Tag: Chip

Categories
Ξ TREND

China breathes a sigh of relief: chip designer Loongson is preparing a brutal 64-core CPU


The US sanctions on China do not only prevent the arrival of the most advanced lithography equipment that ASML, Tokyo Electron or Canon have to the plants of chip manufacturers in the country led by Xi Jinping; They also hinder the arrival of the most advanced processors that have NVIDIA, Intel or AMD, among other companies. This situation has forced Chinese CPU designers to step on the accelerator and develop their own general-purpose chips and for artificial intelligence applications.

Loongson is one of the most capable Chinese processor designers. We have been tracking it for months because in the current climate of geopolitical tension it has become more relevant than ever, and there is no doubt that its cruising speed is high. At the end of December 2022, this company launched its 32-core 3D5000 CPU, a general-purpose processor with LoongArch microarchitecture implemented by this company on the MIPS architecture. But since then he has not rested on his laurels.

The LS3C6000 processor is a little beast for servers

Loongson already has its new LS3C6000 processor ready, a general-purpose CPU for servers implemented on the LoongArch instruction set and equipped with 16 LA664 cores. Nothing really surprising so far. The picture begins to get interesting if we notice that, according to Loongson, the microarchitecture of this chip allows it to deliver performance comparable to that of an AMD processor equipped with the same number of Zen 3 cores.

The LS3C6000 processor is implemented on the LoongArch instruction set and equipped with 16 LA664 cores.

It is evident that it is most prudent for us to approach this statement with some suspicion given that Loongson is an interested party. And, furthermore, at the moment no independent media has published its own analysis of this processor. Even so, it is not at all unreasonable to accept that this company actually already has a microarchitecture capable of going head-to-head with AMD’s Zen 3. In any case, we have not yet investigated the most impressive feature of this development of this Chinese company.

And the cores of the LS3C6000 CPU are interconnected using high-performance links that Loongson has named DragonChain. On paper, this technology will allow this company to scale this CPU and launch new versions in the short term. equipped with 32 or 64 cores.

Additionally, Loongson is confident that its LoongArch 6000 microarchitecture will allow it to match the IPC (instructions per clock cycle) of AMD’s Zen 3 processors. Although this is true, it will still be one step behind Intel and AMD, but there is no doubt that for this Chinese company it will be a success to reduce the distance that separates it from the most advanced designs that its American competitors have.

Categories
Ξ TREND

Intel boosts quantum computing with Tunnel Falls, its first silicon spin qubit chip


Intel has released its new chip focused on quantum research, Tunnel Falls, its first silicon spin qubit device, which it has made available to the academic community to explore this technology and accelerate research development.

A qubit (or quantum bit) is the basic unit of information in quantum computing. While a bit can only represent a single binary value – that is, 0 or 1 – the qubit can represent a 0, a 1 or any ratio of 0 and 1 in the superposition of both states.

In this way, it allows quantum algorithms to process information in a much shorter time than the time required by a classical system. For this reason, quantum computing is being used to carry out discoveries in different fields such as health care, energy or environmental systems.

Within this framework, Intel has launched Tunnel Falls which, as described in a statement, is the company’s first silicon spin qubit device. Specifically, it is a 12-qubit silicon chip manufactured on 300-millimeter wafers at the D1 manufacturing plant.

This device takes advantage of Intel’s most advanced industrial transistor manufacturing capabilities, with technologies such as extreme ultraviolet (EUV) lithography and gate and contact processing techniques.

As Intel explains, in the case of silicon spin qubits, the information – the 0s and 1s – is encoded in the spin (up or down) of a single electron. This translates to each qubit device being “essentially a single-electron transistor.” As a result, Intel can manufacture it using a flow similar to that used in a complementary metal oxide semiconductor (CMOS) logic processing line.

Furthermore, another of the advantages of silicon spin qubits is that they outperform other qubit technologies for “the synergy it offers with cutting-edge transistors.” This is because they are about the size of a transistor and are therefore “up to a million times smaller” than other types of qubits, which measure around 50 square nanometers. The size of the silicon spin qubits thus allows for “more efficient” scaling, according to Intel.

Following this line, Intel also benefits from being able to use CMOS manufacturing lines to produce this chip, since it allows it to use “innovative” process control techniques to “improve performance and performance.”

In fact, in the case of 12-qubit Tunnel Falls, a 95 percent throughput rate has been achieved across the wafer, as well as voltage uniformity resembling that of CMOS logic processes. Also, each wafer provides 24,000 quantum dot devices.

