FPGA Leadership across Multiple Process Nodes
Big Data Hush offers a comprehensive multi-node portfolio to address requirements across a wide set of applications. Whether you are designing a state-of-the art, high-performance networking application requiring the highest capacity, bandwidth, and performance, or looking for a low-cost, small footprint FPGA to take your software-defined technology to the next level, Xilinx All Programmable FPGAs and 3D ICs provide you with system integration while optimising for performance/watt
GPU vs FPGA Performance Comparison
Image processing, Cloud Computing, Wideband
Communications, Big Data, Robotics, High-definition
video…, most emerging technologies are increasingly
requiring processing power capabilities. The technology
selection for each application is a critical decision for
system designers. Being GPU power the conservative
approach to scale processing capacity, using FPGA for
software acceleration is becoming the best option for an
increasing number of applications.
This paper evaluates the 2016’s state-of-the-art
technology for both GPU and FPGA devices, and
performs a qualitative and quantitative comparison. The
analysis must be considered as a preliminary guideline
for technology selection. Some key parameters cannot
be directly compared, and different interpretations of
the results can be derived introducing other variables.
What is important for your design?
The short answer to this question is that FPGAs are
power efficient and GPUs are cost efficient (Figure 1);
but taking a design decision based on simple rule-ofthumbs
is usually risky.
FPGAs are designed to perform concurrent fixed-point
operations with a close-to-hardware programming
approach, while GPUs are optimised for parallel
processing of floating-point operations using thousands
of small cores. Most of the differences between the two
technologies, and their applicability to software
acceleration, are herein derived from these high-level
architectural definitions.
Comparing processing capabilities is not straightforward.
GPUs performance is measured in GFLOPS; they are
capable to accelerate native CPU algorithms based on
floating-point operations, simplifying code adaptation
from high-level programming languages. On the other
hand, FPGAs processing power is measured in GMACS;
they require designing algorithms for fixed-point data
types to maximize efficiency, taking massive benefit of
bit-wise operations.
GPUs gain advantage when considering total floatingpoint
processing power, development effort, device cost,
and flexibility. However, FPGA is starting to be the logical
choice for an increasing number of applications.
FPGA also provides huge processing capabilities with a
great power efficiency, reducing thermal management
and space requirements. This feature allows the
integration of acceleration hardware in small housings,
on-board equipment, or extreme temperature
environments.
Interfaces are another FPGA’s strong point. Being GPUs
limited to PCIe, interfacing with devices implementing
any other standard or custom interfaces will require
additional electronics. FPGA has a huge interface
flexibility, recently improved by the integration of
programmable logic with CPUs and standard peripherals
in SoC devices.
Latency is another parameter to be considered when
running processing software in specialised hardware.
GPUs improve CPUs performance, but FPGA provides
deterministic timing in the order of nanoseconds. This is
especially important for encryption, audio coding,
network synchronisation or control applications that
need to manage small and well-known latencies.
Regarding the price of a software acceleration solution,
GPUs are cost efficient both in development and
hardware installation. FPGAs require specialised design
engineers with knowledge in a number of different
technology areas (electronics, HLD, algorithms,
communications, etc.). The price comparison for midrange
devices is not a drama considering FPGA power
efficiency. The real issue is the engineering effort, which
is being mitigated introducing new development
environments to add abstraction layers to the lowestlevel
design. In addition, autocoding techniques are
starting to reduce implementation times, although they
do not significantly reduce the required know-how.
Finally, RTL-based design enables FPGA to be used as
technology path to ASIC development.
Figure 2 and Table 1 summarise this qualitative analysis
for a faster understanding of the technology trade-offs.
GPU performance in numbers
A selection of 28nm graphic cards and FPGA devices are analysed and used for comparison purposes. GPUs performance is derived from commercial graphic cards characteristics.
Table 2 lists a selection of the best cards for the money as representative samples for the 2016’s state-of-the-art technology combining older models with newer flagship graphic cards.
GPUs price ranges from less than 100€ to more than 600€, showing huge processing powers over 7,000 GFLOPS for single precision operations, having no rival when evaluating floating-point computational capacity of a single chip.
