Category: DATA

Photo by Taylor Vick on Unsplash

THE BUILDING BLOCKS OF SEMICONDUCTOR CHIPS FOR DATA CENTERS

The connected world is leading to real-time information exchange, and this is why consumers to enterprises expect the request to be processed in the fastest time possible. The devices used to send such requests can only process and store a certain amount of data. Anything beyond a threshold requires the use of data centers, and that also means transferring/receiving data over the air.

The computing industry has relied on data centers since mainframe days. However, the importance of data centers has mainly grown due to the connected systems. These data centers have to run 24×7 and also have to cater to numerous requests simultaneously.

To ensure a quick and real-time response from data centers, three major systems have to work in synchronization:

Software: If a smartphone user is sending the request, then the data needs to be encrypted in packets before sending it over the air to the remote location where the massive data centers are located. This means the software solutions, both on the client and the server-side, have to work in harmony. This is why software is the first major system required for accurate data center operation.

Connection: The second major system is the network of wired and wireless systems that aid the transmission of data from the client to the server (data centers). If a robust connection is not available then data centers will be of no use.

Hardware: The third and most critical piece is the silicon chip or hardware that makes up the data center. These tiny semiconductor chips end up catering to all the request that comes from different part of the world. To ensure the request is fulfilled in real-time, a smart silicon chip is also required that can handle the data-efficient without adding bottlenecks.

The growing internet user base along with data-driven computing solutions has to lead to the high demand for data centers. To cater to all such growing services, different types of data centers are required. Some data centers are small in size (less number of servers) and some are giant. Data centers with more than 5000 servers are also called hyperscale data centers. In 2020, there were more than 500 hyperscale data centers running 24×7 and were catering to request coming from any part of the world.

Data Centers Require Different Types Of Semiconductor Chips.

To run these hyperscale data centers requires large facilities, but the key piece is still the tiny semiconductor chips that have to run all the time to handle different types of requests. Due to the growing focus on data centers, there is a need to change the way new semiconductor chips are being designed for data center usage.

This is why all the semiconductor chip solutions that end up getting used in the data centers should be built around the following blocks:

Processing: Semiconductor chips for data centers should be designed to process not only a large amount of data but also a new type of data. This requires designing semiconductor chips that can cater to the request in the shortest time possible, while also ensuring there are no errors during the processing.

Security: Data centers receive different types of data processing requests and this data can have any information from credit card processing requests to personal information to login credentials. Semiconductor products by default have to focus on the security aspect when designing silicon solutions for data center usage.

Workloads: Rapid software development has lead to different types of data. Data eventually lead to the formation of workloads that the computing system has to process. Given the rise of AI/ML/DL, there is a need to process the data elegantly. This requires doing away with traditional processing blocks and instead of adapting to more workload-centric architecture that can enable a high level of parallelism to train and infer information out of it.

Adaptive: Smart world not only requires data capturing but also demands adaptive decisions. This often requires on-the-go training and modeling to ensure the user request is fulfilled intuitively. This is why there is a demand to drive AI drive architectures that can train data efficiently (eFPGA or NPU) and ensure any new (and never seen) request is handled without errors.

Storage: Memory is one of the major building blocks of the computing world. The surge in the use of cloud storage is leading to new innovative storage systems that can provide more storage per dollar. This requires driving new semiconductor solutions so that data centers become powerful but at the same time are compact enough to not consume a large amount of energy.

Efficiency: Data centers are considered to be one of the most power-hungry systems in the world. The year-on-year growth in hyperscale data centers is only going to increase the power consumption. To balance the processing need with the power consumption, semiconductor solutions have to consider the energy consumption per user request. By building an efficient semiconductor chip, the data centers can expand in number without impact the total power consumption.

The above building blocks are not specific to XPUs for data centers only. These blocks are valid for another type of semiconductor chip that data centers require. This can be a networking solution (PCIe) or a new data transfer (HBM) interface. Eventually, the above points discussed are the major reasons why data centers require different types of semiconductor chips.

THE BOTTLENECKS TO DRIVE NEXT-GEN SEMICONDUCTOR CHIPS FOR DATA CENTERS

Designing and manufacturing semiconductor chips for any type of solution requires a thorough understanding of different issues and opportunities. The same strategy is applicable when coming up with semiconductor solutions for data centers. For data centers, the complexity is much more than consumer systems mainly due to the need to provide bottleneck-free semiconductor products that run all the time.