For all these reasons, as stated by the director of Quantum Hardware at Intel, Jim Clarke, the launch of the new chip is “the next step in Intel’s strategy to build a complete commercial quantum computing system.”

COLLABORATION WITH THE QUANTUM RESEARCH COMMUNITY

Intel has made Tunnel Falls available to the quantum research community in order to promote the development of this technology, facilitating its investigation. To do this, the company is collaborating with the University of Maryland’s Laboratory of Physical Sciences (LPS) and College Park’s Qubit Collaboratory (LQC).

As detailed by the technology, academic institutions do not have large-volume manufacturing equipment like Intel. In this sense, by facilitating Tunnel Falls, researchers can start working with this chip in their quantum computing projects, instead of having to make their own.

As a consequence, this collaboration makes possible a broader range of experiments, learning more about the fundamentals of qubits and quantum dots, and developing new techniques for working with multi-qubit devices, according to Intel.

Specifically, Intel is collaborating with LQC on the Qubits for Computing Foundry (QCF) program through the US Army Office of Research. This cooperation will help democratize silicon spin qubits by allowing researchers to gain experience working with this technology.

The first quantum labs to participate in the program are LPS, Sandia National Laboratories, the University of Rochester, and the University of Wisconsin-Madison. But, in addition, LQC continues to work with Intel to bring Tunnel Falls to other universities and research laboratories. Likewise, the information obtained in these investigations and experiments will be shared with the community so that the progress is even greater.

For his part, LPS Head of Quantum Information Science, Charles Tahan, has detailed that the LPS Qubit Collaboratory, together with the Army Research Office, intends to address the “difficult challenges facing qubit development.” ”, as well as “developing the next generation of scientists who will create the qubits of tomorrow”.

In the same way, Sandia National Laboratories has valued the possibility of using the Tunnel Falls chip. “The device is a flexible platform that allows Sandia quantum researchers to directly compare different qubit encodings and develop new modes of operation of qubits, which was not possible before,” said Dwight Luhman, Ph.D. and technical team member.

Categories
Ξ TREND

Intel’s plan against an unattainable TSMC: beat Samsung and consolidate itself as the second largest chip manufacturer


Just 24 hours ago we participated in the annual meeting in which those responsible for Intel in Spain take stock of their business during the year that we have left behind, and also share with the invited journalists your forecasts for the new year. This event has been led by Norberto Mateos, consumer director for the EMEA area and general director of Intel Spain, and one of his statements seemed juicy enough to analyze in this article.

According to Norberto, Intel intends to consolidate itself as the second largest semiconductor manufacturer in the global market. At first glance it may seem that this phrase does not say much, but nothing could be further from the truth. It says a lot. And he says it because it makes something important official: the distance that the Taiwanese company TSMC, which is the largest manufacturer of integrated circuits on the planet, maintains over its two most advantaged competitors, who are none other than Intel and Samsung, is insurmountable in the short and medium term. term.

And at this juncture, Intel has set its eyes on Samsung. Currently, TSMC’s global market share is slightly over 50%, while that of Intel and Samsung ranges from 17 to 20% for both companies. It is clear that a realistic plan requires paying attention to the nearest competitor and taking the necessary steps to distance yourself from it and overcome it. This is precisely what Intel plans to do with Samsung. And the IDM 2.0 strategy () that Pat Gelsinger launched shortly after becoming the general management of this company in February 2021 is everything in this purpose.

IDM 2.0 aims to increase Intel’s competitiveness and reinforce synergies

The steps that this company has taken under the leadership of Gelsinger during the last three years do not only aspire to respond to the needs of Intel itself in the field of semiconductor manufacturing; They also seek to expand their client portfolio and position this company as one of the largest producers of integrated circuits for third parties. The chip manufacturing plant infrastructure that Intel had in 2021 was not sufficient to materialize this double objective, so Gelsinger took the only possible path.

Over the next few years, Intel will invest at least $80 billion in the development of several chip factories.

Intel is making multibillion-dollar investments to expand and strengthen its network of semiconductor manufacturing, packaging, assembly and verification plants. The two factories that it is currently setting up in Arizona (United States) will cost it 20 billion dollars, but this is far from all. It is also building a $25 billion chip factory in Kiryat Gat, Israel; prepares the launch of a 30 billion dollar plant in Magdeburg (Germany); will invest 4.6 billion dollars in new facilities that will be located in Wrocław (Poland), and, finally, it will spend 13 billion more dollars on the expansion of its factory in Leixlip (Ireland).

