• How AI's appetite for HBM memory is driving up the cost of ordinary — As manufacturers shift production lines to the more profitable HBM chips powering AI accelerators, a RAM shortage for standard computers is expected to…
• Apple's N1 chip beats Broadcom with half the bandwidth — Apple's first in-house WiFi chip, the N1, uses a narrower 4nm design to outperform leading Broadcom chips in real-world use while cutting reliance on…
• Germanium: the element inside your phone, missiles, and quantum — A look at germanium's unique optical and electronic properties, its role in 5G, thermal imaging, and missiles, and the chalcogenide glass competing to…
• The $400 million machine at the heart of modern chip manufacturing — A look at how ASML's extreme ultraviolet lithography machine works, and the startup betting X-rays can do the job better and cheaper.
• Lab-grown diamonds are becoming a critical component in chips and — Lab-grown diamonds, which conduct heat five times better than copper, may soon enable faster chips and room-temperature quantum computing qubits.
• The video switcher that assembles a live TV broadcast in real time — A look at the video switcher in a broadcast control room — the device that merges cameras, remote feeds, and graphics into the final picture viewers see…
• Apple I: the garage-built computer now selling for nearly a million — A look at the Apple I's origins, design philosophy, and how a machine that sold for $666 became a collector's item fetching nearly $1 million at auction.
• ASUS's gold-plated GPU pushes luxury PC hardware to a new extreme — ASUS's ROG Astral Dhahab line, sold exclusively in the Middle East, is set to get a 5 kg solid 24K gold variant worth over half a million dollars, paired…
• From a 2 MB disk in 1966 to the quantum limits of storage — A 2.9 kg disk built for UNIVAC computers in 1966 puts six decades of storage progress in perspective — and raises the question of where quantum physics…
• Google's patent puts 3D face recognition directly under the display — A newly filed Google patent describes an under-display 3D face recognition system that could rival Apple's Face ID while eliminating the need for a notch…
• Active optical HDMI cables: when the premium price makes sense — Unlike standard copper HDMI cables, active optical variants convert signals to light internally, solving interference and distance problems in demanding…
• BOE makes screens for Apple and the world — and almost no one knows — A look at BOE, the Chinese display giant behind TV panels, smartphone screens, and a new folding dashboard prototype — operating largely out of the…
• BYD's YangWang U9 hits 392 km/h and challenges the EV speed record — The Chinese-built YangWang U9 supercar packs 1,306 hp across four motors and tops out near 392 km/h, placing it just behind the Rimac Nevera as the…
• Camera sensor size explained: why it matters and how smartphones — A concise explainer on how sensor size affects image quality, low-light performance, and bokeh — and why software processing lets smartphones punch above…
• A full charge in 4 minutes: the real cost of ultra-fast charging — Realme set a world record with 320W charging, but the tradeoff is accelerated battery wear — which explains why Apple and Samsung deliberately cap their…
• CPU sockets explained: how a processor actually connects to a — A concise look at LGA and PGA socket types, how pin count affects data throughput, and the differences between Intel and AMD connection standards.
• PCIe 5 SSDs offer double the speed — but at a steep price — PCIe 5 SSDs reach 14 GB/s transfer speeds, but high prices and limited compatible motherboards make waiting the smarter move for now.
• That $12 10TB drive from AliExpress is a scam — here is what is — A Reddit user who bought a supposed 10TB external drive for $12 found a cheap, no-name microSD card inside running firmware that fakes the reported…
• The Sony camera that saved photos directly to floppy disks — The Sony Mavica MVC-FD5 was a real digital camera that stored photos on standard floppy disks.
• The quiet engineering inside a professional XLR cable — An XLR cable uses three internal wires and phase inversion to cancel electromagnetic interference — a clever design that keeps professional audio clean.
• Why two 14-inch screens are not always equal in size — Aspect ratio determines actual screen area — a 14-inch display at 4:3 offers roughly 12.6% more workspace than a 16:9 screen of the same diagonal size.
• RGB vs. aRGB: the real difference behind PC lighting — A clear breakdown of how standard RGB and addressable aRGB lighting work, and why mixing components from different manufacturers can make PC builds…
• DDR4 vs DDR5: why RAM type matters when buying a new PC — A practical breakdown of DDR4 versus DDR5 RAM — covering clock speeds, CL latency, and what the differences actually mean for everyday performance.
• DDR4 vs DDR5: why your next PC's RAM type actually matters — A clear breakdown of how DDR4 and DDR5 differ in clock speed, latency, and real-world performance, and why it should influence your next PC purchase.
• ThinkPad sub-series explained: how to tell T, X, P and E apart — A practical breakdown of Lenovo's ThinkPad sub-series — T, X, P, L, E and Yoga — so you know exactly what you're paying for before you buy.
• How to choose a PC power supply: the key specs explained — A practical breakdown of PC power supply specs — wattage, 80+ efficiency ratings, modularity, and active PFC — with an emphasis on why brand reliability…
• A cheap power supply can silently kill a new PC — A real-world diagnosis of a PC shutting down unexpectedly shows how a low-quality power supply can degrade silently within months and why it pays to…
• Why Apple stacked two OLED panels in the new iPad — Apple's Tandem OLED bonds two panels together to reach 1,000 nits full-screen brightness, three to five times that of a typical computer display.
• Your PC is furniture — it's time to treat it like one — A custom-built PC can reflect your personality and complement your workspace just as much as any other design element in the room.
• MSI Project Zero wants to hide every cable behind the motherboard — MSI's Project Zero is a hardware ecosystem designed to route all cables behind the motherboard, eliminating visible wiring from the glass-panel side of a…
• Zigbee explained: the protocol powering your smart home — A clear breakdown of how Zigbee works, why it differs from Wi-Fi, and what role a smart home hub plays in bridging the two protocols.
• What is inside Apple's $129 Thunderbolt 4 Pro cable — An internal scan of Apple's Thunderbolt 4 Pro cable reveals a complex printed circuit supporting data, video, and up to 100W charging, alongside Intel…
• LCD vs OLED: how your screen actually works — A concise breakdown of how LCD and OLED screens work, and why the difference matters the next time you buy a device.
• The architecture of Apple's UltraFusion processor bridge — Bypassing traditional upgrades, Apple uses an UltraFusion bridge to merge two processors into a single chip transferring 2.5 terabytes per second.
• What makes Apple Silicon so fast and energy-efficient — A technical breakdown of Apple Silicon's edge, covering its SoC design, unified memory, ARM architecture, Metal API, and Apple's end-to-end control over…
• CPU vs GPU: why core count matters more than speed — A clear breakdown of how CPUs and GPUs differ in architecture, and why massively parallel but simple cores give GPUs the edge in gaming, crypto, and AI.