And exactly what happened to Intel? The once irrelevant athlete in the manufacture of processors and microcircuits immediately yields to competitors in almost absolutely all possible indicators. CPUs through AMD turned out to be more thoughtful, and manufacturing technology through TSMC is more efficient. It looks like Intel has absolutely gone astray.
Already for the bazaar of mobile PCs, where the culprit has been the irrelevant leader for decades, Intel processors dropped the freshly blended Renoir through AMD.
Circumstances are so bad that Apple has announced schedules to end its relationship with the manufacturer and start bleeding ARM-based personal chips. Worse, rumor has it that Intel itself is planning a partnership with TSMC for the sake of releasing separate product options once, including the first custom graphics card. Indeed, this can become a complete humiliation for the sake of the organization.
Or is it just speculation? Against all the difficulties, it happened that Intel earned an unprecedented amount of money for a year - 72 billion. Ultimately, the general problem of the manufacturer is that it does not keep up with the dynamics of demand from the side of the so-called hyperscale data centers. These are companies like Amazon, Microsoft, Google, Facebook, and others, which simply lack a sufficient number of Xeon processors. However, consuming significant began to believe that soon Intel would return to its previous rut in comparison with the manufacture of chips and CPU microarchitectures.
How can one attest to the troubles and misfortunes of Intel in a nutshell? “10 nanometers,” I would say. And it's not so much the failure of chip technology - such arguments can be thrown down for any microarchitecture-developing company that has been resting on its laurels for decades. But 10 nanometers! This is a catastrophe.
By this “10 nanometer” consideration, we mean an electrotechnical move or assembly used to make computer chips. Theoretically 10 nm is the size of the smallest ingredients inside the chip. However, for practice, the names related to scientific and technical processes, and the actual dimensions of the components, for example, the gates of transistors inside the desktop processor, have ended up being compared from each other recently. And, most quickly, there is no such component inside the Intel processor, the size of which is positively 10 nm.
This lack of unconditional bondage between the size of the ingredient and the image of the site stops more problematic, sometimes the skill touches the comparison of scientific and technical processes of rival manufacturers. But more on this later. And now we are interested in Intel's 10nm process and its disadvantages. It was originally expected that it will be implemented back in 2015. Immediately now, the mother of my children is 2020, however, the set of provisions with 10nm chips is small. You will not be able to purchase desktop PCs or server CPUs built for the above technical process. Exceptionally mobile processors for the sake of laptops and tablets switched to 10nm technology, of course, only those with low and ultra-low power consumption. Others have upgraded to 14 nm.
These facts need to be analyzed taking into account the generally accepted stereotype of Intel itself - Moore's Law, and therefore it is necessary to consider the opposition to the laws of physics, which the creators of microcircuits have encountered in recent years. However, even huge adversities in the production of semiconductors can arise, sometimes separate transistors will reach the size of a handful of atoms and will obey obscure photonic effects, for example, tunneling. But this is a completely different story.
Probably all Intel's problems are combined with excessive ambition, moral obsolescence of separate production technology, perhaps complacency and lack of investment.
Intel's general manager Bob Swan said Intel’s difficulties with 10nm technology are “kind of a conclusion to what we have done in the past. Then we tried to conquer Victoria no matter what. And when extremely weighty times approached, we set even more fundamental goals. That is why it took us a long time to acquire them ”.
ncreased expectations from microcircuits
On the 10-nm scientific and technical section, this fundamental setting means the deepening of the density of transistors by 2.7 times. Rather, per unit area of the crystal in a 10-nm node, there are 2.7 times as many transistors as in a 14-nm node. More specifically, the processors made after the 14-nm process technology hold 37.5 million transistors for a square millimeter, while as if in some square millimeter of 10-nm crystals 100 million are held. The main deepening of the density of transistors implements 10-nm technology significantly more fundamental in comparison with other technical processes.
