At the end of 2018, Intel’s largest FPGA innovation center in the world landed in Chongqing. At the beginning of its establishment, Intel made clear its goals, hoping to build Chongqing into China’s FPGA development through accelerated data centers, FPGA innovation summits, FPGA innovation competitions, training and certification. The talent cultivation center, ecological innovation center, and industry gathering center will promote the development of the entire FPGA ecosystem.
Among them, in order to solve the most difficult talent problem, the Intel FPGA Innovation Center set a goal of cultivating more than 10,000 FPGA professionals in China within three years with ecological partners in 2019.
Now, the three-year deadline is approaching, and the Intel FPGA Ten Thousand People Plan has also reached the last mile. The ecological construction driven by talent training has begun to show results.
Over 8,000 FPGA students graduated, and the employment rate of full-time training is 100%
“Our goal is to train 10,000 FPGA engineers in three years, and so far, we have trained more than 8,000 FPGA engineers and developers.”
At the Intel FPGA China Innovation Center’s cutting-edge technology and new product launch conference on Wednesday, Zhang Rui, general manager of Intel’s FPGA Innovation Center, introduced the latest situation of Intel’s FPGA Ten Thousand People Program.
Although compared with CPU, GPU and other markets, the market size of FPGA is much smaller, but the market space is still considerable. IDC predicts that by 2024, the global acceleration server market will reach $28.4 billion, and the global FPGA market will reach $7.155 billion. Among them, the development space of electronic test instruments, communications and networks, aerospace, transportation and other fields exceeds 59%.
The future market demand for FPGA is huge, but the supply of talents is seriously insufficient.
According to the “FPGA Industry Development Status and Talent Training Research Report” jointly released by Intel and IDC in 2020, the current FPGA talent supply in the market can only better meet the needs of primary talents, and more than 70% of the interviewed companies or institutions believe that the current intermediate level There is a serious shortage of engineers above the proficiency level. 60% of the respondents believe that the lack of a training system, the inability to meet the needs of student training, and the disconnection between theory and practice are the biggest challenges faced by colleges and universities in cultivating FPGA talents. Other challenges include high learning threshold, long learning cycle, lack of faculty, etc.
Therefore, if you want to build an FPGA ecosystem, talent development is the key at the moment. In response to various problems in FPGA talent training, Intel PFGA Innovation Center has explored a clear training model for nearly three years.
Zhang Rui said that the Innovation Center has now officially published 13 textbooks, and has established a very complete and professional training system. “This training system can be passed on as a theoretical system for professional FPGA engineers. We also have in-depth cooperation with many universities to offer relevant elective courses, compulsory courses, and joint laboratories and joint development projects.”
The so-called complete professional training system means that the innovation center has divided the FPGA talents into primary, intermediate, advanced software direction, advanced hardware direction, and expert level, and set up different curriculum systems for each level, for example, if students want To get the FPGA junior engineer certification, you only need to pass the online written test, but the expert-level certification also needs to pass the comprehensive project experimental assessment and interview.
In Zhang Rui’s view, off-the-job training is the most outstanding part of the innovation center’s FPGA talent training. “At present, the employment rate of our offline full-time training is 100%. In 3 to 4 months, the trainees have one more life choice, which has changed the trainees’ lives to some extent.”
Leifeng.com learned that the training objects of off-the-job training courses are mainly fresh graduates of electronics or artificial intelligence related majors for undergraduates and higher vocational colleges, mainly for the cultivation of “software + hardware” compound talents that enterprises need. The training will be carried out in each stage, and the trainees who complete the training will have the ability to develop FPGA software and hardware, and be able to engage in FPGA logic development and artificial intelligence application development. And with the maturity of the training system, the full-time training cycle has been compressed from four months to three months.
Product technology has also made new progress, and FPGA ecological construction has accelerated
In addition to the steady progress of talent training, Intel also released two new FPGA products – Agilex FPGA and Stratix10 NX FPGA at the FPGA China Innovation Center’s cutting-edge technology and new product launch conference on Wednesday.
Among them, Agilex FPGA adopts Intel’s 10nm SuperFin process technology, Intel’s proprietary Embedded Multi-die Interconnect Bridge (EMIB) integrated 3D Heterogeneous System-in-Package (SiP), and innovative chip-based architecture, which can provide various applications Customized connectivity and acceleration features, compared to Stratix 10 FPGAs, Agilex FPGAs deliver a 45 percent increase in performance and up to 40 percent lower power consumption, enabling use in areas such as edge, networking, and data centers.
Stratix 10 NX FPGA is an artificial intelligence enhanced version of Stratix. Based on traditional Stratix, artificial intelligence engines or modules are added to adapt to rapidly changing AI application scenarios. In contrast, all reads and writes of the tensor module in GPU go through memory, while the tensor module of 10 NX is directly interconnected, which is much more flexible and efficient than traditional GPU.
These FPGA acceleration cards can be connected to Intel Xeon Scalable processor-based servers through PCIe, and support specific industries that need acceleration through the cloud acceleration center, including industry, artificial intelligence, AIoT, 5G, etc.
“Since the launch of the Cloud Acceleration Center in March 2019, we have successively supported or incubated various enterprises including hundreds of industry leaders, mainly targeting some small and medium-sized customers and colleges and universities,” said Zhang Rui.
Optech, an optical communication R&D and supply chain management company, has introduced the Intel Stratix 10 full-stack solution into its own products. Optech said that in addition to introducing Intel FPGA products, its FPGA talents mainly come from the Intel FPGA Innovation Center.
Whether it is in FPGA talent training or product innovation, Intel is making steady progress, and it is one step closer to the goal of promoting the development of the FPGA ecosystem.
This article is reprinted from: https://www.leiphone.com/category/chips/fgGx61cP0ixjBked.html
This site is for inclusion only, and the copyright belongs to the original author.