VLSI Technologies

History

In the 1980s, while the industry was experiencing a boom in home computers, other equally exciting events took place in the scientific part of the industry.

It is well known that Internet it was invented at the Agency for Advanced Defense Research and Development (USA DARPA), but this is not the only DARPA project that has a powerful impact on the entire industry. VLSI Project is just one of these developments: its relatively small fame is simply incommensurate with its importance. VLSI stands for Very-large-scale integration - an ultra-large integrated circuit, or SBIS. In the early eighties, everything went to the transition to such schemes, but in their development, engineers faced serious problems.

With the increase in the number of transistors that fit on the chip of the integrated circuit, it became more and more difficult to design processors, and when the number of transistors began to exceed a hundred thousand, old methods began to lead to errors. A new way of designing was needed, and it would hardly surprise anyone that the solution was to use a computer.

Caltech professor Carver Mead and Xerox PARC programmer Lin Conway proposed creating a computer-aided design (CAD) system that would help make processors of virtually any complexity. At that time, a supercomputer would be needed to work with such a program, so DARPA had to finance not only the creation of CAD, but also everything around: the development of workstations and even an operating system. Later, Sun Microsystems and Silicon Graphics will grow out of these projects, and a new UNIX branch, Berkley Distribution Software (BSD), will be created as the OS.

Mead and Conway believed that if the development of processors was better automated, then small firms or even students could do them during the course of their studies. The idea turned out to be not only correct, but also very successful: with the help of new tools, processors became much easier to design and it became possible to do this in isolation from production. Moreover, the new software made it possible to reveal hitherto hidden features of the processor structure.

In 2014, processor architectures are usually divided into two classes: CISC (Complex Instruction Set Computing) and RISC (Reduced Instruction Set Computing). There is a fundamental difference between these approaches, but it did not appear immediately.

Early eight-bit processors like the Intel 8080 or Motorola 6800 were only able to execute a few simple instructions. For example, there was no special instruction for multiplying numbers, this action required several processor commands - offsets and additions. This approach seems inconvenient, and therefore the decision to add more capacious instructions was intuitive.

It was also believed that operations embodied directly in hardware will be performed much faster than those performed in the form of programs. So in subsequent developments, the creators of the processors began to add support for more and more new instructions. Multiplying two numbers, for example, turned into one command, but the microcircuit device became more complicated, as it began to include a separate subsystem designed for multiplication. This is how processors with a complex set of commands appeared. This family includes subsequent Intel chips and other processors that were popular in the 1980s.

Not to say that a comprehensive set of teams does not have its advantages, but they had to pay a good price. If the first processors performed one simple instruction in one tick of the clock frequency generator, then more complex instructions began to require several cycles.

As part of the same VSLI project, UC Berkeley professor David Patterson conducted a study that groped for a different approach to processor engineering, which he called RISC. It turned out that if you limit the set of instructions to only those that can be executed in one cycle, then you can increase the speed of their execution and thus increase the overall performance. Everyday logic suggests that this should not happen: the programs are longer! But when it comes to systems of hundreds of thousands of components, everyday logic can relax, and modeling and simulation will give the right answer.

At the same time, Patterson managed to significantly reduce the influence of the bottleneck of the von Neumann architecture - the slow channel between the processor and RAM. RISC has more registers than CISC, and this allows you to access RAM less often - especially if the program is passed through an optimizing compiler and uses resources profitably. This approach works even better in multi-core or multiprocessor systems where multiple computers access the same memory. The less often they do this, the less often each of them has to wait for their turn and, accordingly, the greater the performance increase.

2022: Intel to pay $950 million to steal technology for its processors

On November 15, 2022, a jury in Texas fined Intel $948.8 million for illegally using VLSI Technology's proprietary technologies in its processors.

The VLSI firm is linked to venture capital firm Fortress Investment Group, which in turn is part of the giant. During SoftBank Group the trial, the plaintiffs stated that Intel, without the necessary permission, used Cascade Lake Skylake and secure processing technologies in its microprocessors of generations. data

Intel will pay $950 million to steal technology for its processors

An Intel spokesman said the corporation "strongly disagrees" with the court order and plans to appeal, and that the case is "one of many examples showing that the U.S. patent system is in dire need of reform." VLSI did not comment on the verdict.

In March 2021, Intel lost another patent lawsuit with VLSI regarding microchip manufacturing technologies. Then the Texas court ordered Intel to pay almost $2.2 billion in favor of the plaintiffs. In April 2021, VLSI lost in another patent dispute with Intel: then it was about developments regarding methods for adjusting certain processor parameters. VLSI received the corresponding patents from NXP Semiconductors NV.

Intel lawyers say that the corporation's engineers are developing their own advanced technologies and that its modern microprocessors do not require outdated solutions, patents for which VLSI owns. Intel also says that VLSI is a "patent troll" created by Fortress Investment Group for the sole purpose of filing lawsuits against IT market participants and receiving billions of dollars from them as compensation for patent infringement.^[1]

2021: Intel to pay $2.2 billion for VLSI Technologies

In early March 2021, a jury in the United States sentenced Intel to pay a fine of $2.18 billion for the illegal use of technologies for which patents belong to VLSI Technology. Read more here.

1990: Participation in the creation of ARM

In 1990, the company, along with Acorn Computers and Apple, became one of the creators of ARM. VLSI Technologies is the direct heiress of the VLSI Project, which was involved in the design and production of integrated circuits. What was important to the future^[2] joint venture was that VLSI could provide ARM's own computer-aided design system]. VLSI itself needed a new processor customer.

Notes