Ambiq’s Apollo510 microcontroller pairs the company’s subthreshold energy-efficient design technology with vector acceleration performance to bring more AI capability to edge devices.
Renesas is testing demand for RISCV-based microcontrollers, introducing the first of what could be a series of such processors, the R9A02G021, targeting low-cost applications.
The new AI accelerator in Renesas’s RZ/V2H provides a boost in AI performance and efficiency to target high-performance robots that inhabit factory floors and smart homes and buildings.
NVIDIA has announced its Blackwell architecture, although the results are mixed. It is faster, but not as much so as the industry may have been expecting.
Microchip’s PIC16F13145 microcontroller includes a new configurable logic block that raises the amount of peripheral glue logic available by more than eight times compared with its predecessors.
Renesas and TSMC demonstrated two uses of fast MRAM at ISSCC. Renesas targets faster NVM; TSMC targets working memory. Their tradeoffs are very different.
With the expansion of the RA8 series, Renesas is integrating Arm’s highest-performing Cortex-M CPU in microcontrollers optimized for graphics display and motor control applications.
In the Meteor Lake CPU die, Intel has pushed the capacitance density of on-die capacitors to new heights by taking a page out of DRAM’s book of tricks.
Arm’s new Neoverse CSS N3 and CSS V3 provide a subsystem for N3 and V3 cores. For customers that won’t differentiate in that subsystem, the CSSs provide a head start.
The DA14592 is Renesas’s smallest multicore BLE SoC, featuring a power-thrifty radio transceiver that drops power by more than 23% for send and receive compared with Renesas’s other BLE chips.
NeuReality, an Israeli startup, is building an offload engine for AI-application pre- and postprocessing that removes the need for a CPU in a server and works directly with an accelerator.
Edge-AI vendor BrainChip has updated its Akida event-based IP offering. Its new Akida 2 architecture, available only as IP, supports temporal networks and Transformer encoders at low power.
As Pat Gelsinger begins his fourth year as CEO, three projects to revitalize Intel’s fabs, CPUs, and AI accelerators could restore its technology lead, but they won’t reach the market until 2025.