The processor is often said to be the “heart” of the computer. And in a way it is like that. It can also be seen as the “brain”, showing in both cases that we are facing a vital component for the functioning of our computer . The memory or the hard disk, not to mention the graphics processor are also relevant, of course, but if we had to define an order of importance, the processor (or CPU) would be in first place.
The graphics processor, memory and hard drive would follow in that order. Then we would have the screen, the connectivity and other elements such as the keyboard, the battery (in the case of a laptop) and other accessories. Actually, all the elements are necessary for a computer to work, of course . A computer without memory or a storage system can hardly work. At least the computers that we know and with which we are familiar.
A little engineering, but little
To understand the role of the processor, we must talk about somewhat arid concepts in principle. Current computers follow the Harvard architecture model, which in turn is related to von Neumann. This architecture defines the essential mode of operation of computers, with arithmetic and logical instructions that are executed in the electronics of the processor. These instructions are the elements of which the computer programs are composed . The instructions are stored in memory. And the data, in turn, are stored in memory equally.
Both the programs and the data are stored in the storage units, but the computer has to take them to the RAM so that the processor can access them , in a continuous process of reading from the storage media, writing in memory , processing in the processor and writing in memory to then take the data to the storage units.
The processors are capable of executing instructions within a well-defined repertoire and with strict and irremovable operating rules. There are different sets of instructions, but those that are handled in most current computers are the so-called x86 instructions .
The processors execute (more or less) sequentially the x86 instructions of which the programs are composed, while they operate on the data . The operations are of the arithmetic type, using binary logic with 64-bit registers. At first, only 4 bits were worked. Then with 8 bits. After 16 bits, 32 bits and from a decade ago, 64 bits.
The rate at which the operations are executed is marked by the clock frequency. So, if we have a processor with a frequency of 3 GHz, we have to execute operations at a rate of 3,000,000,000 operations per second . That is, in a nanosecond three operations are executed. The more speed, therefore, the greater number of operations performed. As we will see, it is not exactly like that. But let’s go step by step.
Architecture and multicore
Currently, the best-known x86 processor manufacturers are Intel and AMD. Each of them has designed the electronics of their processors in a different way. In both cases they have to be compatible with x86 instructions, but Intel does it one way and AMD another . Each manufacturer has developed its own architecture. That is: a unique way to process instructions. It is as if the objective was to reach the same destination, but each of them chooses a different route, trusting that it will be the “good” considering the distance, possible traffic jams, galibo limits or the number of traffic lights . The end is the same, but there are many possible ways to get there.
Actually, a processor can execute more than one operation in each clock cycle. Before we said that the programs are executed in a “more or less” sequential manner. That is: in a clock cycle several instructions can be executed. Traditionally, Intel has been better than AMD in terms of IPC. That is, the “route” chosen by Intel to reach its destination has traditionally been more efficient if we look at the metaphor of the previous paragraph. But recently, AMD has made a radical change in its architecture, which is now called Ryzen, standing up to Intel. That is to say: AMD has recalculated the route to arrive and now has achieved a more agile journey.
In addition, processor manufacturers have been integrating several “cores” in each processor for years. In this way, several programs or processes can be executed at the same time . Currently there are processors with up to 32 cores, with two, four, six and eight the usual number of cores (cores) that work simultaneously in a processor.
The performance does not increase linearly with the number of cores: four cores does not mean that we have a 4x performance on a CPU with a single core. It can be 2.5x, for example.
Manufacturing and consumption technology
Processor manufacturers (Intel and AMD are the best known) also evolve manufacturing technology. A CPU can have several billion transistors, with nanometric sizes. Every so often (between one year and three years), manufacturers find a way to make the size of the transistors of which a processor is smaller . This makes them “fit” more in the same space and consume less energy. The increase in transistors has followed for five decades the so-called Moore’s Law , although in recent years it is beginning to question its validity. The number of transistors does increase (at the moment), but the performance does not grow at the same rate.
Another parameter that defines a processor is the energy consumption. We can say, to simplify, that the TDP is the parameter that defines what a processor consumes. Although it is not really like that from a rigorous point of view. The TDP characterizes a processor: thus, there are processors with a TDP of 5W, 15W, 25W, 45W, 65W or even more than 100W. The TDP is measured in Watts, and for the same architecture, the more Watts the TDP has, the more processing power the processor will have .
This is important, because Intel, without going any further, has “Core i7” processors with several cores. But a Core i7 of four cores with a TDP of 15W will have a performance about three times lower than a Core i7 quad-core with a TDP of 45W . They may look the same if we look at the specifications, but they are not in practice.
At the moment we have that a processor will have a performance that depends on its architecture (IPC or instructions per clock cycle), the number of cores, the clock speed in MHz and the TDP . Intel and AMD have their own architectures, both compatible with x86 instructions, although with substantial differences in how they are executed.