Deepening the density by a factor of 2.5 and the passage through the 22-nm to 14-nm patterns were impressive, however, the passage from 32 nm to 22 nm led the deepening of the density by 2.1 times, and the transition from 45 nm to 32 nm - by 2, 3 times. Representation of the nature of these changes helps to chew on the differences between Intel nodes and nodes of rival vendors. For example, Intel's 10nm design assumes a saturation of 100.8 million transistors per square millimeter. This ratio is slightly higher than the TSMC ratio - 96.5 million transistors (later TSMC announced 113.9 million transistors for a square millimeter for an improved 7-nm process technology). All three 7nm sections from Samsung do not reach the 100 million mark.
The thing is that 10-nanometer development through Intel was infinitely principled - so much so that in 2017 the fraternity added a note "Hyper Scaling" to indicate the increased density. In retrospect, it is possible to argue that the expectations were too high. This is because Intel has prepared the final section for the DUV flow lithography base. In a nutshell, the size of the ingredients in a microcircuit is determined by the length of the optical wave used in lithographic processes. These processes etch the ingredients onto a given surface of the silicon wafer, and computer processors are cut from the silicon wafers.
This is not a bisector of dependence. The effect of yes can be shown by various techniques and auxiliary options, for example, masks used are valid in the property of a multiplier, which shorten the dimensions of the ingredients further than the practical optical wavelength.
DUV Chip Cooking Equipment utilizes 193 nm swell UV radiation. However, to consume the limit of the relatively density of the location of the transistors near the provided wavelength. Intel has exceeded that limit.
As a result, there is an unpleasant delay in production for almost five years. It's eternity in terms of Intel volume dynamics and Moore's Law. Already immediately consume the symptoms that the 10nm process is not classified what it is forced to be. So Ice Lake, the freshly baked tenth generation mobile processors, accelerates slower than their 14nm predecessors. Ice Lake's fastest 10nm processors, the Core i7-1065G7, generate a lot of haste around 3.9GHz, while the eighth generation Core i7-8665U is faster for a solid 900MHz. This is a hell of a lot, and therefore, for some reason, the production process goes wrong.
Another proof that the 10nm process did not live up to Intel's expectations is the duplication of weak 10th generation processors. Along with today's Ice Lake CPUs, the newly minted Comet Lake family is descending, and both belong to the 10th generation.
Similar to Ice Lake, Comet Lake mobile processors fit in low-power and ultra-low-power formats.
But in recognition through Ice Lake, Comet Lake utilizes 14nm, but not 10nm process technology, and extends to 6-core modifications around a high clock speed of 4.9 GHz.
As a result, already now you are offered the opportunity to purchase a subnotebook with a processor bearing the Intel 10th generation mark, however, what is inside the box may differ from the declared one. If the calculator is 2 or 4-nuclear, it is given the opportunity to exist weak or ultra-low-power. In addition, 10-nanometer or 14-nanometer. He is given the opportunity to build on the microarchitecture of 2015 Skylake ages or completely freshly baked - Sunny Cove, and also conform to Ice lake.
On the 10-nm scientific and technical section, this fundamental setting means the deepening of the density of transistors by 2.7 times. Rather, per unit area of the crystal in a 10-nm node, there are 2.7 times as many transistors as in a 14-nm node. More specifically, the processors made after the 14-nm process technology hold 37.5 million transistors for a square millimeter, while as if in some square millimeter of 10-nm crystals 100 million are held. The main deepening of the density of transistors implements 10-nm technology significantly more fundamental in comparison with other technical processes.
Deepening the density by a factor of 2.5 and the passage through the 22-nm to 14-nm patterns were impressive, however, the passage from 32 nm to 22 nm led the deepening of the density by 2.1 times, and the transition from 45 nm to 32 nm - by 2, 3 times. Representation of the nature of these changes helps to chew on the differences between Intel nodes and nodes of rival vendors. For example, Intel's 10nm design assumes a saturation of 100.8 million transistors per square millimeter. This ratio is slightly higher than the TSMC ratio - 96.5 million transistors (later TSMC announced 113.9 million transistors for a square millimeter for an improved 7-nm process technology). All three 7nm sections from Samsung do not reach the 100 million mark.
