The global semiconductor industry has entered the Red Sea competition, and there are more and more IC design companies, especially in China, the number of companies has shown an explosive growth trend in recent years. In this industry context, competition is becoming more and more intense, and product homogeneity is unavoidable. Taking mobile phones as an example, several major brand manufacturers are using Qualcomm’s baseband and AP Chips. Although their performance is very good, it is difficult to do so. Make their own mobile phones have prominent highlights and features, and the homogenization competition is serious. It is precisely because of this that more and more mobile phone manufacturers have begun to develop their own main chips. The same is true for Internet companies, many of which are developing their own server chips, for the same reason.
System manufacturers are like this, and IC design manufacturers in the upstream of the industry chain are also facing similar problems. In order to make their own chips have unique characteristics, many manufacturers are showing their magic power. It is an IC design and system solution, and also self-developed EDA tools and IP, especially the software and hardware simulation tools developed by it, which are very distinctive and provide great support for the company’s self-developed chips.
On November 7th, at the China International Import Expo, MimicTurbo, the self-developed EDA tool of Chipqiyuan, attracted attention, and was especially favored by FPGA giant Xilinx. The two parties jointly signed a memorandum of cooperation to reach an in-depth cooperation intention in FPGA application and EDA tool production in the future. Ms. Tang Xiaolei, vice president of Xilinx Greater China, said that Chip Qiyuan is a long-term partner of Xilinx, and the signing of this memorandum of cooperation means that the cooperation between the two parties is developing in a more stable and in-depth direction. At present, a variety of products of Core Qiyuan and Xilinx’s technical capabilities are highly matched and complement each other. The two parties will carry out cooperation in more dimensions and volume in 2022.
Based on Xilinx’s UltraScale+ VU19P FPGA, Chip Qiyuan officially launched the MimicTurbo GT card sample. The product is equipped with MimicTurbo prototype verification software, which simplifies the deployment of FPGA-based prototype verification and can speed up chip design and software development in the chip development process. Verification works.
Targeted self-developed EDA tools and IP
As an IC design company, the main track of Chipqiyuan is 5G communication and data center chips and overall solutions. Why do you need to develop EDA tools by yourself? Regarding this issue, Mr. Lu Sheng, the company’s chairman and CEO, said: “We do things like this mainly for our own use. Therefore, we will not make all the EDA tool chains like the top three EDA companies. In the entire tool chain, we believe that there is room for improvement, or the tools that EDA companies cannot provide and meet our needs.”
Lu Sheng has more than 20 years of working experience in Silicon Valley and has achieved success in many startups. He said that in the past so many years of his career, his team has been raising various demands to EDA manufacturers, and now many of them are widely used in the market. The EDA tools were developed in a targeted manner by EDA manufacturers after they put forward their needs, benefiting the entire industry.
The early chip verification must wait until the tape-out comes back, which poses a great risk. If there is a problem with the chip, then the cost of this tape-out and the large amount of verification work by the previous engineers will be lost. , and will prolong the chip listing cycle, it is likely to miss the market window. These are difficult to bear for the majority of small and medium-sized IC design companies, and the cost is too high.
Therefore, it is necessary to find a way to minimize these potential risks, and when there is no good solution in the industry, self-developed EDA is a good method, because chip design companies have the most demand for themselves and the problems to be solved. Understand that the developed EDA tools are highly targeted.
In addition, such EDA tools can also help other IC design companies in the industry to solve problems. At the same time, through external authorization, the R&D cost of EDA tools can also be flattened, achieving the effect of serving multiple purposes with one stone.
Therefore, self-developed EDA tools and differentiated competition have become the company’s philosophy of Chipqiyuan. Only in this way can we make high-quality chips with distinctive features that can be quickly marketed.
Lu Sheng said: “Although the self-developed EDA tool is our secret weapon, Chip Qiyuan still upholds an open attitude and is willing to share their research results with the business. At the same time, we are still innovating to accumulate more Experience and secret technical weapons.”