Since manufacturers are only providing the required power supply, maximum consumption are derived from user analysis in stress conditions where top processing performance is achieved. Burning up to 360W for the
high-end models, it demands careful cooling designs (heatsink, fans), heavier power supplies, and users ready-to-pay an increasing electricity bill.
Price efficiency is similar for all the analysed models, ranging from 0.07 to 0.12 €/GFLOPS. High-end graphic cards, however, show improved power efficiency, achieving up to 23 GFLOPS/W. In fact, energy usage is currently the most important constraint to continue increasing graphic cards maximum processing capability, and it is expected to be dramatically improved in the following years.
But as per 2016 state-of-the-art technology, energy burning is the main draw-back of GPUs for software acceleration purposes in a number of applications, and it must be considered together with the relative low price and the huge total processing power. At the end, a high power consumption involves that they cannot be installed in systems with demanding power, space or temperature requirements.
FPGA performance in numbers
FPGA technology is evolving fast, with new models implementing 16nm and 20nm, and increasing clock speeds, interfaces bandwidth, on-chip RAM, and fixed- and floating-point processing capacity. For this analysis, we stick to 28nm device not only for a fair technology trade-off, but also for a reasonable price comparison.
Table 3 shows a selection of Xilinx 7-Series devices, including Zynq and representative FPGA integrated circuits. Zynq SoC combines the flexibility of a CPU with the processing power of the FPGA, lowering the entry barrier for software acceleration using programmable logic. It supports embedded operating systems and standard peripherals, being the perfect candidate for mitigating technology draw-backs and opening the FPGA world to a higher number of sectors, applications and end-users.
SoC allows users to implement in the FPGA only the high computation load algorithms and tasks, with a similar approach to how GPUs interface with CPUs for software acceleration.
As mentioned before, estimating the floating-point processing capacity of a device that is designed for
fixed-point operations and HDL programming is not direct. The result will depend on the implementation approach and the type of algorithm.
In this paper, processing power for each selected device is estimated considering the peak DSP performance in GMACS, and the single precision floating-point performance and power efficiency claimed by Xilinx.
As expected, the results show a decent total processing power, with an excellent power efficiency always higher than 70 GFLOPS/W. This enables the implementation of current FPGAs devices into small and efficient hardware with more than reasonable thermal dissipation and cooling requirements.
An interesting conclusion is derived from analysing price efficiency, which reaches 0.29 €/GFLOPS for mid-class FPGAs as the Artix-7 200T. Although still far from GPUs mass-production cost, it is a comparable cost budget. This is not true for the costly high-end devices of the family, where prices are a clear draw-back for many software acceleration applications.
Table 3 only considers the device price, which is usually the cost driver for FPGA-based solutions.
Comparing Key Performance Indicators
Table 4 mixes together representative graphic cards and FPGA devices for a final comparison of the most important parameters quantified in this paper.
The graphical representation of Figure 3 clearly shows the huge processing power of GPUs, and the great power efficiency of the current FPGA technology.
Introducing the price in the equation shows competitive mid-class FPGA, but still far from the GFLOPS per Euro of the mass-market graphic cards.
The cost of the high-end FPGAs limits them to specific niche applications, while the power burning of the high-end GPUs avoids using them for a number of markets and critical systems.
Conclusion
This paper compares key performance indicators of 28nm GPUs and FPGAs devices, focussed on their applicability to software acceleration purposes. The trade-off analysis is a preliminary guideline for technology selection, a design decision that is a fundamental performance and cost driver.
In addition to processing capacity, power efficiency and cost, other parameters such as latency, development effort, flexibility or interfaces are discussed.
The analysis shows cost efficient GPUs with huge floating-point processing capacity, and power efficient FPGAs with flexible interfaces and deterministic latency. The availability of SoC integrating CPUs with FPGA and
standard peripherals is a step forward mitigating FPGA draw-backs, i.e., high development efforts, and limited flexibility and backward compatibility.
In the near future, we can envisage GPUs with improved power efficiency, and FPGAs with increased computational capacity and lower cost. After that, technology selection will be focussed only on architectural design considerations and specific application requirements, which definitely will be good news for the engineer’s community.
But for the time being, the selection of the device will remain linked to the end-user application, available budget, and development capacity.
