Table of Contents
A Intel opened, on the morning of this Tuesday (19) (13 pm Brasília time), the Intel Innovation 2023, where he was able to tell a little more about the architecture of his next generation, Meteor Lake. The company's focus during the presentation was the power of Artificial Intelligence, seen by the manufacturer as a pillar of the future of computing. Among the highlights revealed by Intel are the concept of an AI PC equipped with the intel core ultra. She also announced the arrival of OpenVINO 2023.1, a “toolkit” that will facilitate and optimize the use of AI. Check out all the event details below.
Watch the video with the summary on the Showmetech channel:
PC with artificial intelligence
One of the main announcements of the first day of Intel Innovation 2023 it was the AIPC, a concept of a computer with artificial intelligence. Based on the new Intel Core Ultra processor, the chip brings a NPU (neural processing unit, in free translation) and Intel 4 technology to deliver greater energy efficiency in artificial intelligence processing. In addition, there is an integrated GPU from the Intel Arc line, but, unlike expected, the focus here is not on graphics processing, but on performance and efficiency capabilities.
Despite not giving further details of what this device will actually be like, Intel demonstrated some AI applications in practice. In one of them, for example, the machine managed to generate excerpts from a song in the style of Taylor Swift, with entirely local processing. At another time, the PC also generated an image based on a description, similar to what we can already do with other software available on the market.
The PC with AI also promises to help users' productivity. In another demonstration, the computer was able to warn Pat Gelsinger, CEO of Intel, who was in front of the computer, that someone was calling for him. He then left the computer and when he returned, the device had summarized everything he had missed while he was away, including a translation of all content from French.
During the event, Pat also argued that the technology developed by the company will be fundamental to the expansion of artificial intelligence. He even highlighted the advances his company is making in production technology and software development tools for AI. The opportunity will only increase as more artificial intelligence capabilities are powered by personal computers, he said.
AI represents a generational shift, ushering in a new era of global expansion in which computing is even more fundamental to a better future for all […] For developers, this creates enormous social and business opportunities to push the boundaries of that is possible and create solutions to the world's biggest challenges.Pat Gelsinger – CEO of Intel
Who also participated during Intel Innovation 2023 was the director of operations at Acer, Jerry Kao, who took an Acer Swift with the Intel Core Ultra to California. According to him, the two companies are working to develop a set of software that takes advantage of the potential of Core Ultra chips. On stage, both also announced a feature capable of accessing any information that the user may have seen on the PC screen.
The company guarantees that there is no reason to worry about security and privacy with the new feature. According to it, all images and texts are encrypted and stored locally, without any internet access.
Another novelty announced by Intel that deals with Artificial Intelligence is the arrival of OpenVINO 2023.1, an open source software toolkit used to facilitate the optimization and application of AI processing at the Edge and in the Client segment — the name given to the area that encompasses end-consumer desktops and notebooks. According to Intel, the version brings improvements that will make access and use of algorithms more simplified on a greater number of devices.
Now, programmers and researchers will have the ability to automatically import and convert models originating in PyTorch and TensorFlow. Additionally, you will be able to choose the most appropriate API based on your specific workload requirements.
Version 2023.1 of OpenVINO becomes even more favorable to generating artificial intelligence, expanding the variety of models available for chatbots, code generation and other applications. At the same time, it improves support for GPUs and memory usage efficiency. More advanced compression techniques were also incorporated to optimize the use of Large Language Models (LLMs).
Lastly, it is important to highlight that these tools now enjoy full support by the Neural Processing Unit (NPU) of the new Intel Meteor Lake processors, which is new to Innovation 2023. When combined with OpenVINO's inherent ability to automatically distribute the Artificial Intelligence processing among available computing resources, these changes represent a crucial milestone in Intel's mission to make AI accessible to everyone by making it available “anywhere, on any device.”
