AI Everywhere – Intel introduces the Core Ultra, Intel EVO and the AI ecosystem up to Gaudi

Intel has taken a significant step in its development with the new “Meteor Lake” CPU platform, marketed as “Core Ultra”. These processors are the first from Intel to use the “Intel 4” process, which is based on EUV lithography, and the first Intel client CPUs where parts of the SoC are manufactured by TSMC. “Meteor Lake” is characterized by a disaggregated structure in which wafer chips from different sources are fused into one whole. The platform uses up to six “Redwood Cove” P-cores and eight “Crestmont” E-cores. A special feature is the ability to completely deactivate the compute tile in order to increase efficiency. The integrated GPU (iGPU) is based on Intel ARC with up to 128 execution units and offers improved efficiency and new functions such as ray tracing units and Intel XeSS for AI upscaling. In addition, “Meteor Lake” supports new standards such as PCIe Gen 5 and WiFi. That’s the short version, but we’ll go into more detail in a moment.

The new Core Ultra for the mobile sector: features and technical details

The Intel Core Ultra is the first processor based on Intel 4 process technology and marks a significant evolution in Intel’s architecture. It leverages Foveros 3D packaging technology, which enables advanced intellectual properties (IPs) to be combined with leading-edge processing to optimize performance and capability. The new architecture of the power cores (P-cores) improves the number of instructions per clock cycle (IPC). In addition, the new efficiency cores (E cores) and low power efficiency cores (LP E cores) provide scalable, multitasking performance that delivers higher efficiency compared to competitive ultra-thin PC products.

The Intel Core Ultra also includes an integrated Intel® Arc™ GPU that offers up to eight Xe cores, AI-based Xe Super Sampling (XeSS), support for DX12 Ultimate and increased graphics performance over the previous generation. The GPU supports advanced graphics features including hardware-accelerated ray tracing, mesh shading, AV1 encoding and decoding, HDMI 2.1 and DisplayPort 2.1 20G. Intel’s latest NPU, known as Intel® AI Boost, is specifically designed to handle long-lasting AI workloads with low power consumption. It complements AI processing on both the CPU and GPU and enables more efficient power utilization compared to the previous generation.

The new processor family also offers a wide range of features, including up to 16 cores (six P-cores, eight E-cores, two LP E-cores) and 22 threads, a maximum turbo frequency of up to 5.1 GHz, as well as a maximum memory capacity of up to 64 GB LP5/x and up to 96 GB DDR5. Wireless features include integrated Intel® Wi-Fi 6E (Gig ) and support for discrete Intel® Wi-Fi 7 (5 Gig) for higher speeds and reliability. Intel® Killer™ software optimizes network connectivity and prioritizes traffic for gaming and creative applications.

Intel® Bluetooth® LE Audio enables an immersive, high-quality audio experience on the PC and supports new Bluetooth® Auracast™ features. In addition, Thunderbolt™ 4 offers connection speeds of 40 Gbps and enables connection to multiple 4K monitors, fast storage devices and other accessories. Support for OpenVINO™ toolkits with minimal code changes and automatic device detection improves AI performance by directing workloads to the right compute engine and enables a more efficient workflow with ONNX and ONNX Runtime for faster inferencing.

The nomenclature is new and also has a bit of a twist. According to various sources, the “ultra-free” names Core 3, Intel Core 5, Intel Core 7 and various variants of the Intel Core Ultra series indicate that these new processors are not only based on Meteor Lake-H technology, but will also incorporate elements of the Raptor Lake refresh. The naming of the processors containing the suffix “processor” appears to follow a systematic pattern, although exact specifications for each variant have not yet been fully revealed.