Something we have not said is that it is usual that the graphics processor is also integrated into the processor . It happens on almost all Intel processors and in some AMD families. It is useful, since we do not need to buy an additional graphic card. But its performance in games is very low, so it is not always the ideal solution. Especially for desktops or gaming laptops.
It is also interesting to know that it is usual for each core to work as “virtual” as if it were two. Thus, a processor that specifies that it has four cores and eight threads will be able to handle two processes simultaneously (more or less) for each core . This technique allows to make better use of the resources of the processor but does not increase by a factor of 2x the performance, much less.
Intel and AMD in the present moment
Intel and AMD have been competing for supremacy in the processor segment since the 1990s. Both companies have been developing new solutions that have allowed them to reach the current moment, with Intel already in the eighth generation of its architecture, codenamed Coffee Lake. And AMD with its brand new Zen architecture and Zen + , optimized version of Zen architecture. Coffee Lake is the successor of Kaby Lake, while Zen is the successor of Bulldozer.
Intel, in addition, has a family of very low power processors with an architecture specially designed so that consumption is minimal, with the Godmont and Godmont Plus architectures, Godmont’s successor. They are processors with a simplified internal electronics compared to Coffee Lake or Kaby Lake, which allows to reduce consumption, although at the expense of reducing performance as well . It is as if, to go to your destination, you chose a longer route (less performance), but with a path free of traffic lights, traffic jams and going at a constant and efficient speed for consumption.
Coffee Lake uses a 14 nm manufacturing technology, while Zen also uses 14 nm and Zen +, 12 nm. In general, a smaller size of transistors, better performance in terms of performance and energy efficiency. Intel is already working with 10 nm processors (Cannon Lake), while AMD has by 2019 the arrival of Zen 2, the successor architecture of Zen and Zen + , with transistors of no less than 7 nm.
Architecture is one thing and processors are another. In this sense, Intel has the Intel Core m, Core i3, Core i5, Core i7, Core i9 and Core X processors, as well as the Intel Pentium, Pentium Silver, Atom and Celeron , while AMD offers users their processors Ryzen and Threadripper . Currently you can find Intel Core processors with the Coffee Lake architecture, as well as Kaby Lake and AMD processors with Zen and Zen +. The differences between one and the other are not excessively large and if the prices fit, you can make an optimal purchase of a computer with a processor that is not state-of-the-art.
In the case of AMD, it is your business to focus on the Ryzen or Threadripper processors , be they Zen or Zen +. But better if you do not choose equipment with previous technologies. Both in the case of Intel and AMD, there are processors for laptops and desktops, as well as integrated graphics or without integrated graphics . In the case of laptop processors, the usual thing is that integrated graphics are the norm. For desktops, there are options with and without them.
As you can see, manufacturers do not always make it easy when it comes to choosing. We have the option to buy a desktop or laptop with the processor and the components that we choose (memory, disk, motherboard, graphics, etc.). If you want to have an idea about the potential performance of a processor, look at the TDP. The higher the TDP, the more processing power. For the same TDP, look for the processors of higher ranges. For example, a Core i7 of Intel with TDP of 15W for portable, will have a greater performance than a Core i5 and this a greater performance than a Core i3.
Within the same family of processors, and for a given TDP, look for processors with a more recent architecture. That is to say, a Coffee Lake Core i7 of 15W of TDP for portable, will yield more than a Core i7 Kaby Lake with a TDP of 15W . For the same number of cores, the processors with hyperthreading or SMT (two threads per core) yield more than those without SMT. And, generally, to more cores, more performance. Going from less to more we have, in Intel, the Atom, Celeron, Pentium, Pentium Silver, Core m, Core i3, Core i5, Core i7, Core i9 and Core X. Processors with integrated graphics has models with integrated graphics of AMD its maximum expression (Intel Core i5 and Core i7 “G”). The Intel Core i5 and i7 for enthusiasts are the “K”.
In AMD, the second-generation Ryzen outperform the first-rate Ryzen. The TDP is still a valid reference, as well as the choice of a superior range for the same TDP. Thus, a Ryzen 7 yields more than a Ryzen 5 and this in turn, more than a Ryzen 3 . The greater number of cores, the better. The AMD Threadripper is the ultimate expression of what AMD can do in the field of processors. AMD processors with integrated graphics are called APUs. For desktop computers we have the Ryzen “G”. And in laptops, AMD also has Ryzen processors optimized for this type of equipment. They are the Ryzen “U”. The Ryzen 2xxx are more recent than the 1xxx, so they will be more efficient and faster. All in all, you can find interesting offers if we sacrifice performance by price.