In addition to EDA tools, self-developed IP is equally important. Lu Sheng said that this is also the most difficult task in the entire team. It is more difficult than chip design, because designing a chip only needs to “connect” the existing IP. For design engineers, IP is like a black box, you don’t need to know what’s inside, you just need to know how to use it. For IP R&D engineers, it is difficult to make various standards into IP so that chip companies can use it.
For the entire chip R&D team, if they have the ability to modify IP, there is an additional means of differentiated competition. For example, Lu Sheng said that when he was working on Broadcom’s 100M/1000M Ethernet chip, he made a differentiated design by changing the IP, that is, when he detected that the other end of the chip was also his own chip, he could The speed is switched to 200M or 2000M, so that when customers use my chip on both ends of the network, the network bandwidth can be greatly improved, which is the advantage of differentiated competition. In many cases, these are non-standard, and by continuously differentiated design, realizing innovation, and bringing value to the market and customers, it is very possible to lead new standards step by step, so as to be in an advantageous position in the future competition.
Differentiated competitive weapon MimicPro
For an IC design company, its core competitiveness is the chip design method, that is, know-how. Based on this, Lu Sheng said that years of accumulation of technology and experience have enabled him to understand the problems that IC design companies and chip design engineers need to solve. , as well as the tools and methods needed, with a deep understanding. As a result, he saw a common pain point in the chip design process. MimicPro, a software and hardware simulation tool developed by Chipqigen, is a very valuable EDA verification environment.
The tools represented by MimicPro serve the front-end work of the chip design process. Under normal circumstances, in a chip design process, the energy and resources spent on the front end account for 70% of the total, including time, manpower and funds. To shorten the cycle, reduce costs, strive for a successful tape-out, and achieve rapid time-to-market, it is necessary to find all bugs in the shortest time. In order to solve these problems, EDA manufacturers are increasing investment in research and development, in order to provide customers with tools that can solve various new problems. However, in many cases, only chip design engineers can understand some problems, and EDA manufacturers can’t see them at all. In this regard, Lu Sheng believes that the separation of prototype simulation and hardware simulation is a big problem. Some of the underlying logic in the core technology of EDA manufacturers limits their design, and prototype simulation lacks the ability to analyze and correct errors. The hardware simulation tool is already very perfect, but the speed has certain limitations. At the same time, there is a fault in the development environment, that is, the problems found in the prototype simulation may not be realized in the hardware simulation, because its speed is not enough to support its Operating system and upper layer software.
Lu Sheng said: “We stand from a user’s point of view, based on the need for 5G and data center communication chips, and the need for integrated software and hardware verification, we developed MimicPro, which integrates the advantages of hardware simulation and prototype simulation, error correction With the functions and fast features, we integrate two separate environments into one integrated environment, which can do a lot of work.”
MimicPro can also meet automotive-grade applications. The most important function in the car is to do the so-called fault response. Under the premise of smooth operation, if there is a fault, how will the system respond, and where is the problem of the entire system at the end, whether it is a software problem or a hardware problem, MimicPro is all about can lead to great solutions.
In July of this year, MimicPro, a prototype verification and simulation system based on Xilinx UltraScale FPGA, was officially mass-produced and supplied, and MimicPro with 32 FPGAs will also be officially delivered to customers next month.
According to Lu Sheng, in addition to in-depth cooperation with Xilinx, Amlogic is already using the MimicPro simulation tool in China.
Mainly focus on DPU chips and smart network cards
EDA tools are only a feature of Chipqiyuan. The company’s main business is 5G and data center chips and solutions, especially DPU chips and smart network cards, which are the company’s main products. Using DPU and intelligent network card, compared with x86 architecture products, can greatly improve efficiency, and greatly reduce the procurement cost of servers, and can also greatly reduce electricity bills.
It is precisely because there are so many advantages that at present, many of the world’s top chip manufacturers are developing their own DPUs. Of course, the names of each will be different, such as Intel’s IPU, after Nvidia acquired Mylos, it exploded. The global technology media is paying attention to DPU. In addition, Marvell and AMD are doing similar things, especially AMD’s acquisition of Xilinx, which is to develop smart network cards. The actions of these giants fully demonstrate that the DPU and smart network card market has unlimited potential.
Lu Sheng said that at present, Coreqiyuan is the only domestic manufacturer that mass-produces DPU chips and smart network cards. In this field, Coreqiyuan is directly targeting international giants, and their goal is to empower globalization and lead the industry. .
Lu Sheng said that Chipqiyuan started the research and development of DPU very early. This is also based on his accumulation of technology and experience in Silicon Valley for more than 20 years. Lu Sheng has seen the 10 years of the semiconductor outbreak in Silicon Valley, that is, from 1994 to 2004. , At that time, the industry moved forward rapidly in accordance with the time plan of Moore’s Law. Later, the development of the semiconductor industry slowed down. Until now, the influence of Moore’s Law has become weaker and weaker. In the past 20 years, he has seen many regularities in the entire industry. At present, what is happening in China’s semiconductor industry and the process is very similar to that of Silicon Valley in the past 10 years.
Such accumulation allows Lu Sheng and his team to predict what may happen in the future of China’s chip industry, including technological bottlenecks and the emergence of alternative technologies. Based on this, Chipqiyuan saw the development prospects of DPU and smart network cards two and a half years ago, and began to carry out related work, including mergers and acquisitions of related chip manufacturers. In 2017, they were developing smart network cards using FPGAs and the company’s TCAM chips (network search engines, used in core routers). At present, a new solution based on DPU is adopted.
In 2020, Chipqiyuan’s smart network card products entered the China Mobile supply chain. At present, the company’s goal is to seize the Chinese market as quickly as possible and provide customers with solutions with higher performance and lower power consumption.
A tape-out success
With the increasing complexity of applications, it is difficult to make a chip successful, especially a tape-out, and it is impossible to do without some strong “foundation” support.
In the more than 20 years of working in Silicon Valley, several companies that Lu Sheng participated in were successively acquired by large chip companies. The first company I worked for was listed on Nasdaq, and the second company focused on GPUs, which was invested by the chairman of Nintendo. The technical team with him as the core achieved a successful tape-out at that time. 400 million US dollars sold to ATI, and later, ATI sold this division to AMD.
After that, Lu Sheng began to make network switch chips. The company was acquired by Broadcom in 2000. Later, he went to Marvell. At that time, Marvell acquired Intel’s mobile phone chip department. Lu Sheng’s team only spent 6 months. In time, a mobile phone chip was made, and it was still a tape-out success.
These experiences gave him a deep understanding of switches, WiFi, Bluetooth, and mobile phone chips. Lu Sheng said that in order for these high-end and complex chips to be successfully taped out at one time, a unique design method and the support of high-end software tools are necessary. In addition, there are also high requirements for IP. In order to make chips not the same and have differentiated competitiveness, the corresponding IP must be differentiated, which requires self-development.
Therefore, since the establishment of Chipqiyuan, it has been benchmarking high-end chips. In particular, to maintain a successful level of tape-out, it must have its own core competitiveness. Self-developed EDA tools and IP are aimed at high-level chips. Design the foundation.
It is precisely based on such technical strength and foundation that Coreqiyuan is favored in the capital market. Recently, the company announced the completion of the Pre-A4 round of financing of hundreds of millions of yuan. This round of financing was led by China Internet Investment Fund (“China Internet Investment”), China Resources Capital Runke Fund, Xingwang Investment, Yuntai Capital, Zhenghai Capital. Following the investment, the old shareholder Yimei Investment (Shanghai North High-tech Big Data Fund) continued to follow the investment firmly. This financing will further support the R&D investment of the next-generation DPU chip, accelerate the ecological layout in 5G and cloud data centers, and continue to strengthen the leading position of the chip in this field in China.