Over the past few years, new data center-focused semiconductor companies have emerged and they have been providing different solutions. Fungible’s Data Processing Unit and Ampere’s ARM-powered XPU are a couple of such examples. However, in the end, the goal of all the semiconductor solutions is to focus on a set of features that ensures all the request is catered by the data centers in the real-time without adding any bottlenecks.

When it comes to bottlenecks in the world of computing, the list is endless. These bottlenecks can originate either from software or hardware. Eventually, the software features have to be mapped onto the hardware so that both software and hardware can work in synchronization to drive next-gen solutions.

The next-gen semiconductor chips need to focus on a few criteria’s to drive bottleneck free semiconductor powered data centers:

Features: Traditional semiconductor chips for data centers were (and still are) purely focused on performance. As the world is increasingly adopting connected systems, there is growing demand to balance performance with efficiency. This requires a new set of features that can ensure that tomorrow’s data centers are more efficient than today’s. These features can range from using advanced transistor-level techniques to new packaging solutions.

Data: The amount of data that hyperscale data centers have to crunch will keep increasing every year. The storage aspect of it will also grow along with it. This growth is leading to huge cooling systems and thus adds to total energy requirements. This challenge of managing data while lowering the impact on power consumption is pushing new solutions. More modular approaches are needed to drive next-gen semiconductor solutions.

Parallelism: Any given chip of any type in a data center can receive any amount of request. To ensure there are no bottlenecks, intelligent parallelism techniques are required. Some of the parallelism techniques require software support but many often do require hardware features (cache, data pipeline, etc.) that can support parallelism. Networking and XPUs solutions often have to consider this problem while designing chips for data centers.

Speed: While there is a growing concern of power consumption (by data centers) due to performance requirements, there is also demand to drive faster response out of data centers. This requires designing semiconductor chips for faster processing. Balancing the power, performance, and area aspect for data centers is becoming more difficult than ever. This is leading to more modular data centers but it is still going to demand semiconductor chips that can provide a high-speed solution without adding to the power requirement.

Network: Data centers have to communicate with different systems that are located in remote areas, and such communication requires heavy usage of networking solutions. To drive communication efficiently, robust networking chips are required that can handle the data without any errors. This demands designing and manufacturing semiconductor solutions with reliability and error correction. In the long run, network chips are going to play a vital role and require the bottleneck-free design to drive new data centers.

Architecture: Intel is considered the leader in XPU solutions for data centers. To design XPUs, Intel has been relying on its homegrown x86 architecture. In the last decade, the emerging workloads have changed a lot and that requires new XPU solutions. To provide newer solutions, emerging companies are focusing more on ARM and RISC-V to power their solutions. The major driving factor to use ARM or RISC-V is the ability to adapt and change the architecture to suit future requirements. Picking up the architecture is vital to avoid any kind of bottlenecks in the XPUs for next-gen data centers.

In the last two years, the world has moved towards data center solutions mainly due to the remote feature required by different services. The growth in the number of smartphone and smart device users is also driving the need for new and efficient hyperscale data centers. To cater to the future demand of green hyperscale data centers, the existing and emerging semiconductor companies will keep coming up with a newer solution.

In the long run, newer data-centric semiconductor solutions are only going to benefit the computing industry and the race to wind data centers has just begun.

Photo by Jorge Salvador on Unsplash

THE COSTLY LIFE CYCLE OF SEMICONDUCTOR DATA

The importance of data in different industries has only grown over the last few decades. This has in turn given rise to different new techniques and tools that enable enterprises to capture and analyze data on the go. It will not be wrong to say that today, the most prized commodity in the market is data.

The same story is applicable in the semiconductor industry. The shrinking transistor size and the need to enable efficient devices has increased the importance of capturing data from all the possible semiconductor product development process. However, the major hurdle in doing so is the cost associated with the process.

Semiconductor Data Is The Next-Gen Oil.

Shrinking transistors enable better devices and empower the long-term product development roadmap. Though, when it comes to the cost part of it, things start to get complicated. Cost is also the major reason why there are only three players (Intel, Samsung, and TSMC) that are battling the 5/3 nm race. The cost required to set up the FAB and the support system (equipment, facilities, tools, resources, and many other things) is too high and often requires long-term investment-wise planning. Even building a MINI-FAB today, will require billions of dollars to set up, and there on will take years to break even.

