Introduction Everyone has heard of FPGA more or less, such as Bitcoin mining, or Microsoft said before that it will use FPGA instead of CPU in the data center. So what exactly is it? Why use it? Compared with CPU, GPU, and ASIC, what are the characteristics of FPGA? FPGA is a chip that can reconfigure circuits and is a hardware reconfigurable architecture. Through programming, users can change its application scenarios at any time, and it can simulate various parallel operations of hardware such as CPU and GPU. By interconnecting with the high-speed interface of the target hardware, the FPGA can complete the low-efficiency part of the target hardware, thereby achieving acceleration at the system level. What Is an FPGA? Catalog Introduction Ⅰ FPGA vs CPU vs GPU vs ASIC Ⅱ Five Advantages of FPGA 2.1 Performance 2.2 Time-to-Market 2.3 Cost 2.4 Stability 2.5 Long-Term Maintenance Ⅲ New Applications of FPGA Ⅳ Development Trend of FPGA Ⅴ FAQ Ⅰ FPGA vs CPU vs GPU vs ASIC The core difference between FPGA and CPU, GPU, ASIC chips, etc. is that the connection and logic layout of the underlying operation unit are not solidified. Users can program the logic unit and switch array through EDA software to configure the function, so as to realize the integration of specific functions.FPGA appears as a semi-custom circuit in the field of application-specific integrated circuits (ASIC), which not only solves the shortcomings of custom circuits, but also improves the limited number of original programmable device gate circuits. Compared with ASIC chips, an important feature of FPGA is its programmable characteristics, that is, the user can specify the FPGA to realize a specific digital circuit through the program. Furthermore, FPGA chips are one of the best choices for small batch systems to improve system integration and reliability. Figure 1. FPGA Basic Structure So why is FPGA so fast? This is all because the computer's CPU(central processing unit) and GPU(graphics processing unit) belong to the von Neumann structure, with instruction decoding and execution, and shared memory. FPGAs, on the other hand, are instruction-free and memory-free architectures that make FPGA chips much more energy-efficient than CPUs or even GPUs. Figure 2. Von Neumann Structure In the von Neumann architecture, since the execution unit (such as the CPU core) may execute any instruction, so an instruction memory, a decoder, an operator of various instructions, and branch and jump processing logic are required. Due to the complex control logic of the instruction stream, it is impossible to have too many independent instruction streams. Therefore, the GPU uses SIMD (single instruction, multiple data) to allow multiple execution units to process different data at the same pace, and the CPU also supports SIMD instruction. The function of each logic unit of the FPGA has been determined during reprogramming, and no instructions are required. Figure 3. Computer CPU If the GPU is used for acceleration, in order to fully utilize the GPU computing, the batch size cannot be too small, and the delay will be on the order of milliseconds. Using FPGA to accelerate, only microsecond-level PCle delay is required. Why is FPGA so much lower latency than GPU? This is basically an architectural difference. FPGAs have both pipeline parallelism and data parallelism, while GPUs have almost only data parallelism (with limited pipeline depth).For example, FPGA chips can change the running hardware design on the chip every few seconds, while chips such as CPU and ASIC are already solidified when they leave the factory and cannot be changed. If ASIC, CPU, GPU, etc. are built buildings, and the routes of rooms, corridors, and stairs in the building have been fixed, while the interior of FPGA is similar to the magic staircase in Hogwarts, which can change the route of room to room at any time. In addition, FPGA does not need to compile the instruction system at the software application level like CPU and GPU. To program FPGA, use hardware description language, and directly compile and burn it into a combination of transistor circuits, that is, directly use transistor circuits to implement user algorithms.The biggest feature of FPGA is its flexibility. It can realize any digital circuit you want and can customize various circuits. Reduce the shackles of special chips, truly tailor-made for your own products, you can flexibly change the design during the design process, and have field programmability, so it is especially suitable for applications that require continuous changes in physical operation logic, such as AI algorithm optimization, data center applications, etc. Architecture Throughput（int ops） Delay Flexibility CPU ~1T N/A Very High GPU ~10T ~1ms High FPGA(Stratix V) ~1T ~1us High FPGA(Stratix 10) ~10T ~1us High ASIC ~10T ~1us Low The FPGA is set up by the RAM stored on the chip to reset its working state, so the on-chip RAM needs to be programmed when working. Users can use different programming methods according to different configuration modes, which can be said to be very flexible and convenient. The FPGA has the following configuration modes:🔺Parallel Mode: Parallel PROM, Flash configures FPGA.🔺Master-Slave Mode: One PROM configures multiple FPGAs.🔺Serial Mode: Serial PROM configures FPGA.🔺Peripheral Mode: The FPGA is used as a peripheral of the microprocessor and programmed by the microprocessor. Computational performance compared with CPU: For example, Stratix series FPGAs perform integer multiplication operations, and their performance is equivalent to that of a 20-core CPU, and for floating-point multiplication operations, their performance is equivalent to an 8-core CPU.Computational performance compared with GPU: FPGA performs integer multiplication and floating-point multiplication operations. There is an order of magnitude difference in performance compared to GPU. The computing performance of GPU can be approached by configuring multipliers and floating-point operation components. Figure 4. CPU and GPU Architecture Diagram The core advantage of FPGA for performing computation-intensive tasks: tasks such as search engine sorting and image processing have strict requirements on the return time limit of results, and it is necessary to reduce the delay of computing steps. Under the traditional GPU acceleration scheme, the data packet size is large, and the delay can reach the millisecond level. Under the FPGA acceleration scheme, the PCIe latency can be reduced to the microsecond level. Driven by long-term technology, the data transmission delay between CPU and FPGA can be reduced to less than 100 nanoseconds.The FPGA can build the same number of pipelines (pipeline parallel structure) for the number of data packet steps, and the data packets can be output immediately after being processed by multiple pipelines. The GPU data parallel mode relies on different data units to process different data packets, and the data units need to be input and output consistently. For stream computing tasks, the FPGA pipeline parallel structure has a natural advantage in latency. FPGA is used to process communication-intensive tasks and is not limited by network cards. It outperforms CPU solutions in terms of packet throughput and delay, and has strong delay stability. Therefore, FPGAs have obvious advantages over CPUs when performing large data processing tasks with high repetition rates.By programming the FPGA, the user can change the internal connection structure of the chip at any time to realize any logic function. Especially in industries with immature technical standards or rapid development and change, FPGA can effectively help enterprises reduce investment risks and sunk costs, and is a functional and economical choice. Figure 5. Computer GPU With the evolution of intelligent market demand, highly customized chips (ASIC SoC) have led to a sharp increase in market risks due to the large scale of non-repetitive investment and long R&D cycle. Relatively speaking, FPGA has advantages in the field of parallel computing tasks, and can replace some ASICs in the field of high performance and multi-channel. The demand for multi-channel computing tasks in the field of artificial intelligence (AI) drives the evolution of FPGA technology to the mainstream. Figure 6. ASIC SoC Ⅱ Five Advantages of FPGA 2.1 Performance Taking advantage of hardware parallelism, FPGAs break the sequential execution model and complete more processing tasks per clock cycle, surpassing the computing power of digital signal processors (DSPs). BDTI(Big Data Test Infrastructure), a well-known analysis and benchmarking company, has published benchmarks that show that in some applications, FPGAs can handle many times more processing power per dollar than DSP solutions. Controlling input and output (I/O) at the hardware level provides faster response times and specialized functionality to meet application needs. 2.2 Time-to-Market Despite increasing time-to-market constraints, FPGA technology offers flexibility and the ability to rapidly prototype. Users can test an idea or concept and complete verification in hardware without going through the lengthy manufacturing process of custom ASIC design. This allows users to make incremental modifications and iterate FPGA designs in hours, saving weeks. Commercial off-the-shelf (COTS) hardware provides different types of I/O connected to user-programmable FPGA chips. The increasing popularity of high-level software tools reduces the learning curve and abstraction layers, and often provides useful IP cores (pre-built functions) for advanced control and signal processing. 2.3 Cost The non-recurring engineering (NRE) cost of custom ASIC design far exceeds the cost of FPGA-based hardware solutions. The huge initial investment in ASIC design shows that OEMs need to ship thousands of chips each year, but more end users need custom hardware capabilities that enable the development of tens to hundreds of systems. The nature of programmable chips means that users can save on manufacturing costs as well as long lead times for assembly. System requirements change from time to time, but the cost of changing the FPGA design is negligible compared to ASCI's huge expense. 2.4 Stability Software tools provide the programming environment, and FPGA circuits are the real "hard" implementation of programming. Processor-based systems often contain multiple layers of abstraction that can schedule tasks and share resources among multiple processes. The driver layer controls hardware resources, while the operating system manages memory and processor bandwidth. For any given processor core, only one instruction can be executed at a time, and processor-based systems face the risk of tightly time-bound tasks taking over each other at all times. FPGAs, on the other hand, do not use an operating system, and have true parallel execution and deterministic hardware that focuses on each task, reducing the chance of stability issues. 2.5 Long-Term Maintenance As mentioned above, FPGA chips are field-upgradable without the time and expense involved in redesigning ASICs. For example, digital communication protocols contain specifications that can change over time, and ASIC-based interfaces can create maintenance and forward compatibility difficulties. Reconfigurable FPGA chips can accommodate future modifications. As a product or system matures, users can enhance functionality without spending time redesigning hardware or modifying board layouts. Ⅲ New Applications of FPGA At present, the FPGAs mainly produced by Xilinx and Altera with the highest market share, which are all based on SRAM technology, and need to be connected to an external memory to save the program when in use. When powered on, the FPGA reads the data in the external memory into the on-chip RAM, and after completing the configuration, it enters the working state. When power off, the FPGA returns to a white chip, and the internal logic disappears. In this way, the FPGA can not only be used repeatedly, but also does not require a special programmer, but only a general EPROM and PROM programmer. So Actel, QuickLogic and other companies also provide FPGAs with anti-fuse technology, which can only be downloaded once. They have the advantages of anti-radiation, high & low temperature resistance, low power consumption and fast speed. They are widely used in military and aerospace fields. FPGA cannot be erased and written repeatedly, which is troublesome and expensive in the early stage of development. Lattice is the inventor of ISP technology, which has certain characteristics in small-scale PLD applications. Early Xilinx products generally did not involve military and aerospace markets, but now a number of products such as Q Pro-R have entered such fields.In the industrial field, FPGA chips are widely used in the industrial field, and are widely used in video processing, image processing, CNC machine tools and other fields to realize signal control and operation acceleration functions. With the development of intelligence and automation technology, the industrial field is gradually shifting from human resources as the core element to intelligent unmanned factories with automation as the core element.Smart electric vehicles will be the mainstream development direction of the automotive industry in the future. At present, the application of FPGA in automotive cameras and sensors is relatively mature. In the artificial intelligence system of automatic/intelligent driving vehicles, the applicability of FPGA will be the most suitable for processing sophisticated ADAS and autonomous driving. Figure 7. FPGA for Auto In the field of automotive electronic system interface and control, FPGA chips are used to control and drive electric vehicle motor control systems, connect various in-vehicle equipment such as driving systems, instrument panels, radar, ultrasonic sensors, etc. control. In the field of video bridging and fusion, FPGA chips can be used to realize functions such as signal bridging of multiple image sensors, 3D surround view video fusion, reversing auxiliary video, and assisted driving video.In the field of communication, the number of 5G base stations has increased, and the FPGA usage of a single base station has increased, driving the increase in FPGA demand. According to estimates, the FPGA consumption of a 5G single base station is expected to increase from 1-3 blocks in the 4G period to 4-5 blocks in the 5G period. Figure 8. RFSoC FPGA Board Target 5G eFPGA technology is superior to traditional FPGA solutions in terms of performance, cost, power consumption, profitability, etc., and can provide flexible solutions for different application scenarios and different market segments. The economic trend of increasing design complexity and falling equipment costs has stimulated the market demand for eFPGA technology. Ⅳ Development Trend of FPGA First of all, with the commercialization of the new generation of communication technology, the demand for products such as communication base stations, servers, and intelligent terminals will further expand, thereby driving the increase in the market demand for FPGA chips. At the same time, smart cities, smart factories, and consumer electronics pay more attention to the functionality of various smart IoT devices, which will drive the wide application of FPGA chips in smart IoT devices. With the development of the Internet of Vehicles technology, the scale of the use of FPGA chips in the automotive industry will increase day by day to build a more complete Internet of Vehicles and realize smarter autonomous driving functions. Therefore, with the rapid penetration of 5G, the vigorous development of AI and the increasing trend of automotive intelligence, it is expected that the demand for FPGAs in the three fields of communication, AI and automotive electronics will continue to increase in the future, which will also promote The FPGA industry continues to grow. Ⅴ FAQ 1. What is FPGA and why it is used?The acronym FPGA stands for Field Programmable Gate Array. It is an integrated circuit that can be programmed by a user for a specific use after it has been manufactured. ... These blocks create a physical array of logic gates that can be customized to perform specific computing tasks. 2. Is FPGA faster than GPU?The difference between GPU and FPGA performance is not a static factor, but it does depend on the size of the data set. A study by Sanaullah and Herbordt [7] revealed that FPGA can compute small samples of 3D FFT tens of times faster than GPU. The difference is less clear when the data set gets bigger. 3. Is FPGA faster than CPU?A FPGA can hit the data cell faster and more often than a CPU can do it meaning the FPGA causes more results to occur during an attack. It all goes faster when an FPGA is used. And as a side benefit, no trace of all this is left on the CPU because it's never touched when an FPGA is used. 4. Are FPGAs efficient?Efficiency and Power: FPGAs are well-known for their power efficiency. A research project done by Microsoft on an image classification project showed that Arria 10 FPGA performs almost 10 times better in power consumption. 5. Is FPGA programming hard?FPGA vendors have touted their wares as ideal replacements for DSPs, CPUs, and GPUs – even for all of them in a single device – but they are notoriously difficult for software engineers to program as they are not anything like a conventional processor. 6. What can you do with FPGAs?Uses for FPGAs cover a wide range of areas—from equipment for video and imaging, to circuitry for computer, auto, aerospace, and military applications, in addition to electronics for specialized processing and more. 7. What is the difference between processor and FPGA?Microprocessor vs FPGA: A microprocessor is a simplified CPU or Central Processing Unit. ... An FPGA doesn't have any hardwired logic blocks because that would defeat the field programmable aspect of it. An FPGA is laid out like a net with each junction containing a switch that the user can make or break. 8. What language is used to program FPGA?VerilogTraditionally, FPGAs are programmed using pro-level hardware-description languages such as Verilog or VHDL. 9. How many times can you program an FPGA?There is effectively no limit to the number of times a device can be reconfigured; the configuration is stored in SRAM, which has no write limit. most Fpgas can be passively loaded from a processor, one word at a time. That processor can get the FPGA image from anywhere. 10. What are the advantages of FPGA?FPGA advantagesLong-term availabilityUpdating and adaptation at the customerVery short time-to-marketFast and efficient systemsAcceleration of softwareReal-time applicationsMassively parallel data processing 11. How do you make an FPGA?FPGA design checklistMake sure you have plenty of time to spare.Find a decent computer.If you can afford it, add a big display.Decide which operating system to use.Consider using a virtual machine (VM).Select an FPGA vendor.Pick out a suitable development board.Select an embedded processor to use. 12. What is FPGA for beginners?FPGA stands for Field Programmable Gate Array. As you may already know, FPGA essentially is a huge array of gates that can be programmed and reconfigured any time anywhere. Huge array of gates is an oversimplified description of FPGA. FPGA is indeed much more complex than a simple array of gates. 13. What is FPGA in Verilog?FPGAs are nothing, but reconfigurable logic blocks and interconnects can be programmed by Hardware Description Language like Verilog/ VHDL to perform a specific functionality. 14. Do we need to program the FPGA once powered off?If you have a SRAM-based FPGA, like the Spartan 3, then you have to program it each time it is powered up. The reason for this is that the SRAM which stores the configuration is volatile and loses the programmed configuration after power is switched off. 15. How is FPGA different from microcontroller?One of the main differences between a microcontroller and an FPGA is that an FPGA doesn't have a fixed hardware structure, while a microcontroller does. While FPGAs include fixed logic cells, these, along with the interconnects, can be programmed in parallel by using HDL coding language.
Ivy On 2022-01-26
1309
info@kynix.com
00852-6915 1330
