FPGAs can be used to accelerate the most time-critical network functions.
One thing guaranteed in the financial sector is change. And when it comes to regulatory change, that means change with a capital C. But in order to take advantage of this change, businesses need to be ready to respond – and quickly.
So the question you need to ask is: “Does your business have the tools at its disposal to react quicker and more effectively than ever before?” Because it needs to.
Step forward a potential game-changer in the form of FPGAs. But what are these mysterious wonders?
Well, FPGAs – or, to give them their full title – are silicon devices that can be dynamically reprogrammed with a hardware design and data path that exactly matches a user workload.
Here’s everything you need to know.
What’s in a Name?
The second half of the name, Gate Array, describes the logic gates, the simple building blocks of every digital circuit. When placed together in rows and columns they form an array. Even the most complex processor can be broken down into blocks of logic gates, simple ANDs and ORs connected together to form shifters, adders, multipliers, flip-flops, registers, and accumulators. This granularity is often lost in today’s large-scale integration.
The “Programmable” part of the name indicates that the way in which the logic gates are interconnected isn’t predefined. What this means is that there are no predefined configurations for an FPGA, when it powers up it is essentially a sea of uncommitted logic. The configuration is handled by design files, normally stored off-chip and loaded at power-up. These files tell the FPGA how to interconnect the logic to create larger functions.
The “Field” part refers to the fact that the configuration takes place at power-up, every time power is cycled. When an FPGA starts up it has the potential to be configured in any number of ways, and in fact, it doesn’t even need a power cycle; many FPGAs can now be reconfigured in real-time while parts of the device are still running. This ability to perform virtually any task, using hardware designed specifically for that task, is where the value of FPGAs in high performance applications like data centres can really be felt.
Working with FPGAs
In very basic terms, systems that are defined purely by their software are constrained by their hardware. This is the case with MPUs (Memory Protection Units); whatever they do in software is always executed on the same underlying hardware, the abstraction between the two is handled by the compiler.
With an FPGA, the task – normally defined in software – doesn’t need to conform to the hardware architecture, in fact the opposite is true; the hardware conforms to the task, which no longer has to be defined by software, at least not in the conventional sense.
To put it simply, if you describe a multiplier in FPGA software, it gets implemented as a multiplier in hardware. In even more basic terms, every line of code in FPGA software runs at the same time; architectures don’t come any more parallel than that.
The language used to define the hardware structure of an FPGA is generally referred to as a Hardware Description Language, or HDL. The process of compiling an HDL is termed synthesis and the design flow for an FPGA is similar to that of an ASIC (Application-Specific Integrated Circuit).
As such, one of the key stages of design involves verifying that the hardware meets the specification. As this is real hardware, running concurrently, one of the most important parameters to check is timing. Timing closure can be one of the most arduous but crucially important phases of design. Fortunately, there are highly advanced EDA tools at hand to ease the entire design cycle.
FPGAs in Action
One of the simplest ways to take advantage of FPGAs in a real-world application, like a data centre, is to use an Intel® FPGA Programmable Acceleration Card (Intel® FPGA PAC). Intel offers PACs based on the Arria® 10 GX FPGA and the Stratix® 10 SX FPGA. One example is the Intel® FPGA PAC D5005, which can be implemented in many market segments, such as financial and analytics.
So, Why Should You Care?
FPGAs have evolved over several decades and have always been at the forefront of performance. As the most powerful form of standard IC available to developers, FPGAs can be used to accelerate the most time-critical network functions.
For financial services businesses, there’s a whole host of benefits. Right at the top of the list are risk analytics and high-frequency trading and speed is of the essence in both – whether you’re looking to manage the risk or make the best trading decision. The fact that data can be decoded in tens of nanoseconds puts your business one step ahead of the game. With other key benefits such as ultra-low latency trading, algorithm trading and several use cases in data analytics, FPGAs have a massive role to play.
Intel is dedicated to delivering higher performance in all its forms and FPGAs are an important part of that strategy. By supplying the very best hardware platforms, complemented by the industry’s most advanced EDA tools for FPGA development, businesses have the potential to achieve the very best performance.
This versatility enables the provision of fast processing power efficiency, and low-latency service – which could lower the total cost of ownership and maximise compute capacity within the power, space, and cooling constraints of data centres.
To download Intel’s “FPGAs for Dummies” eBook just click here.
For More Information: