all about Intel’s new super-powerful AI chip

Intel has announced its new Gaudi2 AI chip, twice as powerful as the first generation. The company aims to compete with Nvidia and AMD in all fields, and also makes new GPUs. Find out everything you need to know.

The field of artificial intelligence is evolving. By 2022, according to IDC, spending on this technology is expected to increase by 20% to reach $ 433 billion. the the market growsand computer manufacturers don’t want to miss the AI ​​train.

In this context, Intel announced its new chip of artificial intelligence on May 10, 2022 during its Vision conference. The purpose of the company is specific to regain market share from NvidiaAMD and other competitors.

Gaudi2 vs. Gaudi1

This new Gaudi2 chip is designed by Habana Labs, which is based in Israel and acquired by Intel in late 2019 for $ 2 billion. He is twice as fast than the first version, and should be integrated into servers by the end of 2022.

L ‘GEMM front-end architecture (general matrix multiply) in Gaudi1 is back-ended by 10 Tensor Processor Cores (TPCs), but only eight of them were exposed to users. This chip is notably implemented in TSMC’s 16 nanometer process, and offers 24 MB of on-chip SRAM, four banks of HBM2 memory for 32 GB of capacity and 1 TB per second of bandwidth.

So far, Intel did not disclose the details about the architecture of Gaudi2. We do know, however, that this will be based on TSMC’s 7-nanometer process and will make it possible to increase to 24 TPC instead of 10. The new 8-bit FP8 data format is supported, as in the Hopper GH100 GPU launched in March 2022 by Nvidia.

This new data format allows to have both low resolution inference data and high definition training data in the same format, without having to convert models when from training to inference.

Gaudi2 chip is embedded 48MB of SRAM. It includes HBM2e memory strips that offer 2.45 TB/sec bandwidth. Their number has not yet been revealed.

It is equipped with 24 Ethernet ports at 100GB/second, or one for each TPC. The component must be plugged into a PCI-Express 5.0 port, and can consume 650 watts.

This new chip is a must offer performance is multiplied by 2.5 in comparison with Gaudi1. However, it is not yet known whether any changes to the architecture and rhythm have been made.

Chips such as those in the Gaudi range are allowed speed up math calculations specific to artificial intelligence. Alternatively, we can also mention the Nvidia H100 which was designed to accompany the AI ​​revolution.

They simplify and reduce the value of training in AI modelswho learns by processing complex real-world data to find patterns.

These elements allow in particular improve voice recognition or the autopilot system of autonomous vehicles. Intel’s automotive arm, Mobileye, is training its AI systems with Gaudi.

A third generation Gaudi3 chip is already in development and will bring higher performance, more memory and better network capabilities.

GPU vs AI chips

With Gaudi2 and its new GPUs, Intel’s goal is restore the leadership position of the computer market. Over the past two decades, this status of the company has gradually disappeared.

kasi, the CPUs that popularized it no longer in the spotlight. Today, GPUs are being exploited for artificial intelligence and the main manufacturer of these graphics processing units is Nvidia. That’s why Nvidia’s market cap is estimated at $ 424 billion, more than double Intel’s $ 181 billion.

Many manufacturers have developed specific AI chips, but Nvidia prefers to continue to focus on GPUs. These components can also be used for supercomputers and HPCs. This flexibility is Nvidia’s main selling point.

Ang Businesses love the versatility of GPUs, because it allows them to remain productive at all times and regardless of the evolution of an AI model. General Motors ’autonomous automotive business, Cruise, for example, leases Nvidia GPUs over Google Cloud’s infrastructure to take advantage of their more mature AI software and extreme flexibility.

Also, GPUs and their software can be fast adapt to the constant changes driving the AI ​​industry. For example, they can adapt to new architectures, new types of layers or to the integration of AI models.

The AI ​​chip war

Aside from Intel, many startups work with specialized AI accelerators. We can mention Graphcore, SambaNova Systems, Tenstorrent or Cerebras. According to the CEO of the latter, GPUs are more suitable than CPUs for AI, but remain too limited compared to dedicated chips.

A war looms in this new market for the next five years. To get out of the game, Intel could use an aggressive pricing strategy.

L ‘AI is no longer the preserve of giants such as Amazon and Google, and falling costs could open the door to new applications such as fraud detection, crop tracking or medical image analysis.

AI chips and GPUs: Intel is fighting in all fields

However, Intel is betting both in versatile GPU and AI accelerators expert. the GPU Ponte Vecchio provides its power to the Aurora supercomputer at Argonne National Laboratory which is expected to go live by the end of 2022.

Thereafter, in 2023, Intel will begin to sell the Ponte Vecchio to the wider market. The company also plans to build alternatives to this GPUcheaper and in larger quantities.

Ang The GPU branch is headed by Raja Koduriwho previously developed GPUs for AMD and Apple before joining Intel in 2017. He is also a leader the new set of Arc GPUs focused on video games.

The first product in this range, bearing the code name Alchemistis already on sale and new products will be released later this year for laptops and gaming PCs. The Battemage and Celestial successors is built on the roadmap that extends to 2025.

In conclusion, Intel is fighting Nvidia and AMD in all fields. The American firm wants to position itself as the third player in this profitable market, and meet the needs of all potential customers …

Leave a Comment