Integrated network cards or integrated network controllers are a fundamental piece in today’s PCs, since without them it would not be possible for us to connect to the network of networks or to other computers. But, what has been its evolution?
What are integrated network cards?
An integrated network controller or NIC is nothing more than a processor in charge of communicating two computers with each other through a network. Since the communication of information over long distances does not use the methods that are used internally in the hardware, a piece is needed that converts the electrical signals into signals that propagate through a different medium.
This requires that each of the computers that communicate with each other have the integrated network card to communicate with each other. In the event that we have several computers within the same network that need to communicate with each other and if we have several computers, a central router is necessary that is responsible for carrying out the communication and the corresponding network requests between the different computers connected to said network. .
Networks on the first PCs
The first computers that went on the market were not designed to be able to work in a corporate network, since this was reserved for the mini-computers located in the different companies where each user had a terminal. Actually, in these environments, there was no network, but a central computer whose access and processing time was shared by a series of users, which were few.
Nevertheless. one of these early computers, the Apple II, soon necessitated the need to communicate information in an office environment. All thanks to the success of the first spreadsheet, VisiCalc, but back then file sharing in a work environment consisted of passing storage disks from one user to another. Things did not improve with the launch of the IBM 5150, since for IBM the most lucrative market was the business market and its mini-computer division gave them large amounts of capital, so the IBM 5150 also did not have the ability to connect through a network with other serial computers.
For years the ability to connect to a network on a personal computer was an option that required a network card that was sold separately. Because it was something that home users did not need and if businesses this motivated the high prices for network cards in a personal computer.
Network cards reach the home PC
The first generation of computers with the ability to connect to the Internet did so through Modems, a piece of hardware that was connected directly to the telephone network to connect to the network. Its biggest limitation? It only allowed accesses of a few Kilobytes per second and although year after year they were evolving in terms of their speed in a short time they were limited in terms of their capacity.
It was with the arrival of what we call broadband internet and therefore with the jump to much higher speeds that modems gradually disappeared to be replaced by routers that required the use of an ethernet connection on a PC to be able to use them. At the same time, years of integration and evolution in the PC had meant that what were previously extremely expensive network cards could be integrated into the motherboards of both desktop and laptop PCs.
Today the capabilities of the network card are integrated into the chipset of our computer, but the evolution of network cards has gone beyond providing access to the Internet and they have become a fundamental piece in the face of evolution. of hardware, not only to connect several computers together, but for creation both within the PC and even within an APU or SoC.
Evolution of integrated network cards
NICs have undergone a gradual evolution over time, increasing the capacity of the network processes that they can manage at the same time. Most NICs that are mounted on PCs have limited capacity, so it is the CPU that has to manage these processes. Unfortunately in order to save costs and the fact is that a piece that is often overlooked when looking at the specifications many PCs come with this NIC.
The second type, however, does have the ability to reduce the load on the CPU when managing the amount of requests made by the PC. These NICs are usually found integrated into advanced and therefore high-end motherboards. Not only do they free up the CPU, but their communication is much smoother. Of course, they are not programmable and therefore it is not possible to use their ability to move data more efficiently to different situations.
The third and most advanced type are SmartNICs and they will be key for the future, not only do they have more processing capacity than any other type of NIC. But they also serve to accelerate the movement of data between CPUs and GPUs in supercomputers that are used for data centers where the amount of network processes is much higher. The idea in this case is to use SmartNICs and the fact that they are programmable to manage data movements more efficiently. Today, SmartNICs are implemented through the use of SoCs, specialized ASICS such as NVIDIA DPUs based on Bluefield Mellanox technology or using FPGAs.
What does a “network card” do to my processor?
In the future, the intercommunication of the different components in an APU or a SoC will be done through a SmartNIC integrated in the chip itself. This concept, which is known as NoC, consists of converting each of the classic elements of an SoC into an element within a network that communicates to the central SmartNIC for communication between them and with the memory.
Regarding its road maps, NVIDIA has shown us GPUs with a SmartNIC integrated inside them and in the case of AMD, the purchase of Xilinx aims at integrating SmartNICs based on Xilinx FPGAs within AMD processors. . It must be borne in mind that as the number of CPU cores and GPUs grows more and more, the conventional structures to communicate the elements of an SoC become less and less efficient.