Setting up smaller research and development facility is an efficient way to capture semiconductor data, but it is not feasible to rely on smaller labs/setups for too long. In order to meet the worldwide demand, the facilities eventually have to expand.

– MINI-FAB >$1 Billion

– MEGA-FAB > $4 Billion

– GIGA-FAB > $12 Billion

– MINI/MEGA/GIGA = Defined Based On Wafer Capacity.

This makes the process of capturing and handling the semiconductor data crucial. Any data point that comes out of the pre-silicon or post-silicon stage has to go through a specific life cycle before being stored for long-term usage. This life cycle data handling process itself adds additional cost apart from the FAB investment. In the long run, the semiconductor companies understand the importance of setting up the data life cycle flow and have always invested relentlessly both in the process that leads to silicon and also the process required to generate data out of different silicon products.

Below is an overview of how the semiconductor data is handled and why each of these processes is vital. In nutshell, these steps are no different than how any big data will get handled. When it comes to the semiconductor industry, the major difference is the effort (cost and resources) it takes to generate data from different types of semiconductor products that often require large setups.

Generation: Generating semiconductor data requires a silicon wafer (with dies that are testable) and a test program that can run on the full wafer. Both of these processes demand different sets of tools and resources. A dedicated FAB is tasked up with creating a silicon wafer that has the product printed (repeatedly) across the full area. Which in itself is a costly process. On the other hand, a dedicated tester environment (OSAT) with different hardware and equipment is required to drive the test program. Such a long and delicate process not just requires product but also manufacturing, logistics, handling, and equipment resource. The sum of all these investments eventually allows semiconductor data to be generated. And without going into details, is understood how costly and time-demanding process this is.

Cleaning: Generating data out of the silicon is the first step. As explained above, it requires different sets of hardware equipment to drive the semiconductor data generation. The data in the majority of the cases are generated in a standard format, but still require a lot of post-processing and elimination techniques to make sense of it. This cleaning process is more on the software side and demands data processing tools that can help engineers understand different features of the silicon data. The cost associated is due to the setting up of the flow that allows semiconductor companies to capture the data at the source, which can then be easily sent to the servers for engineers to retrieve. There on, the cleaning steps start.

Profiling: Simply collecting random silicon data is not useful. The semiconductor product itself is going to be used in different systems and environments. This environment will push the product through different operating conditions. To ensure the product works under different conditions (temperature, process variation, current, and voltage settings), the development phase pushes the product/silicon through several testing criteria. These are often based on the industry-accepted standards (AEC, JEDEC, IPC, etc.). Carrying out all these tests to gather the promising semiconductor data (that will qualify the semiconductor product for the market) is challenging. The cost associated with it is also on the rise and thus adds another costly layer towards capturing the semiconductor data.

Mapping: The semiconductor data is often captured in big chucks. This can be at the wafer level or lot level. In both cases, it becomes really important to ensure that the data can be traced back to the die/part it originated from. The process to do so starts much before the test data is available. This can be via different marking to memory-based traceability techniques. This again points to the fact that the data mapping also requires base resources to achieve the target of not only ensuring the semiconductor data is available, but it is also easy to map it back to the source.

Analysis: Post all the possible data is captured and is available for engineers, the main task starts. While clean data (no skews or deviations) is the dream of every engineer, however, even with the cleanest data it becomes crucial to question different aspects of it. And, if there is a fail part, then finding the root cause is a must. This process requires sophisticated data exploration tools that bring efficiency. These tools should also be able to connect back to any historical/relevant data that can answer any deviation or miss alignment with new data. If data cannot answer all the questions, then comes the interesting part of putting together the root cause analysis plan. All this is not only a time-consuming process but also demands costly resources.

Visualization: Analysis and visualization go hand in hand. However, not all the tools are great at both the analysis and the visualization part. This pushes semiconductor data engineers towards exploring the data using different tools. In the majority of the cases, these tools are procured from the software data industry. But it also happens that the companies are willing to invest internally to come out with an easy visualization technique that can provide information as soon as the data is ready. This does require a dedicated team and tools that require capital.

Monitoring: Monitoring is another aspect of semiconductor data. It can be about the process steps involved during the semiconductor fabrication or about all the equipment being used for semiconductor product development. Each of these data points has to be monitored in real-time to ensure that there are no miss-steps during the fabrication or testing part of the product development. The environment required to set up monitoring and capturing monitored data again demands investment.

Storage: Given how time-consuming and costly the process to generate the semiconductor data is, it is very vital to ensure that every data point is stored for long-term usage. The next product development might 9from nowhere) require data from different products to establish a scale or reference. Doing so is only possible if the data is stored in a standard process and is easily retrievable. This is also the major reason to invest in long-term semiconductor data storage servers, which indeed requires long-term investment too.

In today’s data-driven world, semiconductor companies have to invest in the resources required to drive each of the steps in the cycle. Capturing, analyzing, and storing data in the world of semiconductors is more vital given how difficult and time-sensitive the product development process is.

Without thoroughly analyzing the real silicon data, it is impossible to conclude whether the product is going to be a success or not. This is why data is one of the major building blocks of the semiconductor industry.

THE REASONS FOR COSTLY SEMICONDUCTOR DATA

The life cycle of semiconductor data presents a clear picture of the resources required to capture the relevant semiconductor data points. As the industry moves towards much smaller and advanced technology-node along with innovative package technology, the cost associated with it will also be on the rise.

All these developments will have a major impact on the resources required to capture the semiconductor data, as each new semiconductor technology development will demand upgraded FABs and OSATs along with all the support tools and other hardware resources to equipment.

Below is the major reason why the costly semiconductor data is here to stay and how it is also impacting today’s node process and packages out in the market.

Equipment: The different sets of tools and equipment required to drive the successful development of a semiconductor product is key to generating relevant data. Each new design methodology, technology node, and packaging process demands new sets of equipment. This adds to the cost of development and leads to FAB/OSAT upgrade or expansion. This added cost is necessary to ensure that the semiconductor data can be generated and analyzed successfully. Thus showcasing clearly why the semiconductor data cost is on the rise.

Data Tool: The raw data is the first step towards analyzing how the product is behaving on the real silicon. To take a step further, investment is required to procure advanced data analytics tools. The feature-based subscription cost associated with it is also on the rise and is heavily impacting the data analysis part. On top of this, every other year there are a new set of programming solutions that pushes semiconductor data engineers towards a new way of analyzing data. This also requires investment not only in the tool but also in the training and skill upgrade part of it.

Skills: To make the most of the data also demands skills that take years to master. In today’s day and age, the explosion of new technology (on the software side) is also pushing engineers to capture new skill sets on the go. This requires companies to not only invest in core product development resources (FAB to OSAT) but also in people who can explore data with limited information and present the full picture.

Resources: Apart from human resources the data also demands a unique support environment. This can be a factory setup that enables data generation to a data warehouse that is storing all the data-related information. Such resources require a dedicated knowledge team and tools. All the cost associated with such process ends up producing the relevant semiconductor data. Without resources, it is impossible to do any task (not just data exploration).

Process: Technology-node to package to materials (and beyond) all go through a different life cycle and process. This process involves working in a dedicated lab that does require a unique set of tools. To ensure the process are right the tools have to be efficient and the combination of process and tools eventually leads to trustworthy data. The journey of capturing the semiconductor data is thus heavily dependent on these costly processes.

Research: To drive next-gen FETs and XPUs, continuous research is required and it also demands data to validate the new technology/solution. This means a dedicated setup/lab with the next-gen process, equipment, and data tools. All this adds to the costly process of generating the data for research and development activities.

The journey of semiconductor data is very interesting and high-tech. It certainly involves a lot of processes and steps that are dependent on different facilities, equipment, and human resources. As long as the goal is to come up with a new silicon solution, all these semiconductor resources will keep demanding high investment, and in the long run, it is the right thing to do.

The growing importance of semiconductor solutions in every aspect of life is also raising the question as to whether the semiconductor data is the next-gen oil.

Photo by Kvistholt Photography on Unsplash

THE IMPACT OF CLOUD IN SEMICONDUCTOR INDUSTRY

In the high-tech industry, software plays a crucial role. The ability to manage the task with a click of a button has a profound effect on different day-to-day activities. Software-driven digitization is also a key enabler in achieving productivity, which in turn allows companies/industries to capture the market worldwide.

Over the last few decades, the software delivery model has evolved. From desktop to browser to smartphone apps to cloud, the different software delivery models have increased productivity. Industries worldwide have realized the potential of software-driven solutions and have rightly adopted strategies to maximize software deployment via the cloud. The same is true for the semiconductor industry.

An advanced industry like semiconductor, where a single mistake can amount to zillions of dollars in losses, certainly needs solutions that can help plug any gaps in the product development process. These solutions can range from both the technical and business aspect of the product life cycle.

To achieve high design and manufacturing standards, the semiconductor industry also has been heavily relying on cloud-based software solutions. This is why over the last couple of decades all major segments of the semiconductor industry have adopted cloud strategy.

There are certainly cases where the transformation is happening today, but irrespective of the digital transformation journey, semiconductor design, and manufacturing houses realize the importance of hoping on to the cloud solutions and take advantage of every detail by capturing and connecting all the possible data points.

There are numerous ways in which cloud strategy is impacting the semiconductor industry:

Optimization: To track all the project and product details, cloud-based solutions play a vital role. The tools provide an ability to capture every change (technical and business) and ensure that the information is available on the go. This way of capturing the details allows optimized operations that ensure that the design to manufacturing flow is well connected to the specifications and the execution flow.

Defect And Error Free: By making use of smart and connected equipment with cloud-powered data analytic tools, semiconductor companies can ensure that the end product is defect-free. Deploying smart cloud tools that can capture gaps in the process/recipe ensures that there are no errors in the manufacturing flows.

On-Time: Delivering products on-time to the customer is key to capturing the large market, and it often requires swift coordination between multiple cross-functional teams. To connect and share the information seamlessly needs cloud-powered data-backed solutions that can track every detail from forecasting to material handling. Capturing the minutest of the details ensures that the product reaches the end customer on time.

End-To-End Process: Semiconductor industry is built on top of several segments, which includes the design houses to equipment manufacturers to FABs to manufacturing OSAT sites, and many more. All these individual segments should be connected end-to-end to provides a holistic view of how the product development, manufacturing, and delivery flow. This is where cloud-based solutions are useful, and such solutions also bring transparency.

To create a robust and high-quality semiconductor product, numerous data points are required. This is possible only with the help of tools and solutions that connect and provide a detailed view of different stages of the product development cycle. This is why a cloud-based end-to-end solution is heavily used in the semiconductor industry.

As more FABs and OSATs get established in different parts of the world, the cloud strategy should be one of the top priorities. Doing so, will only increase productivity and enable a better customer experience.

THE USE OF CLOUD IN SEMICONDUCTOR INDUSTRY

Cloud-based solutions are used at every step of semiconductor product development. The efficiency and productivity such tools bring to different stages of product development, ensures that the end-product is defect-free.

Different segments of the semiconductor product development require a unique cloud strategy. These can range from specific EDA tools to massive data storage instances. In the end, all these data points are connected/tied to a specific product. This way any team from anywhere can access the needed data/information to execute the task at hand.

Research And Development: Designing devices to form the circuits and layouts that will eventually get fabricated into a silicon product, is the first step towards developing a turnkey semiconductor solution. For several years, the semiconductor industry has relied on software-based solutions. In the last decade, the same software has moved from desktop to cloud. This has enabled designers and researchers to access files and libraries on-the-go. On top of that, it has taken away the need to have high-performance systems that were earlier required to carry out tons of simulations. Today, one can simply deploy and get on with other work while the simulations are running on the cloud. Using the cloud, companies can ensure that there is no IP infringement as the check can run on the fly and tools can compare design solutions against the massive poll of cell libraries. All this ensures that the right product is developed by following the industry-standard protocols.

Data Analysis: Data generation occurs at every stage of silicon development. Whether it is simulations in our design stage or fabrication/testing in FABs/OSATs. This is why advanced statistical analysis is carried out on every data point that is generated. Doing so on the fly means deploying solutions that can run closer to the data generating source. Also, if the data source is error-free, then data engineers can question it and take a deeper look by using analytical tools. In both scenarios, highly advanced tools are required. This means making use of cloud-based solutions that are easy to access and are loaded with features to enable accurate data exploration.

Supply Chain Management: Delivering products to the end customer often requires connecting several dots from different systems. This is why supply chain management is needed. It ensures the product and its processes are tracked using the unique bill of materials. This often requires relying on specific cloud-based tools that can swiftly retrieve any information related to the products to provide its full history from inception to delivery. Such a task without cloud tools will certainly invite errors.

Market Analysis: Capturing market trends to understand which products will give maximum profits is key to success. This often requires capturing different data points and then aligning the product roadmap as per the market (which means the customer) requirement. This ensures that the CapEx is diverted towards high revenue products. Such planning and projection are not possible without capturing different data points and customer developments. This is where cloud-based market analysis comes in handy and ensures that the projects are profitable.

Resource Management: Semiconductor equipment, tools, and labs require high CapEx and such investment is viable only if there is a high ROI. To achieve positive ROI, the resources need to be looked after and that requires efficient management. This means periodic maintenance to ensure minimum downtime. For such management, cloud-based solutions are deployed so that whenever tools go down or require maintenance, the data can be captured to raise the alert beforehand.

Factory Operation: Running FABs and OSATs 24×7 is key to ensuring that the facilities breakeven as quickly as possible. This requires capturing second to second activity of every corner of the factory. This is possible only when a connected system is deployed that can raise alarm in case of downtime and also provide remote status of every piece of equipment and tools. Cloud-based solutions already play a key role in this can thus are heavily used by semiconductor FABs and OSATs.

Logistics: Shipping is a big part of the semiconductor product development cycle. The wafers often come from a different outside supplier and then get fabricated at a different location, only to get tested in a different part of the world. This means optimized logistics and real-time tracking of material. While the majority of the logistics solution providers are heavily using cloud-based solutions to track and deliver packages, it so often happens that the semiconductor companies also have to invest internally in systems that can raise shipping labels apart from tracking where the customer deliveries are. This is why cloud-based logistics solutions are handy for semiconductor companies.

Archiving: Saving every data point related to the product is vital. This data ranges from design files to test data to financial records. All these need to be stored for a very long term so that whenever a query or comparison has to be done, the data can be easily retrieved. Quick retrieval and analysis are only possible if cloud solutions are deployed.

The use of the cloud in the semiconductor industry will keep growing. In the next few years, more sophisticated and smart tools will be deployed. Factory operations are already utilizing high-tech cloud solutions for product fabrication. Design and Supply Chain Management are not behind. The cloud market for the semiconductor industry will only keep growing with more new players entering the product development every year.

Photo by Mathew Schwartz on Unsplash

THE IMPACT OF SEMICONDUCTOR ROOT CAUSE ANALYSIS

Delivering high-grade products is the ultimate goal of every manufacturer. To do so, different industry-specific standards and processes are used. The same applies to the semiconductor manufacturing industry.

In semiconductors, as the technology-node shrinks, the cost to manufacture tiny silicon increases too. To ensure there is no waste of materials, time, and cost, several different strategies are deployed to screen the part before it moves ahead in the fabrication/manufacturing line.

One such strategy is root cause analysis. It so often happens, that the product being developed might encounter an issue during the qualification, testing, or packaging stage. Does not matter where the failure occurs, it is critical to understand the root cause. For an industry like semiconductor manufacturing, where every failing part can jeopardize not only the product quality itself but also the system to which it will eventually get glued in to. Such a scenario can certainly lead to millions of dollars in losses apart from a negative business reputation.

This is why holding on to every failing product and carrying out detailed root cause analysis is one of the major pillars of the semiconductor industry.

Over the last few decades, as the devices are getting tinier than ever, the importance of root cause analysis is growing due to the several impacts it has on the product:

Cost: Finding the root cause of why the product failed, empowers the team with data points to take necessary actions. It so often happens that the root cause analysis is more about setup or lab where the qualification or testing is being carried out and has nothing to do with the product itself. All root cause analysis and actions eventually can save cost by eliminating the need to re-do either the design or the qualification. Root cause analysis of failing products also ensures that the product will not fail in the field, thus saving years of investment.

TTM: Given the stiff competition in the semiconductor industry, the ultimate goal of every design and manufacturing house is to ensure that the product is launched within the planned time frame. In case of failures, root cause analysis provides a way to capture any severe design or process issue early on. This allows course correction to ensure the product still makes it to the market in time.

Quality: Qualification is an important process before the product gets released for production. Root cause analysis of any failing product during the development stage ensures that the product meets the high-quality industry standards. The standards increase as per the target domain and so does the importance of root cause analysis of the failing product/part.

DPPM/DPPB: Defective Parts Per Million/Billion is an industry-standard metric. The goal of semiconductor manufacturing is to lower DPPM/DPPB. In case of field failures, root cause analysis comes in handy, as it ensures the severe fails are captured to lower the DPPM/DPPB further. More so, when the product/part will be used for critical applications like automotive or wireless communications.

Root cause analysis has a major impact on semiconductor manufacturing. With new FETs and equipment launching every year, the role played by analyzing any smaller to big product-related failure is vital for both FAB and FAB-LESS companies.

THE STEPS OF SEMICONDUCTOR ROOT CAUSE ANALYSIS

To capture the root cause that leads to the failure of a product, several steps are followed. Each of these steps plays a key role in providing data points to establish the root cause to drive defect-free manufacturing.

Inspection: Any given part/product that fails first and foremost has to go through a detailed inspection. Depending on the state of the product (whether in wafer or die or assembled packaged form), the inspection is performed. The images generated by these are then used to establish if any non-technical cause (handling, stress, recipe issue, etc.) leads to the failure of the product. X-Ray and SEM are widely used as these processes provide internal product details.

Data: The data points for the failing part/product come from different stages. All the parts have marking to trace their origin. As a starting point, the first data point is collected by testing the product without damaging it. Another data point can be inspection data that FAB and OSAT might have on the failing part/product. Apart from these two, several other data points like recipe error, handling issue, the material used and many more, are captured to establish whether the root cause can be determined based on the data itself.

Testing: Testing can be done performed on: ATE and BENCH. On ATE, the part/product is tested to capture failing tests. This provides an early hint about which blocks of the part/product are failing. In case, there is no firm conclusion, then BENCH validation allows a way to find if the part indeed failed or there was a setup error. In many cases, different testing scenarios are also used to capture testing-related data points.

Reproduce Failure: This is the basis of the root cause analysis. Once the failure mode is understood, then the setup is replicated with an exact error to understand if the part/product repeatedly fails or not. This often requires a good setup and good parts/product, with the only varying parameter being the setup to capture the failing scenario/setup.

Localization: Root cause analysis can also lead to the discovery of issues in the product itself. This often requires the team to perform the root cause analysis to establish the location at which there are design issues. This is achieved with the help of all the above steps and most importantly using the equipment to show that when the part/product is biased, then it fails due to a hotspot found on the specific location of the layout. Advanced equipment are used to capture such data.

Documentation: In the end, everything needs to be documented. Documentation is a major part of not only capturing the reasons for part/product failure but also enables cross-functional learning. In the future, the documentation can be used to minimize the efforts to establish root cause in cases where the failing trend is similar. 8D and other problem-solving methods are used to document detailed root cause analysis.

Finding the cause of failure is like finding a needle in the haystack. The complexity the products are bringing, (due to advanced technology-node) is making root cause analysis an important and also difficult task at the same time.

THE PILLARS OF SEMICONDUCTOR ROOT CAUSE ANALYSIS

Performing root cause analysis is certainly a team effort. Experienced people play a vital role in reducing the time it takes to establish the root cause.

However, there are two major pillars without which root cause analysis is not possible. These pillars also require massive investment to establish and often have a high operating cost.

Lab: A dedicated space is required where experienced engineers can take the failing part and establish the root cause by biasing the part in a controlled environment. This requires skill-based training and the ability to understand any failing product/part that comes into the lab.

Equipment: To run a root cause analysis lab, advanced equipment are also needed. These equipment can perform detailed inspection, apart from carrying out different setups/testing to reproduce the failure.

The majority of the semiconductor companies have an in-house dedicated lab to perform root cause analysis. Some companies prefer to outsource. Eventually, it is all about managing resources and lowering the cost without compromising on the DPPM/DPPB.

As the industry moves towards more complex products that will power everything everywhere, capturing the root cause of every failing product during development or the production stage, will become more vital than ever.