September 30, 2020
The modules are based on the Hailo-8 artificial intelligence chip and appear in Mini PCIe cards that can be installed immediately with end fittings. The company estimates that they are 26 times faster than Intel’s Myriad-X module, and 13 times faster than Google’s Edge TPU module.
The Tel Aviv company Hailo has announced the Hailo-8 M.2 acceleration cards that appear in Mini PCIe format and are intended to accelerate the heating activity of deep learning networks in end devices. The modules are based on the Hailo-8 artificial intelligence processor developed by the company. The modules are adapted to the standard software infrastructures in the field of neural networks, such as TensorFlow and ONNX, which are supported by Hailo’s Dataflow Compiler.
The modules provide a processing power of 26 trillion operations per second (TOPS) at a power efficiency of 3TOPS / W. The module appears in sizes of 22 by 60 mm and 22 by 80 mm and is supported by the Linux operating system. Soon the company will also add support for the Windows operating system. The company reported that in a comparative test between its module and Intel’s Myriad-X module and Google’s Edge TPU module, it was faster at a high rate: the number of images from a standard test that it processed in one second was 26 times higher than Intel’s and 13 times higher. Google.
The card is intended for a new and evolving market of artificial intelligence in a plug-in format, which is designed to meet the demand for high-performance and energy-efficient end fittings. This is reflected in the growing demand for fanless AI-based end accessories, which allow multiple cameras to be connected to a single end device that provides smart video processing.
Posted in categories: Artificial Intelligence, News, Semiconductors
Posted in tags: artificial intelligence, halo, chips
*The article has been translated based on the content of Source link by https://techtime.co.il/2020/09/30/hailo-8/
. If there is any problem regarding the content, copyright, please leave a report below the article. We will try to process as quickly as possible to protect the rights of the author. Thank you very much!
*We just want readers to access information more quickly and easily with other multilingual content, instead of information only available in a certain language.
*We always respect the copyright of the content of the author and always include the original link of the source article.If the author disagrees, just leave the report below the article, the article will be edited or deleted at the request of the author. Thanks very much! Best regards!