In This Article
What is 7 nm Processor?
The 7 nm processor refers to the lithography process followed for making the chip. Technically, it is the manufacturing or fabrication process of the semiconductor. In simple terms, it signifies the gap between two transistors.
- The 7 nm processors are considered to be one of the latest and most advanced processors as of now.
- These processors offer higher performance, more density and consume less power and are used in several industries and for different purposes including high to mid-end mobile devices.
- The higher cache on chip and drive current enhances the processing power of these processors while lowering the operational costs for the end users.
- The small size of these processors causes problems in planning, designing, placement, routing, and Quantum Tunneling to name a few.
Understanding 7 nm Processor
Typically, the term 7 nm implies the technology node. It is typically based on the Fin Field -Effect Transistor or FinFET technology which is basically a specific type of multi-gate MOSFET technology.
It is considered to be one of the latest and most advanced FinFET process nodes that is used in processor design and fabrication as of now.
The term 7 nm however does not signify any geometry or any particular dimension of the processor, especially not the size.
It is simply the term used to identify the generation as well as the technology used in it.
To put it in simple words, it is just a commercial name given to the transistor and is used for marketing purposes.
However, after 2009 the term ‘node’ is also used as a commercial name and has become more popular.
In short, the term 7 nm does not have any direct relation with the size of the processor, the gate length, gate pitch or the metal pitch.
The shrunk transistors in a 7 nm processor offer improvement in utilization of the silicon area as well as power efficiency.
However, the tradeoff is more complexity in processor design and fabrication process as well as in its cost.
The producers of the 7 nm processors follow different design rule management systems for volume production.
As said earlier, there aren’t many of them but the notable differences in design and measurements are worth mentioning.
For example, the 7 nm metal patterning followed by the TSMC as of now uses SADP or the Self Aligned Double Patterning lines.
This involves cuts interleaved within the cells on a distinct mask according to the requirement to reduce the height of the cells.
For a 7 nm processor the most crucial factor for performance is the fin. To form the fin, TSMC follows SAQP or Self Aligned Quad Patterning process.
Also, the design rule checks followed by them ensure that multi-patterning is avoided and enough clearances for the cuts are provided for the only one cut mask that is required.
Then the 7 nm chip of the GlobalFoundries, also referred to as 7LP pr Leading Performance, follows a different design rule management process that offers nearly 40% higher performance with more than 60% lower power consumption.
Add to that, it also offers double scaling in terms of density and a reduced cost by as much as 30% to 45% for each die as compared to its older 14 nm process.
In their processor variants, the CPP or the Contacted Poly Pitch and the MMP or the Minimum Metal Pitch are known to be 56 nm and 40 nm, respectively. And, it is produced with SADP.
The manufacturing company also had plans to use EUV 13.5 nm lithography to make their more advanced 7LP+ processors but later on decided against it and also stopped all their process development plans and practices for 7 nm processors and beyond.
And, as for Intel, their new Intel 7 is designed on the basis of their preceding 10 nm node and was known previously as 10 ESF or 10 nm Enhanced SuperFin.
You can expect to have an increase of anywhere between 10% and 15% in terms of performance per watts by using these processors.
In the meantime, the older 7 nm process which is now referred to as Intel 4 is supposed to be released sometime in 2023.
Though only a few specific details about this particular processor have been made public by Intel, the transistor density is quite noteworthy.
It is estimated to be 202 million transistors per square millimeter at the minimum.
Also the naming of process nodes is different of different major manufacturers along with other specs of their respective processors that include:
- SRAM bit cell size
- EUV implementations
- EUV-limited wafer output
- Transistor density
- Transistor fin pitch
- Transistor fin height
- Transistor gate pitch
- Minimum metal pitch and
However, multiple patterning in the 7 nm processors needs special mention because it plays a major role in the field of yield and cost since the use of EUV lithography in 7 nm process is still very limited.
It is the multiple patterning processes that still determine the resolution for most of the crucial layers.
And, with the more use of EUV, as and when it happens, there will be some additional considerations.
Another vital reason for using multiple patterning even now for most of the layers is that the immersion tools are much faster now.
For that matter, the layer completion output by using EUV is not as desired as it should be on the layers that specifically need immersion quad-patterning.
Therefore, multiple patterning is an important part of the 7 nm technology that is good to use on all layers in order to complete the layers more productively and to produce a 7 nm processor that will be more powerful and effective to put to different uses.
Processors with smaller form factors such as the 7 nm processors are extensively used in several industries and purposes such as:
- Artificial Intelligence
- 5G or 5th Generation mobile network and infrastructure
- Internet of Things or IoT devices
- Hyper scan data centers
- Consumer applications
- Graphics Processing Unit
- Automotive and
- Networking, information, and communication.
It is the design benefits and enhanced performance of the 7 nm processors that allows it to be used in almost all high-to-mid end mobile devices and other electronic devices that are designed for performing highly complex computing in quick time.