The thing is that 10-nanometer development through Intel was infinitely principled - so much so that in 2017 the fraternity added a note "Hyper Scaling" to indicate the increased density. In retrospect, it is possible to argue that the expectations were too high. This is because Intel has prepared the final section for the DUV flow lithography base. In a nutshell, the size of the ingredients in a microcircuit is determined by the length of the optical wave used in lithographic processes. These processes etch the ingredients onto a given surface of the silicon wafer, and computer processors are cut from the silicon wafers.
This is not a bisector of dependence. The effect of yes can be shown by various techniques and auxiliary options, for example, masks used are valid in the property of a multiplier, which shorten the dimensions of the ingredients further than the practical optical wavelength.
DUV Chip Cooking Equipment utilizes 193 nm swell UV radiation. However, to consume the limit of the relatively density of the location of the transistors near the provided wavelength. Intel has exceeded that limit.
As a result, there is an unpleasant delay in production for almost five years. It's eternity in terms of Intel volume dynamics and Moore's Law. Already immediately consume the symptoms that the 10nm process is not classified what it is forced to be. So Ice Lake, the freshly baked tenth generation mobile processors, accelerates slower than their 14nm predecessors. Ice Lake's fastest 10nm processors, the Core i7-1065G7, generate a lot of haste around 3.9GHz, while the eighth generation Core i7-8665U is faster for a solid 900MHz. This is a hell of a lot, and therefore, for some reason, the production process goes wrong.
Another proof that the 10nm process did not live up to Intel's expectations is the duplication of weak 10th generation processors. Along with today's Ice Lake CPUs, the newly minted Comet Lake family is descending, and both belong to the 10th generation.
Similar to Ice Lake, Comet Lake mobile processors fit in low-power and ultra-low-power formats.
But in recognition through Ice Lake, Comet Lake utilizes 14nm, but not 10nm process technology, and extends to 6-core modifications around a high clock speed of 4.9 GHz.
As a result, already now you are offered the opportunity to purchase a subnotebook with a processor bearing the Intel 10th generation mark, however, what is inside the box may differ from the declared one. If the calculator is 2 or 4-nuclear, it is given the opportunity to exist weak or ultra-low-power. In addition, 10-nanometer or 14-nanometer. He is given the opportunity to build on the microarchitecture of 2015 Skylake ages or completely freshly baked - Sunny Cove, and also conform to Ice lake.
Microarchitecture difficulties
The Sunny Cove listing correctly brings us to another great failure of Intel - microarchitecture. Prior to the release of 10nm Ice Lake chips for ultraportable laptops last year, the horde of processors for desktops, laptops and servers relied on the 14nm process, which debuted in 2014, and the Skylake architecture, which emerged in 2015. Both have ruled a thousand times, however there were no chunky changes in the updates.
Above all, the dark phenomenon of the Nehalem microarchitecture in 2008, Intel could only recommend 4 cores for processors in popular PC modifications. This continued until the 2017 release of the Coffee Lake microarchitecture, an evolved version of Skylake, and a further increase to six cores. For about a decade, Intel has not increased the number of cores for commercially discarded product modifications.
In a little less than two years since the noble age, Intel raised the bar to 10 cores for popular desktop processors by releasing Comet Lake, a secondary updated Skylake tweak belonging to the 14nm processor family. It turns out that during 10 flights there were no shifts, and then a depression was created 2.5 times after the lapidary time interval. Nothing to do could lead to such a shrill increase in the number of nuclei after a long stagnation? The root cause for this is the origins of AMD's Zen architecture and Ryzen processors, the first of which ended in 2017. Without going into lengthy reasoning, Intel was lazy until, happily, it did not have a competitor.
Of course, with as many as ten cores, Intel is infinitely inferior to AMD, which currently invites 16 cores in the famous Computer with 3rd generation Ryzen processors. Their advantage is encompassed yes in that they are built for a 7nm process warehouse through TSMC.