Intel will not assume 100% of these investments because it will receive juicy subsidies from the governments of the countries involved. Even so, his bet is intimidating from an economic point of view. However, its recipe has one more ingredient that we have not yet investigated: the synergy it maintains with TSMC. Yes, these two companies compete in the field of semiconductor manufacturing, and they will do so even more intensely in the future, but they are also allies.

TSMC also makes chips for Intel. It has been doing this for many years, and this relationship of complicity will endure in the future as an important part of the IDM 2.0 strategy. What’s more, analysts at the financial services company Goldman Sachs argue that Intel will strengthen its relationship with TSMC in the short term referring to this Taiwanese company the manufacturing of part of its semiconductors. According to these technicians, in 2024 Intel will buy chips worth $5.6 billion from TSMC, and in 2025 this figure will increase to $9.7 billion.

We still have to focus on the last major pillar of Intel’s strategy to beat Samsung and consolidate itself as the second largest chip manufacturer on the planet: its intention to develop 5 nodes in just 4 years. At the moment it is on the right track. Several of its plants, including the one in Ireland, are already manufacturing chips with Intel 4 lithography on a large scale. And Norberto Mateos has confirmed to us that nodes 20A and 18A will go into production this year with the purpose of being ready for circuit manufacturing. integrated on a large scale in 2025. Be that as it may, we can be sure of one thing: the competition between TSMC, Intel and Samsung will reach unprecedented ferocity in 2024 and 2025.

Cover image: Intel

Categories
Ξ TREND

The US will not be able to contain China’s technological development. Chip industry experts predict it


The US government is doing everything in its power to prevent China from getting the cutting-edge semiconductors it needs to develop its military industry. Its strategy attacks two different fronts. On the one hand, it seeks to prevent Chinese integrated circuit manufacturers from obtaining the latest generation lithography equipment they need to produce highly integrated chips. In this field, the Dutch company ASML and the Japanese companies Tokyo Electron, Nikon and Canon lead the way.

The other essential ingredient in his recipe is to prevent China from taking over the most advanced semiconductors currently manufactured by NVIDIA, Intel or AMD, among other Western-aligned companies. In any case, the tool that the US Administration is using to assert its interests are sanctions against China and the countries in its orbit, including Iran and Russia. It has been implementing these prohibitions for several years and nothing seems to indicate that it will stop tightening them in the short term.

Every time the US Government approves a new package of sanctions, it strives to remind its allies that these measures are intended to hinder China’s access to the integrated circuits that are necessary to develop latest generation weapons. As could be expected, the Chinese Government has responded by drastically increasing the resources it allocates to the development of its semiconductor industry with a very clear purpose: to become independent in the medium or long term from technologies from abroad.

Authoritative voices are already rising predicting how difficult it will be to stop China

There is one indispensable component to the semiconductor industry that we haven’t talked about yet: the software used to design cutting-edge integrated circuits. It is currently in the hands almost exclusively of companies controlled by the US and its allies, so China urgently needs to have its own software tools specialized in chip design. And it looks like he’s going to have them soon. One of the Chinese companies that are already working in this area is SEIDA, and, curiously, its leader knows the American idiosyncrasy very well.

SEIDA will have its chip design software ready in early 2024

Liguo “Recoo” Zhang is Chinese, but he has lived in the US for several decades and has worked at Siemens EDA, the US subsidiary of this German company that dominates the chip design software market in China. Zhang currently heads SEIDA and in the business plan that he presented in 2022 with the purpose of attracting new investors, he stated that his company will have its chip design software ready in early 2024. However, this is not all. And in that document SEIDA defends that its plan involves “breaking the foreign monopoly.”

Presumably the software SEIDA is working on will be used by SMIC, Hua Hong Semiconductor and other Chinese IC manufacturers. In fact, SMIC is one of the investors in this company. Peilun “Allen” Chang, SEIDA’s chief operating officer, claims that Zhang and other former Siemens EDA employees left the German company as a result of US sanctions.

Their purpose is to transfer their knowledge to China because they are convinced that despite their efforts the US Government will not be able to stop China’s technological advance. These people have developed their professional careers in the US and presumably know the semiconductor industry well, so it seems reasonable to accept that they may be right. We’ll see.