{"id":145376,"date":"2025-10-16T23:11:55","date_gmt":"2025-10-16T21:11:57","guid":{"rendered":"https:\/\/prohoster.info\/blog\/novosti-interneta\/google-predstavil-coral-npu-otkrytuyu-platformu-dlya-sozdaniya-ai-uskoritelej"},"modified":"2025-10-16T23:11:55","modified_gmt":"2025-10-16T21:11:57","slug":"google-predstavil-coral-npu-otkrytuyu-platformu-dlya-sozdaniya-ai-uskoritelej","status":"publish","type":"post","link":"https:\/\/prohoster.info\/en\/blog\/news\/google-predstavil-coral-npu-otkrytuyu-platformu-dlya-sozdaniya-ai-uskoritelej","title":{"rendered":"Google has introduced Coral NPU, an open platform for creating AI accelerators","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"

Google has introduced the open Coral NPU (Neural Processing Unit) platform, providing an open hardware accelerator for machine learning models and software tools for use with standard AI engines. Coral can be used as a foundation for creating energy-efficient system-on-chip (SoC) solutions, suitable for Internet of Things systems, edge computing, and sensor data collection boards, as well as in consumer wearable devices with ultra-low power consumption, such as headphones, augmented reality glasses, and smartwatches. The project's developments are released under the Apache 2.0 license. <\/p>\n

The Coral NPU is designed for running continuously active AI applications on portable devices with minimal energy consumption. The base implementation of the Coral NPU delivers performance of 512 billion operations per second (GOPS) while consuming just a few milliwatts of power. The NPU is engineered for flexible architecture modification based on SoC manufacturers' needs. The first manufacturer to produce chips based on the Coral NPU will be Synaptics, which has announced a line of processors for Internet of Things devices, the Astra SL2610, including a Torq NPU subsystem built on the Coral NPU architecture. <\/p>\n

Typical applications of the Coral NPU include the use of AI for image and sound processing, user interaction, and context awareness. For example, devices can run large language models and applications for facial and object recognition, visual searches, speech recognition, live translation, speech transcription, keyword extraction from speech, gesture control, and voice commands, as well as user activity detection (walking, running, sleeping) and environmental type detection (indoors, outdoors). <\/p>\n

The NPU features a 32-bit RISC-V instruction set architecture RV32IMF_Zve32x, an AXI4 bus, a four-stage instruction processing pipeline with in-order dispatch, out-of-order instruction completion, four-thread scalar, and two-thread vector dispatch. The processor supports SIMD operations for simultaneous processing of 128-bit vectors and is equipped with 8 KB of ITCM memory for instructions and 32 KB of DTCM memory for data.

\"Google<\/center> <\/p>\n

The NPU consists of three processor components that work together: <\/p>\n