In the mobile sector, Intel's circumstances are no better. The new line of 7nm mixed processors Renoir through AMD has eight Zen 2 cores with a capacity of 15 watts. Intel has managed to develop an exclusively 6-core Comet Lake Core-i7 10810U as a competitor. It is a processor with an exclusive clock speed of 1.1 GHz. The 15-watt Ryzen 7 4800U is packed with 8 cores and clocked at 1.8 GHz. Not a positive comparison.
The Sunny Cove listing correctly brings us to another great failure of Intel - microarchitecture. Prior to the release of 10nm Ice Lake chips for ultraportable laptops last year, the horde of processors for desktops, laptops and servers relied on the 14nm process, which debuted in 2014, and the Skylake architecture, which emerged in 2015. Both have ruled a thousand times, however there were no chunky changes in the updates.
Above all, the dark phenomenon of the Nehalem microarchitecture in 2008, Intel could only recommend 4 cores for processors in popular PC modifications. This continued until the 2017 release of the Coffee Lake microarchitecture, an evolved version of Skylake, and a further increase to six cores. For about a decade, Intel has not increased the number of cores for commercially discarded product modifications.
In a little less than two years since the noble age, Intel raised the bar to 10 cores for popular desktop processors by releasing Comet Lake, a secondary updated Skylake tweak belonging to the 14nm processor family. It turns out that during 10 flights there were no shifts, and then a depression was created 2.5 times after the lapidary time interval. Nothing to do could lead to such a shrill increase in the number of nuclei after a long stagnation? The root cause for this is the origins of AMD's Zen architecture and Ryzen processors, the first of which ended in 2017. Without going into lengthy reasoning, Intel was lazy until, happily, it did not have a competitor.
Of course, with as many as ten cores, Intel is infinitely inferior to AMD, which currently invites 16 cores in the famous Computer with 3rd generation Ryzen processors. Their advantage is encompassed yes in that they are built for a 7nm process warehouse through TSMC.
In the mobile sector, Intel's circumstances are no better. The new line of 7nm mixed processors Renoir through AMD has eight Zen 2 cores with a capacity of 15 watts. Intel has managed to develop an exclusively 6-core Comet Lake Core-i7 10810U as a competitor. It is a processor with an exclusive clock speed of 1.1 GHz. The 15-watt Ryzen 7 4800U is packed with 8 cores and clocked at 1.8 GHz. Not a positive comparison.
Principle into the future
Here's a version of the charge. The years that have passed have not been technologically productive for Intel. George Davis, the company's chief commercial officer, spoke of the 10nm flop this way: “This electrotechnical site will certainly not be the best in Intel's history. It is less efficient than the 14nm process technology, and less efficient than the 22nm process technology. " But are the consequences of Intel's ongoing problems really so disastrous?
From the point of view of an economic perspective, for this question it is possible to respond positively - no. In the end, it is not so much the flowing arrangement, otherwise it is too bad, in fact, there is no difficulty at all. In 2019, Intel's earnings have delivered unprecedented numbers. Since mid-2018, the ages of its sales did not fall due to scientific and technical stagnation, and the manufacturer himself felt adversity with a satisfied view of demand for its 14-nm processors.
If you dig deeper, it is possible to come to your senses to the conclusion that along the last border the plot of difficulties is covered in the scientific and technical process. The number of cores in Intel server processors has clearly increased since the arrival of the 14nm era. Immediately Intel invites the whole 28 cores in some kind of processor die. that, the larger the cores in the same process, the fewer processors can be pulled out of one semiconductor wafer, which, in turn, can lead to supply constraints.
But, whatever one may say, Intel does not check any economic difficulties, but even this event seems to be the main reason, because the culprit might strike a grand protest against competitors in the design of products and technologies.
And this special effect is now visible. Ice Lake processors evolved into a freshly baked microarchitecture popularly known as Sunny Cove. It improves clock performance by 18% after comparing it to Coffee Lake, a refinement of the Skylake microarchitecture.
But even this is only a beginning. The decisive word in the renaissance of Intel's microarchitecture was the introduction of Jim Keller to the team, who led the category for microprocessor development.
