(Image source: Intel)
Faced with challenges and pressures from all sides, chip giant Intel began to fight back.
At the inaugural Intel Innovation Summit on October 28, Intel CEO Pat Gelsinger unveiled a series of new products: alder lake, a 12th-generation Intel Core processor based on Intel 7-tech (3rd generation 10nm) technology, and a first-generation Intel Arc Sharp alchemy series of high-performance discrete GPU graphics •Partnered with Google to deploy the ASIC-based IPU processor product "Mount Evans"; Intel Tofino 3 Intel Fabric Processor (IFP); DeepRec End-to-End Toolkit co-built with Alibaba; New OneAPI 2022 Toolkit and new OneAPI Center of Excellence; Disclosure of Information related to the next generation (fourth generation) Intel Xeon Scalable processor "Sapphire Rapids"; and with Alibaba and Baidu , Google Cloud, Microsoft, SiPearl's ecological cooperation, etc.
Kissinger pointed out that the use of oneAPI to develop software can optimize chip advantages and hash rate improvement. He argues that Intel co-founder Gordon Moore's Law hasn't disappeared. "Moore's Law is still valid, and today, we predict that it will remain even faster than Moore's Law in the next decade."
Wang Rui, senior vice president of Intel and chairman of Intel China, said in an interview with Titanium Media App that in order to achieve the goal of "returning to the peak of the industry before 2025", Intel will achieve 4 process upgrades within 5 years, which is far more than the rhythm of 18 months to 2 years of process iteration of Moore's Law.
"In the past few years, we have fallen behind in terms of process, and if we want to catch up, we must be at a faster pace than others to surpass." Wang Rui said that today Intel is running a marathon at the speed of sprinting, "but everyone does not have to worry about running, because this is not a person, a company running, but with the entire industry to run, the baton will not be lost." ”
When talking about Apple, Alibaba, NVIDIA and other companies are all working on self-developed chips, Wang Rui told Titanium Media App that Intel and other companies in the same industry are competing with each other. In the past decade or so, NVIDIA can be said to be the leader in AI and GPU. In the past decade, Intel has also given Nvidia a venue without competition to achieve product iterative innovation and development. Now that Kissinger has returned, Intel's competitors are only themselves, and the company will accelerate the implementation of IDM 2.0, including investment and other ecological construction, to achieve innovation in Intel's entire ecosystem.
"For Intel, the first is that we must improve our execution and technical strength, so that we really have leadership, so why do we have the confidence to say that Moore's Law is not invalid, when we have a leading chip, we have the confidence to negotiate cooperation with customers; the second is that all companies are doing localization, but one day it will become a trend of 'the world will be united for a long time, and it will be divided for a long time', and it will eventually cooperate with Intel." Wang Rui said that Intel is a chip company that relies on an open ecosystem to optimize its own software and hardware platforms, which is one of the core competitiveness. So Intel is confident that it will win back customers like Apple.
This is the first time that Wang Rui has been interviewed by the media since wang rui was appointed chairman of Intel China in September this year. Rui Wang joined Intel in 1994 and has held various positions as General Manager of Intel's Technical Support Division (TEG) and General Manager of Marketing Group China. (For details, see Titanium Media above: "Intel appointed Dr. Wang Rui as chairman of china, and the current president Yang Xu will retire at the end of 2021")
Rui Wang, Senior Vice President of Intel Corporation and Chairman of Intel China (Source: Intel)
Wang Rui pointed out that Intel has been rooted in China for 36 years and will continue to develop in China, and IDM 2.0 will also be laid out in China in the future. She stressed that Intel will continue to play a role as a semiconductor industry leader.
<h2>Intel's 12th Generation Alder Lake chips are here!
</h2>
This time, there are more new Intel hardware products. It mainly includes compute SoC solutions including Core, processor optimization solutions from cloud to edge, data connection center solutions, and Intel AI ecosystem layout.
The first to make a splash is the long-rumored 12th generation Intel Core processor Alder Lake.
The 12th Generation Intel Core Series features a new high-performance hybrid architecture for the first time in intel 7 (3rd generation 10nm) process technology, the biggest change since 2015. The chip is equipped with up to 16 cores and 24 threads, that is, it has 8 performance P large cores + 8 performance E small cores at the same time, and the default turbo frequency can reach 5.2GHz and 3.9GHz, respectively. Up to 36% faster photo editing performance, up to 32% faster video editing performance, up to 37% faster 3D modeling performance, and up to 100% faster multi-frame rendering compared to Gen 11. Intel today introduced the first six desktop processors, including the intel core i9-12900K, the world's top gaming processor.
According to the Core i9-12900K with win11 platform in the Geekbench 5 database, the single-thread score is 1834 and the multi-thread score is 17370. While AMD's 16-core flagship, the Ryzen 9 5950X, scores 1652 for single-threading and 16442 for multi-threading, the Core i9-12900K is 11% and 5.6% higher than it, respectively.
This generation of Cores, which will contain 60 processors, will power more than 500 model designs, from high-end desktops to ultra-thin and light laptops, and is expected to deliver 28 SKU models of processors to OEM partners. Starting today, users can pre-sell the Intel Core i9-12900K on channels such as JD.com, priced at $589 (5,089 yuan), and is expected to ship on November 4.
In addition, Intel today introduced a new generation of Theradiol Alchemy high-performance discrete graphics SoC series, code-named "Alchemist", which will provide the new Xe Supersampling (XeSS) upscaling technology, leveraging the advantages of machine learning and Alchemy's built-in XMX AI accelerator to provide high-performance and high-fidelity visuals and ensure widespread use in a wide range of games and hardware. In addition, Alchemist will also support Intel Deep Link technology, new computing power, and more. Microsoft and Intel are also working together to bring a complete data science toolchain to Windows 11, which will be the first to be available on the new Surface Laptop Studio.
In terms of cloud-to-edge infrastructure, Intel announced optimizations for the next generation of Xeon Scalable processors, Sapphire Rapids, which is working with the open source community and a large ecosystem partner to make it easier for developers to develop on their next-generation processors, and intel Ponte Vecchio GPUs and software platform oneAPI to support SiPearl's microprocessors. In addition, Intel emphasized that cloud application developers can use the latest 3rd generation Intel Xeon Scalable processors in major cloud service providers including Alibaba, Amazon AWS, Baidu, Google, Microsoft, Oracle, and Tencent.
In the Data Connectivity Center segment, Intel announced a deep partnership with Google Cloud to deploy AT&T's upcoming virtual wireless access network for the data center and cloud infrastructure, leveraging Intel Intelligent Fabric's intelligent fabric, ASIC-based IPU "Mount Evans" solution, and partnering with AT&T, one of the largest U.S. communications carriers, as its chip supplier.
In the AI ecosystem segment, Intel announced that the first 10 gigabit Aurora supercomputer in the United States will be equipped with the 4th generation Intel Xeon Scalable processor "Sapphire Rapids" and Intel's next-generation Ponte Vecchio GPU, which is expected to provide more than 20 billion exaFLOPS of double precision (FP64) computing performance per second, which is expected to become the first computer system in the United States to achieve a 20 billion-level system. And Intel plans to achieve a 30x increase in total AI performance through the Saphire Rapids Xeon Scalable processor.
It is worth mentioning that Intel also announced today that it has partnered with Alibaba to develop an end-to-end toolkit, DeepRec, which uses Intel AI technology to empower Alibaba's recommendation engine to facilitate deep learning training of recommendation systems and deploy extremely large-scale recommendation services with trillions of samples. And the Intel Optimization AI Toolkit has cooperated with the open source community, as well as Amazon, Baidu, Facebook, Google and Microsoft, and is compatible with Pandas, scikit-learn, MXNet, PaddlePaddle, PyTorch, TensorFlow, ONNX Runtime and other software architecture platforms to optimize Intel hardware upgrades.
<h2>Committed to optimizing chips with software development</h2>
One of the highlights of this innovation conference is that Intel will advance software algorithms to optimize chip capabilities.
Intel today released a Focused Investment Program for Developers, including an upgraded, unified, and more comprehensive Developer Zone, oneAPI 2022 toolkit, and a new OneAPI Center of Excellence (CoE) designed to optimize access to resources and simplify architectural development across central processing units (CPUs) and accelerators.
It is reported that the oneAPI 2022 toolkit released by Intel has 900 new features, adding cross-architecture software development capabilities for CPUs and GPUs to improve productivity.
Wang Rui said that IDM 2.0 is Intel's new banner, including expansion of production capacity, investment in advanced packaging technology, etc. It takes a long time to see the landing, but financially, the software does not need a large investment, itself needs a team with strong technical capabilities like Intel CTO. From the cloud to the edge, Intel will continue to focus on the software direction.
"We built our own hardware bottom layer, which is our hard power, and the above software development and oneAPI are soft power, and the two together are Intel's core competitiveness." Wang Rui told the Titanium Media App.
In an interview with Titanium Media App and other interviews, Wang Rui said that China's open source market is very large, oneAPI development in China has a certain basic commonality, and the Intel oneAPI toolkit has a strong heterogeneous computing cross-architecture, cross-vendor technical architecture bottom, making it more convenient for developers to develop software, systems, etc.
Rui Wang emphasized the importance of empowering open ecosystems, ensuring that developers can use their preferred development tools and development environments, and consolidating trust and partnerships with cloud service providers, open source communities, start-ups, and more ecosystem partners.
Looking at competitors, NVIDIA has established a huge ecosystem through the CUDA platform, and the most critical has created a series of standards. In short, the CUDA platform is a set of software and hardware ecological standards for NVIDIA to establish parallel computing, most AI chips need to match it, in addition to CUDA, NVIDIA is also building new platforms in more fields.
At present, Intel has also entered the GPU track and strengthened software capabilities. For competition, Wang Rui told Titanium Media App: "In the past decade or so, Nvidia can be said to be the leader in AI and GPU. Intel has also given Nvidia a venue where it has no competition over the past decade. Since Pat came back, we've started investing in this, and of course that's a long road, and we're now in a very small share of GPUs, but it also gives us an opportunity. We must build an ecology, if there is no ecology, just a GPU can not rely on it alone to compete, related software development we will invest. ”
When asked whether Moore's Law is invalid, Wang Rui said that Moore's Law is not a physical law, but an economic law of human demand for semiconductor technology. Intel believes that as long as it can provide economic value, it has grasped the essence of Moore's Law, and there is no need to worry about whether to consider the judgment that Moore's Law is invalid. What Intel wants to do now is to achieve an extension of Moore's Law by upgrading chip technology.
When asked when the current global "chip shortage" will improve, Wang Rui gave a forecast that under the factors of increased demand for chips, unbalanced supply chains, and lack of raw materials, it is expected that the semiconductor shortage will not ease until 2023.
Despite Intel's release of many new products, the stock price has not risen sharply. As of the close of trading at 16:00 EST on October 27, Intel (NASDAQ:INTC) shares fell steadily by 0.18% to close at $47.89 per share, with a market capitalization of $194.769 billion.
(This article was first published on titanium media App, author | Lin Zhijia)