Smaller the form factor of the processor, like the 7 nm processor, more benefits are offered. Some of the benefits are:
- Enhanced and faster performance
- Higher density of transistors
- Higher drive current per unit area due to taller fins and larger effective volume of the channel
- More space for making architectural improvements
- More cores for each CPU
- Less need for higher drive and wider standard cells
- Higher cache on chip
- Lower operational expenses for the end users
- Higher power carrying capability
- Higher frequency for low power
- Capacity to perform at a low supply voltage
- Enabling switching much faster due to higher electron mobility
- Needs shorter wavelengths of light to create the circuit with smaller features
- Less power consumption and therefore
- Less heat generation and less thermal throttling needs.
Overall, the 7 nm processor technology helps various merchants involved in the business such as lithography equipment suppliers, foundry and design players, device OEMs, and semiconductor material/wafer and metrology equipment suppliers.
All these benefits offered by the 7 nm processors enhance the performance of the processor on the whole but indirectly.
However, the architecture of the processor needs to utilize the voltage curve efficiency as well as the extra transistor budget properly for an improved process.
This will transform into better and higher computer performance at a much lower power requirement.
This will, in turn, increase the battery life especially in the smaller form-factor mobile devices such as laptop computers, tablets, and ACPCs.
The 7 nm FinFET technology of TSMC is however proven to allow a more inclusive design ecosystem.
This guarantees a seamless migration path as well as a much faster design cycle time.
This new technology needs very little engineering resources to pull off the product benefits offered by it.
Add to that, it also significantly reduces the time to market a product.
With the processor manufacturing technology developing and the size of the chips shrinking, the chips with lower nanometers come with reduced channel length between the source and the drain.
This increases the complexities in designing the chips more than ever.
The challenges faced by the manufacturers of the processors while creating the physical design of the chips typically depends and varies on type of the design.
For example, the mobile chips come with a smaller form factor than the server chips and therefore the focus is more on the constrained area rather than on the performance and reliability.
However, in the case of the server chips, the priority of the processor manufacturers is more on the performance and reliability rather than on area efficiency.
The smaller the form factor of the processors, the higher and more are the challenges faced by the chipmakers in its designing and production, and so is the case with the 7 nm processors.
Some of the most commonly faced problems by the designers or manufacturers of the 7 nm processors while designing a smaller nanometer chip are its setup, timing and hold violations issues.
Another notable common challenge is due to the elementary defect of the physical design with the silicon transistors of the 7 nm processors is that these are very close to each other.
This results in a process known as Quantum Tunneling which means that the transistors unfortunately cannot be reliably turned off and will stay on for the most part of it.
And, some of the most significant challenges posed by the 7 nm fabrication process are:
- Floor Planning – This is due to the higher number of macro counts as well as the different variations in the macro sizes. It is also due to the clock channel and routing blockages.
- Power Planning – It is due to IR drop issues that are the results of switching and congestion of cells along with the need for a high amount of power within the same area at the same time.
- Placement – This is due to the congestion once again as well as the timing violations.
- Clock Tree Synthesis – This is the challenge that involves building clock trees and meeting skew targets and latency as well.
- Routing – This involves creating the physical connections with the minimum use of wire length and less DRC/ANT but ensuring that a good timing QoR is maintained.
- Signoff – This involves a large variety of checks to be performed which include and are not limited to antenna, DRC or Design Rule Check, LVS or Layout Versus Schematic, LEC or Local Exchange Carrier, STA or Spanning Tree Algorithm, IR or Infrared wireless mobile, EM or electromagnetic sound wave, and more in order to verify that the layout of the processor drawn works just as it is intended.
Another significant challenge faced by the processor manufacturers is related to photolithography.
The light beams using laser during the fabrication process causes a fringe effect which takes a heavy toll and results in a third order effect and also amplifies the second order effects in the transistors.
This means that all of the algorithms in the transistor physics as well as in the backend both change drastically.
The 13.5 nm Extreme Ultraviolet or EUV lithography, on which the manufacturers were counting on originally, cannot be used for this node.
This is because, as of now, the EUV technology is not ready for smaller processors such as 7 nm, still being in its early stages.
This issue prompts the processor designers to resort to the traditional 193 nm immersion with multiple patterning, which involves pitch walking, edge placement error, and high cost and is a major challenge by itself.
This forces the manufacturers to push the lithography rules to the extreme limits.
In addition to the above, the manufacturers also face a significant challenge regarding spotting yield and the reliability detractors which are seriously affected due to the hard line scaling scenarios.
These scenarios increase the vulnerability of the systematic and parametric detractors which, in turn, magnify the reliability risks.
The interconnect RC delay also limits performance scaling.
