Huawei launched the Mate 60 Pro phone with a 7 nm processor fabricated in China, seemingly in contravention of Western sanctions. They did it by employing multiple-patterning lithography.
The AMD Epyc 8004 (Siena) processor targets telco and other embedded designs. Employing the Zen 4C CPU, it’s similar to the recent Bergamo server processor but has fewer CPUs and DRAM interfaces and a smaller package.
Peak raw performance of the Hexagon-based AI accelerator in Qualcomm’s newest flagship smartphone processor is 4× greater than that of the previous generation.
Industry leaders duke it out on the new LLM test included in the new version of the MLPerf data-center inference benchmark while also pursuing efficiency gains.
Intel, Ayar Labs, and Lightelligence ran experiments demonstrating the benefit of fiber in an optical mesh network and its use in accessing remote memory via CXL.
Less than two years after acquiring Pensando, AMD is readying the new Giglio DPU that tweaks the earlier Elba design to raise performance and reduce power.
Nuclei Systems, a Chinese RISC-V processor vendor, announced a licensable RISC-V CPU core certified for ASIL D automotive applications, the most stringent level.
Arm’s new licensable Neoverse CSS N2 provides an entire compute subsystem implemented and verified on TSMC’s 5 nm process. It may save customers months of work.
Israeli startup Inuitive has released the NU4100 chip for 3D image processing. Incorporating the company's image-processing accelerators and technology from Arm and Synopsys, the chip targets autonomous machines, AR/VR, and the smarthome market.
Intel plans to introduce two server processors in 2024, Granite Rapids and Sierra Forest, using chiplets to boost the number of CPUs and deliver new DRAM and I/O capabilities.
Silicon Labs’ EFR32G23 and -28 support Amazon’s Sidewalk protocol. One transmits using subGHz FSK; the other FSK and Bluetooth. The company provides the full software stack.
Ceva’s updated NeuPro-M artificial-intelligence accelerator can better handle floating-point and INT4 data, enabling developers to use different types at different neural-network layers to improve performance with less impact on accuracy.
Samsung and TSMC are forging ahead with 3 nm and 2 nm silicon processing as well as 3D-packaging technologies. Globally dispersed fab construction continues for old and new nodes.