Despite the fact that he is going to leave this rectifier after six months, it is impossible to underestimate the contribution, some of it can be transferred to the development of the company. Keller is one of the most respected, if not the most respected, microprocessor designer.
Here's a version of the charge. The years that have passed have not been technologically productive for Intel. George Davis, the company's chief commercial officer, spoke of the 10nm flop this way: “This electrotechnical site will certainly not be the best in Intel's history. It is less efficient than the 14nm process technology, and less efficient than the 22nm process technology. " But are the consequences of Intel's ongoing problems really so disastrous?
From the point of view of an economic perspective, for this question it is possible to respond positively - no. In the end, it is not so much the flowing arrangement, otherwise it is too bad, in fact, there is no difficulty at all. In 2019, Intel's earnings have delivered unprecedented numbers. Since mid-2018, the ages of its sales did not fall due to scientific and technical stagnation, and the manufacturer himself felt adversity with a satisfied view of demand for its 14-nm processors.
If you dig deeper, it is possible to come to your senses to the conclusion that along the last border the plot of difficulties is covered in the scientific and technical process. The number of cores in Intel server processors has clearly increased since the arrival of the 14nm era. Immediately Intel invites the whole 28 cores in some kind of processor die. that, the larger the cores in the same process, the fewer processors can be pulled out of one semiconductor wafer, which, in turn, can lead to supply constraints.
But, whatever one may say, Intel does not check any economic difficulties, but even this event seems to be the main reason, because the culprit might strike a grand protest against competitors in the design of products and technologies.
And this special effect is now visible. Ice Lake processors evolved into a freshly baked microarchitecture popularly known as Sunny Cove. It improves clock performance by 18% after comparing it to Coffee Lake, a refinement of the Skylake microarchitecture.
But even this is only a beginning. The decisive word in the renaissance of Intel's microarchitecture was the introduction of Jim Keller to the team, who led the category for microprocessor development.
Despite the fact that he is going to leave this rectifier after six months, it is impossible to underestimate the contribution, some of it can be transferred to the development of the company. Keller is one of the most respected, if not the most respected, microprocessor designer.
It became famous due to the development of the microarchitecture of the K8 processor, codenamed Athlon 64 and the main chip from AMD, which became a noble competitor to Intel. Later, Keller worked at Apple, creating photo designs for a series of in-house processors for the ARM warehouse, which then took the lead in the market for phones and tablets. In 2012, Keller returned to AMD, leading the development of the Zen microarchitecture and once again supplying AMD with devices to fight Intel. Following a laconic presence for the apex position of Tesla's electric vehicle builder, Keller took over as senior vice president of Intel in early April 2018.
Given the lag between the design and microarchitecture of processors and the launch of product marketing, it is highly unlikely that the new Sunny Cove cores inside Ice Lake processors are Keller's plow. This will undoubtedly be for the further post-Sunny Cove architecture of Willow Cove. It is planned to be released at the end of this age for the sake of a family of 14-nm backported processors, that is, using the "reverse transfer" of the freshly baked microarchitecture for the "old" technical process, Rocket Lake processors.
The Golden Cove microarchitecture will take an even bigger step forward and lay the groundwork for slated Alder Lake processors later in the year. But even Golden Cove has no way of conforming to a good Keller piece. To achieve the desired result, we need to wait for the Ocean Cove phenomenon, which will be exhausted in 2022 or 2023, however, Keller's quick withdrawal will mean that his influence on this calculation will probably be limited to some extent.
There is happily no overhead about Ocean Cove. Not long ago there were rumors that the productivity of the provided microarchitecture would be 80% higher than that of Skylake. Even though these are purely rumors, we know for sure that Keller has an ingenious formulary price list and that Intel has a monumental, cunning plan that is much larger than it did years ago. As Keller said: "We expect to multiply the number of transistors by 50 at a time and do our best to get the extreme out of each stack."