Also, using VarScan Solution for identifying contact and PPM level changeability from 1 kΩ to 10Ω range is another major challenge in 7 nm processor designing.
In order to identify all these challenges by the customary optical SEM or Scanning Electron Microscope and E-beam systems accurately and quickly new methodologies are required to be used by the foundries in the 7 nm processor manufacturing technology and below.
The high cost of manufacturing the 7 nm processors is another issue.
This is because the cost of the fabrication equipment needed for the 7 nm logic process will most likely be about $160 million, and this is just for making only a thousand wafers per month, depending on the installed capacity.
Add to that, there are also other variable costs of manufacturing such as raw silicon wafers, utilities, and labor.
Apart from that, the longer cycle times are also quite a challenge that the manufacturers need to overcome.
Typically, a 7 nm logic process can take in excess of 80 photo passes while using multiple patterning and 193 nm photolithography.
This is in comparison to the 50 photo passes of a 28 nm logic process.
Mask making is very difficult at each node especially by using the 193 nm wavelength lithography due to its physical limit of 40 nm half-pitch.
The photo mask makers need to eliminate the diffraction issues at the advanced nodes.
This can be done by using RETs or Reticle Enhancement Techniques on the mask where one RET is called Optical Proximity Correction or OPC.
This improves the printability on the wafers by modifying the mask patterns.
However, the assist features of OPC are becoming more complex and smaller with each advanced node which poses another challenge to the processor manufacturers.
Also, the number of masks on each mask set is also increasing at every node, adding to the complexities.
Also, multiple patterning adds to the number of total masks required to produce a given chip.
This not only increases the complexity but also adds to the time taken to manufacture a chip because every feature on it must be written more accurately.
The masks are also required to have perfect shapes with better geometries and spacing which needs a wider process window.
More insistent OPC such as Inverse Lithography Techniques or ILTs are required for it.
This, once again adds to the difficulty level and time taken to pattern or write the mask in manufacturing a 7 nm chip using the traditional e-beam mask writers.
The longer turnaround time, in turn, equates to higher costs of manufacturing.
Even if the EUV lithography is used at 7 nm, mask makers will need to deal with the additional complexities of the EUV masks such as even longer write times.
Then there are other issues as well such as improving channel mobility and reducing contact resistance as well as parasitic capacitance in order to ensure better performance.
The fins are also needed to be re-engineered in order to extend the FinFET.
The fins can be made taller so that they supply more drive current.
Therefore, the chips will be much faster and consume less power at the same time.
However, this will also increase the capacitance of the device.
Then, consider the Backend-Of-The-Line or BEOL where the interconnects within the device are formed.
These are minute small wiring schemes. These are more compact in a 7 nm silicon chip which results in unnecessary Resistance-Capacitance or RC delays in the chips.
Typically, copper resistivity can increase average delay by a significant percentage.
In addition to that, the other challenges faced by the chipmakers in the BEOL include:
- The conventional metallization scheme which is commonly referred to as the Copper Dual Damascene process
- Insufficient scaling of the barrier/liner thickness which results in added resistance and reduced volume for copper fill and
- An increase in metal resistivity though in smaller amounts due to surface and/or grain boundary scattering.
Dealing with the issues of low-k dielectrics is another problem that the processor manufacturers should take into account.
This is mainly caused due to the materials with poor mechanical properties.
That is not all. Another most significant issue with manufacturing lower nanometer processors such as 7 nm is with the process control.
The manufacturers need to create structures consisting of a finite number of atoms to keep it precise, thin and conformal.
Variation or variability control will ensure better scaling in terms of power, performance, and area scaling and therefore it must be minimized.
This needs atomic layer control since every Angstrom and nanometer matters.
This will ensure uniformity in all respects including extreme edge and line edge roughness of the wafer, surface stoichiometry, and intra-die uniformity irrespective of the feature sizes.
Also difficult and critical is the metrology and wafer inspection part of process control to detect the defects at the node and it is also becoming very expensive.
The structures in three dimensions are even more difficult to measure.
This is because there is no single tool that can effectively and accurately handle all metrology requirements for FinFET.
A typical CD-SEM or a Critical Dimension Scanning Electron Microscope will not be enough and therefore it is needed to move to scatterometry.
The processor manufacturers will need to implement much better, more complex, and probably more costly hybrid solutions for it by using other techniques such as AFM, OCD and X-ray in metrology for the advanced nodes.
When the data obtained from each are combined it will help in making a more precise estimate to design the best 7 nm processors.
Therefore, In order to deal with all these challenges effectively and successfully to design the best processor with a small form factor that will offer the users with significant benefits as mentioned above, it calls for the services of dedicated SMEs or Subject Matter Experts.
So, reaching to the end of this article now you surely know what a 7 nm processor is and how useful it is these days.
In spite of its design challenges, this vital processor manufacturing technology helps the smaller mobile devices of today to function smoothly and effectively.