“The development of technologies such as cloud computing, the Internet of Things, and 5G has brought us to a digital world. Data is showing an exponential growth, with forecasts estimating that in 2020 an average internet user will generate 1.5GB of data per day.
The development of technologies such as cloud computing, the Internet of Things, and 5G has brought us to a digital world. Data is showing an exponential growth, with forecasts estimating that in 2020 an average internet user will generate 1.5GB of data per day.
So much data also brings us great challenges. Only by making great adjustments to the infrastructure of cloud computing and data center networks can we meet the needs and challenges of the big data era.
FPGAs: The Key to Building a Smart, Connected Future
According to forecasts, by 2020, there will be 50 billion terminals connected to the Internet, and now the number of connected terminals is only 8 billion. The number of these terminals networked in the future will be higher than the bandwidth required by each of them today. Cisco predicts that by 2020, IP traffic will reach 2,300EB per year.
The data center will be connected to these terminals, and these terminals will be connected to the data center. This method is called a virtuous cycle. This cycle will begin to accelerate with the development of the Internet of Things, which means that the network must be at a higher level. To process more data at a higher speed, the data center must also do more complex calculations, process larger data sets, and even embedded terminals and devices have to do a lot of local computing.
On top of that, data centers are also dealing with some more challenging workloads. For example, for big data analysis and machine learning, we can finally see that there will be some bottlenecks and computing hotspots in the whole cycle, which cannot be dealt with by the CPU alone, so at this time, a new kind of Technology – FPGA.
FPGAs are designed to accelerate computing
Recently, Intel listed six global markets where FPGAs are targeted at the earliest and most transformational applications: 5G wireless communications, radar and aerospace, networking, cloud computing, smart cities and unmanned vehicles. No matter which market it is, they are actually facing the same challenge, that is, the number of connections is increasing dramatically, and the computational complexity is also increasing, which creates bottlenecks and hot spots.
FPGA can help us deal with these challenges well. With an FPGA, your system’s design isn’t locked in, development time isn’t as long, and it’s more flexible than these dedicated hardware.
Taking 5G as an example, we are currently moving towards the era of 5G wireless communication. We can see that 5G will require larger bandwidth and more complex digital signal processing capabilities in the future, but the 5G standard has not yet been finalized, so we I don’t know what kind of power 5G will have after connecting these terminals of consumers and the Internet of Things to the cloud.
Now Intel is in a very unique position to help customers meet the challenges posed by 5G, such as fixed function and software, hardware, programmable endpoints, and end-to-end solutions, in the whole virtuous circle At every point, a mature solution is required.
Especially in the 5G network, FPGA can have the following capabilities, such as accelerating the calculation of MIMO antenna and baseband signal processing, security, and other functions that may restrict the development of the network.
CPU+FPGA+… Heterogeneity will become the norm
In addition to products that directly use FPGA, the CPU+FPGA architecture is also a combination that everyone uses a lot. Intel’s CPU plus FPGA can replace these proprietary architectures, providing some capabilities such as local computing, connection and analysis. It only needs to be achieved by accelerating a few key functions.
In addition, users can also put these many islands on a single Intel architecture, coupled with the FPGA environment, coupled with the 5G network, which can actually reduce the cost of the entire development and reduce the error rate. , and also allows us to have data analysis and machine learning capabilities, so that we can understand what the network-generated data can bring to customers, so that customers can make more informed decisions.
FPGA is crucial for heterogeneous computing or heterogeneous architecture. FPGA is imagined as an advanced multi-function accelerator, on the one hand, it can bring maximum programming flexibility, support highly differentiated products, and you can also reconfigure it in the field to virtually accelerate any digital algorithm. In addition, they also support parallel computing, so they are especially suitable for the future computing world, and more importantly, their performance is better than CPU or GPU in terms of throughput, implementation speed and energy consumption. ten times.
In addition, it can be seen that it can process larger data with lower latency, faster than traditional software-based products running on hardware products. It is because of the above advantages of FPGA that FPGA is equivalent to hardware performance and software programmability, so it becomes particularly important for heterogeneous computing environments or architectures.
Intel FPGAs are flexible, low-latency, high-throughput, and energy-efficient, making them ideal solutions for accelerating AI applications. Altera merged into Intel and became the Intel PSG Division, gaining access to Intel’s industry-leading three-year process technology, bringing greater capacity and faster speeds. The latest STRATIX 10 high-end chips are already sampling, ARRIA 10 products are already in full production, and CYCLONE 10 and MAX10 are already available for mass shipment.
In the next few years, the integration of large amounts of data will make data centers and clouds face huge challenges and changes. Heterogeneous acceleration and dynamic demands in different environments will be met by FPGAs. Intel’s efforts in FPGAs will also provide us life solves many practical problems.
FPGAs will be everywhere, how about learning an FPGA in 2019?
We see FPGAs in more and more designs and in more and more markets, and they themselves are becoming more complex as they are incorporated into more and more systems.
A decade ago, the main markets for FPGAs were industrial, medical, automotive, and aerospace.
These markets are still going strong, but FPGAs are also starting to play a role in artificial intelligence, data centers, the Internet of Things, network communications, test and measurement instruments, wireless and wireline communications infrastructure, and even nuclear power plants.
As one of the most important high-tech technologies, FPGA/IC logic design and development has become one of the most promising industries, especially for FPGA system engineers who are familiar with hardware architecture. According to statistics, China has a huge demand gap for FPGA design talents every year, and the salary of FPGA design talents is also the highest in the industry. However, since there are relatively few FPGA specialties in the school, even if there are, they have been exposed to FPGA. So as people who want to emerge in this industry, domestic FPGA is an “uncultivated” treasure, we must study hard and seize the opportunity.