Edge-Native Software Platform
Recently, large companies have realized the benefits of bringing data centers consumers and sensitive areas, such as distribution centers and factories. This is called Edge Computing. Sectors such as telecommunications, logistics, monitoring, product inspection and interactive media are users of this technology.
According to Intel, by 2025, more than half of the data generated by companies will be outside the data centers or the cloud, and by 2026, half of Edge Computing platforms will be integrated with Machine Learning solutions, a tenfold increase from 2022. In response to this trend, the blue team is developing a complete AI platform called Temporarily of “Edge-Native Software Platform”, which includes software and hardware resources.
The objective of this solution is to boost the development and adoption of Edge Computing, with several modular layers adapted for different business sectors, following the flexibility offered in the brand's other services. Details about how the platform works and the launch date have not yet been released, but we can expect news in the coming months.
As expected, Intel also revealed the new processor families, which will be launched in the coming months. The highlight given by the company was the 4th generation Xeon Scalable Sapphire Rapids, which according to Intel data would be “the best family of CPUs for Artificial Intelligence on the market".
Some of the AI-focused improvements implemented in the 5th generation Emerald Rapids were also detailed, as well as in the Granite Rapids (composed of high-performance P-Cores) and Sierra Forest (based on high-efficiency E-Cores) lines, the latter two scheduled to be launched only next year.
The Xeon Sapphire Rapids processors, launched in January, stood out for being Intel's first data center chips to use a chiplet architecture and integrate several accelerators, including AI-focused solutions. This strategy boosted performance per watt, reduced total cost of ownership (TCO) and, according to the company, secured leadership in Artificial Intelligence processing on CPUs.
With the 5th generation Emerald Rapids, the company seeks to maintain its prominent position in the Xeon line, promising additional optimizations and a leap in energy and operational efficiency. However, the main changes are in the Granite Rapids and Sierra Forest families, which expand the use of chiplets to offer greater flexibility to servers.
The Granite Rapids family is aimed at high-performance processing in general and intensive tasks, including AI, while Sierra Forest is better suited for workloads with higher core count and density. Both share the same platform, allowing adjustments according to customer needs.
The Intel Core Ultra family, also known as Meteor Lake, will be the first line of consumer processors to include a Neural Processing Unit (NPU). This solution will offer low latency AI inference, even without an internet connection, and will work in conjunction with the CPU enhanced by the Intel 4 manufacturing process, as well as a more powerful GPU based on the same architecture as Intel Arc gaming graphics cards.
Cooperation between accelerators is exemplified by different types of Artificial Intelligence models: the GPU is ideal for algorithms that use 3D media or applications, while the NPU is optimized for low-power tasks and long-use features, and the CPU is indicated for lighter AI models with fast response and low latency. The tech giant announced a collaboration with Acer to develop AI features optimized for Intel Core Ultra chips, including “Acer Parallax,” which uses the NPU to create a 3D effect in user-selected images.
Attempt to gain market share
By focusing on Artificial Intelligence, Intel is trying to return to a time when its conferences and news offered a roadmap for the computing industry. Previous mistakes have resulted in the loss of manufacturing capacity, allowing TSMC and Samsung to excel in this regard. Additionally, Nvidia has become a leading provider of AI accelerators, which help process artificial intelligence workloads.
The Intel leader reaffirmed the company's commitment to updating its production technology at an unprecedented pace. As part of this effort, the Intel 3 chip manufacturing technique will be released later this year. In addition to trying to catch up with rivals, Gelsinger is opening its manufacturing network to external customers – even if they compete with internal products.
Finally, software developers will have the opportunity to test their projects early using Intel's Developer Cloud, allowing them to try out their workloads on company-hosted computers over the internet. This will help customers speed up their work when Intel chips are generally available.
- Intel Core Ultra is announced as “processor for AI”;
- 5th generation Intel Xeon arrives with improvements in AI work;
- Intel Developer Cloud promises to democratize AI reach in the cloud.
Text proofread by: Pedro Bomfim (19 / 09 / 23)