At the same time, the 7nm CPUs that follow the problematic 10nm processors will not face the same limitations as their predecessors. For the production of 7-nm processors, photolithography of the extreme ultraviolet spectrum (EUV) with a swell length of up to 13.5 nm will be used. Or, the 7nm process has completely changed. It will be seen, however immediately it is definitely possible to say that Intel's monitoring is extremely optimistic.
Intel plans to force the passage from 7nm to 5nm and beyond. From this we can conclude that the manufacturer will temperamentally improve the freshly baked technology in the antagonism of today's costly, even if this requires investments in research and development work. Above all, with the involvement of EUV lithography, Intel expects to return to the past rates of production - once every 2 years, activating the ages from the 7-nm process technology in 2021 and reaching the release of the 1.4-nm circuit in 2029. “I think EUV will help us get back to the pace with which transistors are increasing unanimously according to Moore's Law,” Davis said.
All this collectively assigned gives the impression that Intel is returning samples of the most advanced architectures and fastest processors. Whether this will happen is another question. Immediately, AMD is probably in a better position than Intel, despite the fact that the latter makes a lot of efforts. AMD's clever design of microarchitectures, including Zen 3 and Zen 4, together with TSMC's technology solutions, will help increase competition between the two manufacturers. However, we will not predict the defeat of Intel.
In the end, the final time, sometimes NetBurst and Pentium 4 ended their days and Intel's circumstances stalled, leading the Core family and primacy for the processor market during the decade.
Given the lag between the design and microarchitecture of processors and the launch of product marketing, it is highly unlikely that the new Sunny Cove cores inside Ice Lake processors are Keller's plow. This will undoubtedly be for the further post-Sunny Cove architecture of Willow Cove. It is planned to be released at the end of this age for the sake of a family of 14-nm backported processors, that is, using the "reverse transfer" of the freshly baked microarchitecture for the "old" technical process, Rocket Lake processors.
The Golden Cove microarchitecture will take an even bigger step forward and lay the groundwork for slated Alder Lake processors later in the year. But even Golden Cove has no way of conforming to a good Keller piece. To achieve the desired result, we need to wait for the Ocean Cove phenomenon, which will be exhausted in 2022 or 2023, however, Keller's quick withdrawal will mean that his influence on this calculation will probably be limited to some extent.
There is happily no overhead about Ocean Cove. Not long ago there were rumors that the productivity of the provided microarchitecture would be 80% higher than that of Skylake. Even though these are purely rumors, we know for sure that Keller has an ingenious formulary price list and that Intel has a monumental, cunning plan that is much larger than it did years ago. As Keller said: "We expect to multiply the number of transistors by 50 at a time and do our best to get the extreme out of each stack."
At the same time, the 7nm CPUs that follow the problematic 10nm processors will not face the same limitations as their predecessors. For the production of 7-nm processors, photolithography of the extreme ultraviolet spectrum (EUV) with a swell length of up to 13.5 nm will be used. Or, the 7nm process has completely changed. It will be seen, however immediately it is definitely possible to say that Intel's monitoring is extremely optimistic.
Intel plans to force the passage from 7nm to 5nm and beyond. From this we can conclude that the manufacturer will temperamentally improve the freshly baked technology in the antagonism of today's costly, even if this requires investments in research and development work. Above all, with the involvement of EUV lithography, Intel expects to return to the past rates of production - once every 2 years, activating the ages from the 7-nm process technology in 2021 and reaching the release of the 1.4-nm circuit in 2029. “I think EUV will help us get back to the pace with which transistors are increasing unanimously according to Moore's Law,” Davis said.
All this collectively assigned gives the impression that Intel is returning samples of the most advanced architectures and fastest processors. Whether this will happen is another question. Immediately, AMD is probably in a better position than Intel, despite the fact that the latter makes a lot of efforts. AMD's clever design of microarchitectures, including Zen 3 and Zen 4, together with TSMC's technology solutions, will help increase competition between the two manufacturers. However, we will not predict the defeat of Intel.
In the end, the final time, sometimes NetBurst and Pentium 4 ended their days and Intel's circumstances stalled, leading the Core family and primacy for the processor market during the decade.