आज क्या बदल गया, यह आपके BOM के लिए क्यों महत्वपूर्ण है और क्या देखना चाहिए। हर तथ्य-दावा को स्रोत तक नक़ल किया जाता है; हमारी स्वयं की विश्लेषण चिह्नित की जाती है और अलग रखी जाती है。
EN · एकल भाषा संपादकीय नीति
SIA and SEMI August 2026 data show global chip sales up 144% year-on-year; Omdia lifts 2026 growth to 94.1% while HBM, advanced packaging and 2nm remain bottlenacked into 2027.
Frore Systems adds an integrated diamond spreader to its LiquidJet coldplate, claiming 35% tokens/Watt efficiency versus skived coldplates for hyperscaler AI factories.
HPE launches its first four Gen13 servers on 6th Gen AMD EPYC, targeting agentic AI and inference. XAIPART maps the parts and sourcing impact.
OMNIVISION's OG05D 5MP global shutter HDR sensor shrinks footprint 30 percent and cuts power 30 percent versus prior 5MP generation, aimed at barcode, factory automation and humanoid robotics cameras.
TI has extended Zephyr support across MCUs, MPUs and wireless devices, with upstream code, SDK updates and CRA-oriented security maintenance relevant to design teams.
Microsoft priced the Surface Laptop Ultra at $2,599-$5,899 with October 16 shipping, debuting NVIDIA's RTX Spark N1X in a 15-inch mini-LED chassis and a breakaway USB-C port.
Intel is preparing a Core 2000HX mobile family on 'Raptor Lake Next' dies, reviving DDR4 compatibility for laptops while Nova Lake-HX takes the Core Ultra 4000HX name.
MediaTek and Airoha are showing Filogic 8800, an enterprise-grade tri-band Wi-Fi 8 router platform, claiming near-100% throughput gain and 50% lower system power across mixed Wi-Fi 5/6/7/8 fleets.
Renesas adds four 100V E-mode GaN FETs (RDS(on) 1.2 to 5 milliohm) claiming up to 63 percent lower soft-switching FOM than comparable GaN devices, targeting 800 VDC / 48V power stages for AI servers, humanoid robotics and motor drives.
Rohm is outsourcing TOLL packaging to Tata Electronics and TO-252 packaging to Suchi Semicon, starting with legacy power MOSFETs while keeping wafer fab in-house.
SK hynix has not started HBM hybrid bonding samples while Samsung has shipped them, raising design-in and supply questions for advanced AI memory.
Samsung's HBM4E reportedly passed NVIDIA Rubin Ultra testing at up to 16 Gbps/pin and 4 TB/s per stack, signaling a rebound after the HBM3E rejection.
प्रति प्रकाशन दिन एक ईमेल। आप इसमें अनुकूल स्रोत के रूप में XAIPART जोड़ सकते हैं।
RSS फीड · हम इन संक्षिप्त सूचनाओं को कैसे स्रोतों से अद्यतन करते हैं
