Academic
Publications
GPU Versus FPGA for High Productivity Computing

GPU Versus FPGA for High Productivity Computing,10.1109/FPL.2010.32,David Huw Jones,Adam Powell,Christos-Savvas Bouganis,Peter Y. K. Cheung

GPU Versus FPGA for High Productivity Computing   (Citations: 2)
BibTex | RIS | RefWorks Download
Heterogeneous or co-processor architectures are becoming an important component of high productivity computing systems (HPCS). In this work the performance of a GPU based HPCS is compared with the performance of a commercially available FPGA based HPC. Contrary to previous approaches that focussed on specific examples, a broader analysis is performed by considering processes at an architectural level. A set of benchmarks is employed that use different process architectures in order to exploit the benefits of each technology. These include the asynchronous pipelines common to "map" tasks, a partially synchronous tree common to "reduce" tasks and a fully synchronous, fully connected mesh. We show that the GPU is more productive than the FPGA architecture for most of the benchmarks and conclude that FPGA-based HPCS is being marginalised by GPUs.
Conference: Field-Programmable Logic and Applications - FPL , pp. 119-124, 2010
Cumulative Annual
View Publication
The following links allow you to view full publications. These links are maintained by other sources not affiliated with Microsoft Academic Search.
    • ...Some tradeoffs have been explored in [8] and [9]...

    Daniel Llamoccaet al. Separable FIR Filtering in FPGA and GPU Implementations: Energy, Perfo...

    • ...Prior work on comparing FPGAs and GPUs for high productivity computing used a set of non-standard benchmarks that target different process architectures [6] such as asynchronous pipeline and partially synchronous tree...
    • ...Using the benchmarks in [6] results in analysis that considers processes at architectural level...
    • ...Our benchmark results lead us to a different conclusion from the one reported in [6] about HPRC systems being marginalised by GPUs...
    • ...In contrast to previous work [6], we select a number of established benchmarks based on the HPC Challenge (HPCC) Benchmark Suite [8] and the Berkeley technical report on Parallel Computing Research [9] to measure the relative speedup for each platform...

    Brahim Betkaouiet al. Comparing performance and energy efficiency of FPGAs and GPUs for high...

Sort by: