Advanced Computing in the Age of AI | Tuesday, August 16, 2022

FPGAs in the Financial Services Industry 
Sponsored Content by Dell EMC

Field-programmable gate arrays (FPGAs) are gaining traction in the financial services field. They’re silicon chips, but they can be programmed as if they were software. Using prebuilt logic blocks and programmable routing resources, FPGAs are configurable through software. They offer financial services and FinTech firms a number of advantages, including performance, energy savings and time to market.

Years ago, FPGA technology was utilized only by IT digital hardware experts. Now, they have evolved to the point where they are replacing custom Application Specific Integrated Circuits (ASICs) used for signal processing and control applications. Industry research projects FPGAs will be a $7.2 billion market by 2022.[1] Dell EMC is at the forefront of FPGA innovation.

What are FGPAs?

FGPAs are programmable, configurable semiconductor devices used for customizing and optimizing hardware in the field. They can be programmed to perform a wide range of functional requirements after manufacturing. FPGAs can be employed in a variety of use cases, literally becoming any digital circuit that’s needed.

Being programmable—and re programmable—distinguishes them from ASICS, which are custom manufactured for a particular task. While FGPAs have some limitations compared to ASICs, they have more capabilities within the same power constraints. In some use cases, such as highly parallel processing and high-throughput workloads, FGPAs can actually be superior to ASIC microprocessors.

Why FGPAs are Growing in Popularity

Most financial services firms face the IT challenge of needing to add volumes of dense, highly performing compute while facing limits on space and energy. Performance is essential for high frequency trading run on High Performance Compute (HPC) solutions. The trading systems, which must be located near the trading floor, confer financial advantage partly based on the ability to execute trades at lightning speed.

Tight on Space and Power

Financial firms often operate out of high rent districts like the London’s “The City,” New York’s Wall Street or Chicago’s Loop. Floor space for data centers is limited and costly. In some cases, there is simply no more electricity available to run bigger and faster HPC systems even if there is space for them. This scenario drives the appeal of FPGAs in finance.

Moore’s law has expanded FPGA processing power to the size where they can be applied to key algorithms. Programmed the right way for finance, an FPGA can outperform even the fastest CPU- and GPU-based HPC solutions. Indeed, as CPU performance growth slows in the industry, FPGA is on the way up. From performance comes profits. And, given the energy constraints of some financial firms, the ability to operate high frequency trading systems in limited space with less energy use makes FPGAs a clear winner.

Cost and Time to Market

A sophisticated FinTech firm can get to market more quickly with an FPGA than with an ASIC. The flexibility and rapid prototyping capabilities make it possible to move product development along a lot faster. ASICs are comparatively slow and expensive to design and produce. The process includes steps like VHDL/Verilog, layout), mask generation and foundry.

FGPAs allow for more rapid concept testing, hardware verification and iterative design cycles. They do not require an ASIC’s large production runs to justify the upfront investment in design. The cost of an FPGA generally consists of the unit price from its manufacturer and the time investment of custom programming it.

Long-term maintenance is another area where FGPAs present a potential advantage over alternatives. FPGA chips are field-upgradable. Engineers can upgrade them off-site without swapping out a chip. So, they don’t require the kind of expense and delay associated with redesigning an ASIC. This can be an issue, for instance, when digital communication protocols evolve. A reconfigurable FPGA can keep up with modifications. In addition, FGPAs are benefiting from a growing availability of FPGA-oriented software tools and off-the-shelf I/O hardware built to connect with FPGAs.

Technical Drivers of FPGA Popularity

The core driver of FPGA technology is its parallel performance. FGPAs are based on a matrix of configurable logic blocks (CLBs) connected through programmable interconnects. Reprogrammable silicon is as flexible as software that runs on a processor-based system except it is not constrained by the amount of processing cores at hand. Different processing operations will not need to compete for necessary resources. Each dedicated processing task will be assigned to a specific section of the chip and function on its own, without any conflict or overlap influence from existing logic blocks. The independent performance of one part of an application will not be affected at all when adding more processing.

In a financial services use case, that kind of preemption might block a time sensitive transaction and cost the firm money. FPGAs do not use an OS. As a result, they improve reliability with true parallel execution. They enable deterministic hardware dedication for all tasks.

Software tool are used for the FPGA programming environment. However, the FPGA is a hard implementation of the programming. This presents several advantages over processor-based systems. Processors are usually composed of various abstraction layers used to schedule tasks and share resources among different processes. There is a driver layer to control hardware resources. An OS manages memory and processor bandwidth. Only one instruction can be executed at any one time per processor core.

FPGAs can be adapted to be used with the latest technologies, transforming C code or graphical block diagrams into digital hardware circuitry. The flexibility to customize FPGAs gives financial services IT professionals more choices and autonomy in applications. FPGAs exist in a one-time programmable (OTP) format; however, the prevailing types are re-programmable SRAM-based. Engineers are looking to FPGAs users to customize microprocessors to fulfill specific needs. Professionals who design integrated circuits are eagerly using FPGAs as they have greater scope and potential in contrast to programmable read-only memory (PROM) chips.

Nallatech and FGPAs

Nallatech is putting FGPAs to work across its portfolio of server solutions with CPU/GPU-SIMD/FPGA-MIMD accelerators and memory. Two certified FGPAs from Nallatech are the 385A and 510T FPGA OpenCL accelerators. The 385A accelerator is certified to run in Dell EMC PowerEdge R630 and R730 servers for network-enabled applications. The 510T accelerator is certified for PowerEdge C4130 and R730 servers for compute-intensive applications.

Coming soon, Nallatech expects to introduce platform integration with a new generation of Intel® Programmable Acceleration Cards starting with the Intel® Arria® 10 GX FPGA enabled by the acceleration stack. This enables Nallatech to offer Intel Xeon processor-based server acceleration solutions.



Add a Comment