Interestingly, the information suggests that these new processors will not only be based on Meteor Lake-H technology, but will also incorporate elements of the Raptor Lake refresh. The naming of the processors that include the suffix “processor” seems to follow a systematic pattern, although exact specifications for each variant have not yet been published in full:

Intel Core 3 Processor XXX
Intel Core 5 processor XXX
Intel Core 7 processor XXX
Intel Core Ultra 5 processor XXX
Intel Core Ultra 7 processor XXX
Intel Core Ultra 9 processor XXX

Intel Core Ultra Processors Press Deck_FINAL_12.13 V2

 

Meteor Lake now relies on “Intel 4”

Intel’s new “Intel 4” manufacturing process represents a significant technological advance and an important development for the company. The process is Intel’s first to utilize extreme ultraviolet (EUV) lithography, a key technology that enables more complex and finer transistor patterns to be created on silicon wafers. This leads to higher performance and efficiency as well as reduced manufacturing defects.

Important technical features of the “Intel 4” process include:

• Reduced structure sizes: The fin pitch size has been reduced to 30 nm, which is about 0.88 times smaller than Intel 7. The contact gate pitch is now 50 nm, compared to 60 nm previously. The minimum metal pitch for the bottom layer (M0) is also 30 nm, which is 0.75 times smaller than Intel 7, resulting in a doubled density compared to the Intel 7 process.
• Improved performance and energy efficiency: Processors manufactured using the Intel 4 process offer up to 21.5% higher clock speeds with the same power consumption compared to current Intel processors. Alternatively, the increased speed can be traded for better energy efficiency by achieving the same clock speed with 40% less energy.
• New materials and design updates: Intel is using a new combination of materials called Enhanced Copper (eCu), combining copper with cobalt. This is intended to improve performance while maintaining resistance to electromigration. Compared to pure cobalt, eCu offers better performance in terms of clock speeds.
• Density and transistor designs: Intel 4 offers double the logic cell density compared to Intel’s previous generation processors, allowing the integration of more arithmetic circuits into one processor. The SRAM cell density has also been improved by more than 20%.
• Meteor Lake processors: With the Intel 4 process, Intel plans to introduce the Meteor Lake processors, which contain six cores for high performance and eight cores for energy efficiency, an integrated GPU and an AI module.
• Evolution to Intel 3: Intel plans to further improve the Intel 4 process to an even more advanced manufacturing technology called Intel 3. This technology is expected to offer up to 18% more performance than Intel 4.

This information makes it clear that the Intel 4 process is an important step for Intel to further increase the performance and energy efficiency of its chips and to remain at the forefront of the semiconductor industry.

Intel EVO

The latest generation of Intel® Evo™ Edition laptops are characterized by a combination of high performance and intelligent functions. They offer improved performance that not only ensures efficient cooling, but also enables quiet use. At the same time, the responsiveness of the device remains high. These laptops are designed to last longer, with a battery life that lasts all day, even with intensive use.

The devices have features for fast wake from sleep and fast charging, making them ideal for mobile use. In addition, they offer advanced video effects and high-quality cameras that meet modern video conferencing requirements. Intel® Unison™ expands usage possibilities by connecting the devices to tablets and smartphones, enabling broader device compatibility.

 

These laptops are not only technically advanced, but also designed with environmental compatibility and sustainability in mind, which is underlined by the EPEAT Silver certification. The integrated Intel® connectivity software optimizes network connections and ensures efficient use for productivity tasks, collaboration and entertainment.

The Engineered for Intel Evo program ensures that accessories, such as Bluetooth accessories, docks, monitors and storage solutions, meet high performance standards to deliver a smooth PC experience. This year, the program is expanding its collaboration with various industry partners to increase the range of compatible and high-quality accessories.

Intel Evo press deck_FINAL

 

AI Everywhere – More than just a buzzword?

Both Intel Core Ultra and the 5th generation of Xeon are also being used in areas that you might not expect. In fact, multiple use cases for edge computing represent the fastest growing segment of computing – expected to grow to a global market of $445 billion by the end of the decade – with AI representing the fastest growing workload. In this market, edge and client devices drive 1.4 times more demand for inference than the data center.

In many cases, customers will deploy a mix of AI solutions. A good example of this is Zoom, which runs AI workloads on Intel Core-based client systems and Intel Xeon-based cloud solutions within its all-in-one communication and collaboration platform to deliver the best user experience and cost. Zoom uses AI to suppress the neighbor’s dog barking and blur the messy home office in the background with nice blurring, as well as create and email meeting minutes.

To make AI hardware technologies as accessible and user-friendly as possible, Intel is building optimizations into the AI frameworks that developers use (such as PyTorch and TensorFlow) and providing basic libraries (via oneAPI) to make software portable and performant across different types of hardware. Advanced developer tools, including Intel’s oneAPI and OpenVINO toolkit, help developers leverage hardware acceleration for AI workloads.

The Intel Core Ultra and AI

Intel Core Ultra represents a significant evolution in Intel’s chip design and aims to strengthen the integration of artificial intelligence (AI) into PCs. This development follows Intel’s previous innovations, such as Intel® Centrino®, which enabled laptops to connect to Wi-Fi. One of the key components of the Intel Core Ultra is an integrated AI acceleration unit called a neural processing unit (NPU). This unit is designed to enable more efficient use of energy for AI tasks, with an improvement in energy efficiency compared to previous generations. In addition, both the GPU and CPU of the Intel Core Ultra are designed to support and accelerate AI applications.

Intel has partnered with over 100 software vendors to develop a variety of AI-powered applications for PCs. These applications span multiple domains, from creative to productive tasks. The goal is for these applications to run efficiently on PCs equipped with Intel Core Ultra, especially when compared to other platforms. For example, the performance of Adobe Premiere Pro is expected to be improved on these systems.

Intel Core Ultra-based PCs are already available on the market and it is expected that a wide range of laptop and PC models will be equipped with this technology over the next year. Forecasts suggest that by 2028, a large part of the PC market will include AI PCs, which could bring new opportunities for work, learning and creative activities.

The Intel Xeon and AI

The 5th generation Intel Xeon processor family, also unveiled today, delivers improved performance and efficiency. Compared to the previous Xeon generation, these processors deliver an average performance increase of 21% for general purpose computing and enable 36% higher average performance per watt for a range of customer workloads. Customers following a typical five-year refresh cycle and upgrading from older generations can reduce their total cost of ownership by up to 77%.

Xeon is the only mainstream data center processor with built-in AI acceleration. The new 5th generation Xeon offers up to 42% higher inference and fine-tuning on models with up to 20 billion parameters. It is also the only CPU with a consistent and continuously improving set of MLPerf training and inference benchmark results. Xeon’s built-in AI accelerators, along with optimized software and enhanced telemetry capabilities, enable more efficient and manageable deployments of demanding network and edge workloads for communications service providers, content delivery networks and broad vertical markets, including retail, healthcare and manufacturing.

For example, during today’s event, IBM announced that the 5th generation of Intel Xeon processors on its watsonx.data platform has achieved up to 2.7x better query performance compared to previous generation processors. Google Cloud, which will deploy 5th Gen Xeon next year, noted that Palo Alto Networks experienced a 2x performance boost in its deep learning threat detection models by leveraging the built-in acceleration in 4th Gen Xeon via Google Cloud. And indie game studio Gallium Studios used Numenta’s AI platform running on Xeon processors to improve inference performance by 6.5x over a GPU-based cloud instance, saving cost and latency in their AI-powered game Proxi.

This kind of performance opens up new possibilities for advanced AI – not only in the data center and cloud, but also in networks and edge applications worldwide.

Fun with the AI

I couldn’t resist this subheading, but I had to. But the background is quite serious. Pat Gelsinger gave an update on Intel Gaudi3, which will be released next year as an AI accelerator. He presented the next-generation AI accelerator for deep learning and large-scale generative AI models for the first time. Intel has seen rapid expansion of its Gaudi portfolio due to growing and proven performance benefits combined with a competitive total cost of ownership (TCO) and pricing. With the increasing demand for generative AI solutions, Intel expects to capture a larger share of the accelerator market by 2024 with its range of AI accelerators, led by Gaudi.

The information and press material for this news was provided in advance by Intel on the sole condition that the NDA deadline of December 14, 2023 at 3 p.m. be met. There was no obligation to publish.