The global semiconductor industry has entered the Red Sea competition, and there are more and more IC design companies, especially in China, the number of companies has shown an explosive growth trend in recent years. In this industry context, competition is becoming more and more intense, and product homogeneity is unavoidable. Taking mobile phones as an example, several major brand manufacturers are using Qualcomm’s baseband and AP chips. Although their performance is very good, it is difficult to do so. Make their own mobile phones have prominent highlights and features, and the homogenization competition is serious. It is precisely because of this that more and more mobile phone manufacturers have begun to develop their own main chips. The same is true for Internet companies, many of which are developing their own server chips, for the same reason.
System manufacturers are like this, and IC design manufacturers in the upstream of the industry chain are also facing similar problems. In order to make their own chips have unique characteristics, many manufacturers are showing their magic power. It is an IC design and system solution, and also self-developed EDA tools and IP, especially the software and hardware simulation tools developed by it, which are very distinctive and provide great support for the company’s self-developed chips.
On November 7th, at the China International Import Expo, MimicTurbo, the self-developed EDA tool of Chipqiyuan, attracted attention, and was especially favored by FPGA giant Xilinx. The two parties jointly signed a memorandum of cooperation to reach an in-depth cooperation intention in FPGA application and EDA tool production in the future. Ms. Tang Xiaolei, vice president of Xilinx Greater China, said that Chip Qiyuan is a long-term partner of Xilinx, and the signing of this memorandum of cooperation means that the cooperation between the two parties is developing in a more stable and in-depth direction. At present, a variety of products of Core Qiyuan and Xilinx’s technical capabilities are highly matched and complement each other. The two parties will carry out cooperation in more dimensions and volume in 2022.
Based on Xilinx’s UltraScale+ VU19P FPGA, Chip Qiyuan officially launched the MimicTurbo GT card sample. The product is equipped with MimicTurbo prototype verification software, which simplifies the deployment of FPGA-based prototype verification and can speed up chip design and software development in the chip development process. Verification works.
Targeted self-developed EDA tools and IP
As an IC design company, the main track of Chipqiyuan is 5G communication and data center chips and overall solutions. Why do you need to develop EDA tools by yourself? Regarding this issue, Mr. Lu Sheng, the company’s chairman and CEO, said: “We do things like this mainly for our own use. Therefore, we will not make all the EDA tool chains like the top three EDA companies. In the entire tool chain, we believe that there is room for improvement, or the tools that EDA companies cannot provide and meet our needs.”
Lu Sheng has more than 20 years of working experience in Silicon Valley and has achieved success in many startups. He said that in the past so many years of his career, his team has been raising various demands to EDA manufacturers, and now many of them are widely used in the market. The EDA tools were developed in a targeted manner by EDA manufacturers after they put forward their needs, benefiting the entire industry.
The early chip verification must wait until the tape-out comes back, which poses a great risk. If there is a problem with the chip, then the cost of this tape-out and the large amount of verification work by the previous engineers will be lost. , and will prolong the chip listing cycle, it is likely to miss the market window. These are difficult to bear for the majority of small and medium-sized IC design companies, and the cost is too high.
Therefore, it is necessary to find a way to minimize these potential risks, and when there is no good solution in the industry, self-developed EDA is a good method, because chip design companies have the most demand for themselves and the problems to be solved. Understand that the developed EDA tools are highly targeted.
In addition, such EDA tools can also help other IC design companies in the industry to solve problems. At the same time, through external authorization, the R&D cost of EDA tools can also be flattened, achieving the effect of serving multiple purposes with one stone.
Therefore, self-developed EDA tools and differentiated competition have become the company’s philosophy of Chipqiyuan. Only in this way can we make high-quality chips with distinctive features that can be quickly marketed.
Lu Sheng said: “Although the self-developed EDA tool is our secret weapon, Chip Qiyuan still upholds an open attitude and is willing to share their research results with the business. At the same time, we are still innovating to accumulate more Experience and secret technical weapons.”
In addition to EDA tools, self-developed IP is equally important. Lu Sheng said that this is also the most difficult task in the entire team. It is more difficult than chip design, because designing a chip only needs to “connect” the existing IP. For design engineers, IP is like a black box, you don’t need to know what’s inside, you just need to know how to use it. For IP R&D engineers, it is difficult to make various standards into IP so that chip companies can use it.
For the entire chip R&D team, if they have the ability to modify IP, there is an additional means of differentiated competition. For example, Lu Sheng said that when he was working on Broadcom’s 100M/1000M Ethernet chip, he made a differentiated design by changing the IP, that is, when he detected that the other end of the chip was also his own chip, he could The speed is switched to 200M or 2000M, so that when customers use my chip on both ends of the network, the network bandwidth can be greatly improved, which is the advantage of differentiated competition. In many cases, these are non-standard, and by continuously differentiated design, realizing innovation, and bringing value to the market and customers, it is very possible to lead new standards step by step, so as to be in an advantageous position in the future competition.
Differentiated competitive weapon MimicPro
For an IC design company, its core competitiveness is the chip design method, that is, know-how. Based on this, Lu Sheng said that years of accumulation of technology and experience have enabled him to understand the problems that IC design companies and chip design engineers need to solve. , as well as the tools and methods needed, with a deep understanding. As a result, he saw a common pain point in the chip design process. MimicPro, a software and hardware simulation tool developed by Chipqigen, is a very valuable EDA verification environment.
The tools represented by MimicPro serve the front-end work of the chip design process. Under normal circumstances, in a chip design process, the energy and resources spent on the front end account for 70% of the total, including time, manpower and funds. To shorten the cycle, reduce costs, strive for a successful tape-out, and achieve rapid time-to-market, it is necessary to find all bugs in the shortest time. In order to solve these problems, EDA manufacturers are increasing investment in research and development, in order to provide customers with tools that can solve various new problems. However, in many cases, only chip design engineers can understand some problems, and EDA manufacturers can’t see them at all. In this regard, Lu Sheng believes that the separation of prototype simulation and hardware simulation is a big problem. Some of the underlying logic in the core technology of EDA manufacturers limits their design, and prototype simulation lacks the ability to analyze and correct errors. The hardware simulation tool is already very perfect, but the speed has certain limitations. At the same time, there is a fault in the development environment, that is, the problems found in the prototype simulation may not be realized in the hardware simulation, because its speed is not enough to support its Operating system and upper layer software.
Lu Sheng said: “We stand from a user’s point of view, based on the need for 5G and data center communication chips, and the need for integrated software and hardware verification, we developed MimicPro, which integrates the advantages of hardware simulation and prototype simulation, error correction With the functions and fast features, we integrate two separate environments into one integrated environment, which can do a lot of work.”
MimicPro can also meet automotive-grade applications. The most important function in the car is to do the so-called fault response. Under the premise of smooth operation, if there is a fault, how will the system respond, and where is the problem of the entire system at the end, whether it is a software problem or a hardware problem, MimicPro is all about can lead to great solutions.
In July of this year, MimicPro, a prototype verification and simulation system based on Xilinx UltraScale FPGA, was officially mass-produced and supplied, and MimicPro with 32 FPGAs will also be officially delivered to customers next month.
According to Lu Sheng, in addition to in-depth cooperation with Xilinx, Amlogic is already using the MimicPro simulation tool in China.
Mainly focus on DPU chips and smart network cards
EDA tools are only a feature of Chipqiyuan. The company’s main business is 5G and data center chips and solutions, especially DPU chips and smart network cards, which are the company’s main products. Using DPU and intelligent network card, compared with x86 architecture products, can greatly improve efficiency, and greatly reduce the procurement cost of servers, and can also greatly reduce electricity bills.
It is precisely because there are so many advantages that at present, many of the world’s top chip manufacturers are developing their own DPUs. Of course, the names of each will be different, such as Intel’s IPU, after Nvidia acquired Mylos, it exploded. The global technology media is paying attention to DPU. In addition, Marvell and AMD are doing similar things, especially AMD’s acquisition of Xilinx, which is to develop smart network cards. The actions of these giants fully demonstrate that the DPU and smart network card market has unlimited potential.
Lu Sheng said that at present, Coreqiyuan is the only domestic manufacturer that mass-produces DPU chips and smart network cards. In this field, Coreqiyuan is directly targeting international giants, and their goal is to empower globalization and lead the industry. .
Lu Sheng said that Chipqiyuan started the research and development of DPU very early. This is also based on his accumulation of technology and experience in Silicon Valley for more than 20 years. Lu Sheng has seen the 10 years of the semiconductor outbreak in Silicon Valley, that is, from 1994 to 2004. , At that time, the industry moved forward rapidly in accordance with the time plan of Moore’s Law. Later, the development of the semiconductor industry slowed down. Until now, the influence of Moore’s Law has become weaker and weaker. In the past 20 years, he has seen many regularities in the entire industry. At present, what is happening in China’s semiconductor industry and the process is very similar to that of Silicon Valley in the past 10 years.
Such accumulation allows Lu Sheng and his team to predict what may happen in the future of China’s chip industry, including technological bottlenecks and the emergence of alternative technologies. Based on this, Chipqiyuan saw the development prospects of DPU and smart network cards two and a half years ago, and began to carry out related work, including mergers and acquisitions of related chip manufacturers. In 2017, they were developing smart network cards using FPGAs and the company’s TCAM chips (network search engines, used in core routers). At present, a new solution based on DPU is adopted.
In 2020, Chipqiyuan’s smart network card products entered the China Mobile supply chain. At present, the company’s goal is to seize the Chinese market as quickly as possible and provide customers with solutions with higher performance and lower power consumption.
A tape-out success
With the increasing complexity of applications, it is difficult to make a chip successful, especially a tape-out, and it is impossible to do without some strong “foundation” support.
In the more than 20 years of working in Silicon Valley, several companies that Lu Sheng participated in were successively acquired by large chip companies. The first company I worked for was listed on Nasdaq, and the second company focused on GPUs, which was invested by the chairman of Nintendo. The technical team with him as the core achieved a successful tape-out at that time. 400 million US dollars sold to ATI, and later, ATI sold this division to AMD.
After that, Lu Sheng began to make network switch chips. The company was acquired by Broadcom in 2000. Later, he went to Marvell. At that time, Marvell acquired Intel’s mobile phone chip department. Lu Sheng’s team only spent 6 months. In time, a mobile phone chip was made, and it was still a tape-out success.
These experiences gave him a deep understanding of switches, WiFi, Bluetooth, and mobile phone chips. Lu Sheng said that in order for these high-end and complex chips to be successfully taped out at one time, a unique design method and the support of high-end software tools are necessary. In addition, there are also high requirements for IP. In order to make chips not the same and have differentiated competitiveness, the corresponding IP must be differentiated, which requires self-development.
Therefore, since the establishment of Chipqiyuan, it has been benchmarking high-end chips. In particular, to maintain a successful level of tape-out, it must have its own core competitiveness. Self-developed EDA tools and IP are aimed at high-level chips. Design the foundation.
It is precisely based on such technical strength and foundation that Coreqiyuan is favored in the capital market. Recently, the company announced the completion of the Pre-A4 round of financing of hundreds of millions of yuan. This round of financing was led by China Internet Investment Fund (“China Internet Investment”), China Resources Capital Runke Fund, Xingwang Investment, Yuntai Capital, Zhenghai Capital. Following the investment, the old shareholder Yimei Investment (Shanghai North High-tech Big Data Fund) continued to follow the investment firmly. This financing will further support the R&D investment of the next-generation DPU chip, accelerate the ecological layout in 5G and cloud data centers, and continue to strengthen the leading position of the chip in this field in China.
The Links: LQ5AW136 LMG7520RPFC