CES 2023 AI Roundup: AMD AI Chips, Nvidia Omniverse Updates

CES, one of the largest tech events in the world, has come and gone for another year.

AI Business brings you some of the AI announcements you may have missed from the show floor in Las Vegas.

AMD’s Alveo V70 AI Accelerator Preview

Chipmaker AMD gave the world a glimpse of Alveo V70, a new high-performance, energy-efficient AI inference accelerator.

Designed for data center and cloud deployments, the hardware is based on AMD’s XDNA with AI Engine architecture.

According to AMD, the Alveo V70 “extends pervasive AI from edge to cloud and delivers industry-leading compute efficiency for the most demanding workloads.”

The chipmakers also provided a look at its new integrated data center CPU and GPU, the AMD Instinct MI300. Built for HPC and AI performance, the MI3000 boasts Zen 4 CPU cores and HBM memory chips atop AMD’s CDNA3 GPU architecture.

Ryzen AI Technology

AMD also unveiled Ryzen AI Technology – the first dedicated artificial intelligence hardware in an X86 processor, part of the new Ryzen 7040 Series Mobile processors.

The new hardware is focused on bringing adaptive AI architecture to laptop computing – providing more performance for real time AI experiences.

Nvidia Omniverse gets new tools

In a special address at CES, Nvidia announced new features for its metaverse application development platform, Omniverse, including enhancements for the 3D modeling tool Blender and a new suite of experimental generative AI tools.

Related:How Artificial Intelligence Will Evolve in 2023

The changes give Blender users access to a new panel allowing them to transfer shape keys and rigged characters. The challenge of reattaching parts of rigged characters can now be solved with a one-button operation from Omniverse Audio2Face — an AI-enabled tool that generates realistic facial expressions from an audio file.

An expansion to Nvidia Canvas will allow all RTX users to access 360 surround images to create and conceptualize panoramic environments.

And the AI ToyBox, which features extensions derived from NVIDIA Research, enables creators to generate 3D meshes from 2D inputs.

Nvidia also announced that early access for the Unity Omniverse Connector, which provides Unity users the ability to send data to an Omniverse Nucleus server is now available.

Read the rest of this article on AI Business.