When we are still entangled, the newly bought mobile phone or computer is not the latest 7nm or 10 nm chip, TSMC announced the first month to complete the 5nm architectural design, and have entered the trial production stage. The semiconductor process achieves such a fast iteration in such a short period of time, which is completely exceeded by many people. Especially for DIY players who have painful AMD 7nm and Intle 10nm new products, this news is like a general existence of sunny day.
Don't you say that many people haven't used 7nm and 10nm, have you been in double? This is not only a fast in science and technology, but also feeling some uncomfortable in his heart ...
Test 3NM is ready to go
In fact, from the perspective of users, we should support the acceleration iteration of products, as this allows us to get more excellent performance with the same money. However, if the frequency of the product update is too fast, it seems that "there is no way" reverse effect. Because no one wants the latest products just bought today, I haven't used it for a few days! "
The refined TSMC has been beneficial to seize the 7nm process by taking the lead in preempting the 7nm process, and the large amount of apple, high-pass, AMD, NVIDIA and other international giants. Even the AMD that has been suppressed by Intel, and has received more attention and expectation by players with 7nm of Taiwan.
This also allows TTEM airy and decided to attack 5nm high in 2020!
TSMC process plan
According to the official data, the new 5nm chip based on Cortex A72 core can provide 1.8 times of logical density, and the performance can be increased by 15%, and the chip area is reduced by 15%. According to the development plan of TSM, the future 3nm and 2nm are already listed ...
Did we use the NM (Nano) unit for many years to fail?
In fact, it is not absolutely, according to the official disclosure of TSM: The 5nm process market is to help customers to support the 5nm system single-chip design of the next-generation high-efficiency operation application products, and the target locks 5G and artificial intelligence (AI) markets with broad development prospects.
This can't help but think of some things about the 7nm chip last year. Who can think of the world's first mass production 7nm chip, it is a mineral enterprise from China - Jia Nan. The technical level and shipments of this company cannot naturally not follow Apple, Qualcomm, but grab it before they have achieved the first.
Mining machine enterprises grab the world's 7nm chip
The reason is that the structure of the mine chip is extremely simple, so it is extremely easy to adapt to the latest production process from the perspective of design and production. For this general computing processor for computers and mobile phones, because of many structural complexities, the transistor density is extremely high, so the impact on design and production, and the improvement rate will have a more demanding requirements.
If the pulse machine chip is like a basketball court, then the computer processor is equivalent to build a aircraft carrier ...
TSMC puts the 5nm process technology as the layout of the future 5G and the AI market, and there is also a wonderful work.
For example, the most important baseband chip in the 5G field, a smaller volume, low fever, and more than 5nm baseband, which will undoubtedly bring direct improvement to the experience of mobile terminal. In the artificial intelligence area, the design of the edge computing chip is similar to the mine machine chip, and the structure is simple and functional, and it is more likely to be deployed to more intelligent terminals.
Google edge calculation TPU is very small
So only from the current trend, the 5nm of TSME is actually produced in 2020, in fact, the traditional PC market, as well as the mobile phone processor market in the short term. Today, the 7nm and 10nm processes of the best will continue to take over the years, at least R & D costs will limit these processor vendors to new processes.
This seems to also indicate that the traditional chip field and future emerging 5G + AI areas will form a clear watershed in the 7nm era. Previously, the global semiconductor process was always treated with process steps, and in the future, the structure is relatively simple, the volume is more compact, and the special chip of the process is more advanced will take another way.
Therefore, experienceism that is simple from digital size to distinguish is more advanced will no longer apply.
Intel's first 10nm processor i3-8121u
Even in now, it is not advisable to the conclusions of "the smaller the number, the more advanced". According to foreign media for Intel's 10nm processor Core i3-8121u, the evaluation of the 10nm processor Core i3-8121u, although "10" relative "7" seems to be behind, the 10nm technology of Intel is more advanced.
The processor is something, in fact, it is to plug in more transistors in a limited chip area, and ensure that these elements can operate efficiently. The Intel 10nm process used the third generation of FinFET stereo transistor technology, the transistor density reached 1.008 billion per square millimeter (in line with the official claim), is currently 2.7 times the current 14NM!
As a comparison, Samsung 10nm process transistor density is only 55.1 million per square millimeter, only more than half of Intel, 7 nm is 1.012.3 billion per square millimeter, barely high. 10nm. As for the retrieval electrical transistor density than Samsung. This is also the root cause of many people to spit "7NM shrinkage".
Therefore, if only the transistor density this key data, the 10nm technology of Intel seems to be "backward", but in fact, it is actually not behind, but even leads.
Intel's 10nm is still looking forward to
And return to our users to see this problem, do we really need blind recovery?
The answer is inevitably negative. We must clearly understand a truth, technology is the endless development, and our needs are threshold, then how do you understand this sentence?
For a simple example, when the processor still stays in a relatively backward process, the power consumption and heat generation are so huge. Sally, the popularity of the professional heat sink directly, for example, the Kyushu Fengshen, overclocking three, high-end, such as owl, etc. Because there is no these huge radiator, I can't press it!
At that time, the progress of each process process would actually bring the improvement of "solidity", because the processor base power consumption is too big at that time. With the continuous iteration of the process technology, the base of the increase is small. The percentage of increased increase in the lifting amplitude is also shrinking.
Therefore, people gradually improve the process technology, gradually become so sensitive.
Mobile phone hot hand has already been returned
If you can't understand, then give a simpler example. A few years ago, Android phone was hot, but now there will be such a tragic experience now. So before you change your phone very urgent, now it is basically enough, there is no special slot, so you will not be so urgent.
Be
New product:
递 科技 GROVE Kit: No hardware driver or embedded development, designed for Microsoft Azure service
Wei Shi VCNL36687S proximity sensor: small size package, integrated signal processing IC and 12-bit ADC
Yadeno Semiconductor Eval-Adxl362 Evaluation Board: Compatible with Arduino Interface, bring your own LCD screen
Infineon IM69D120 and IM69D130 Xensiv MEMS microphone, designed for low distortion and high signal-to-noise ratio
Be
Recommended activities:
Awards Q & A | How much do you know about TOF flight time?
Come to register Mouser, prize rain, everyone has a copy
Mouser helps graduation season, subscribed for 1,75 yuan can apply for 100 yuan subsidy
[Group purchase] This may be the full network price of the highest raspberry purchase method
Be
Invite you to join the panel engineer group (group number 135513647)
More activities, consultation, technical exchange
Our other product:
