Latest from Embedded

Dreamstime_mauriceyom98_321540663
Enhancing FPGA programmability
191448470 © Biancoblue | Dreamstime.com | Infineon
id_191448470__biancoblue__dreamstime
ID 217230663 © Christian Offenberg - Dreamstime.com | electronica.de
promo_messe_munich__id_217230663__christian_offenb
ID 312689142 © Chechotkin | Dreamstime.com
LiDAR in autonomous driving
ID 333722484 © Weddingodessa85 | Dreamstime.com | AI generated.jpg
Translucent car showing electronics | AI generated
Dreamstime_cookiecutter_100807069
Security risks with DRAM
ID 326211957 © Kwangmoo | Dreamstime.com
Security IP CCTV camera using solar energy
Dreamstime_sompongsriphet_152034055
65bbbaff7a5f68001ef23f04 Auto Dreamstime Sompongsriphet 152034055

As AI Takes Off, Chipmakers Pump Up Performance (Download)

Feb. 1, 2024

Read this article online.

It’s hard to miss the fact that AI is trending strongly across the industry. Demand for it is on the rise everywhere, including the edge, leading to upgrades of existing semiconductor products. Suppliers are targeting new data types and new instruction sets as well as introducing new products. 

Each of these initiatives aims to squeeze more computing, optimized for the needs of AI, from small devices that generally must use the bare minimum of power served up at a rock bottom price point. A big challenge is that inferencing used in more advanced AI depends greatly on large language models (LLMs) that require enormous computing resources for training. It also, for the most part, forces edge devices to rely on high-bandwidth communication with more powerful resources in the cloud or a data center, which can be inconvenient and power-intensive.

Comments

To join the conversation, and become an exclusive member of Electronic Design, create an account today!