In most device-accelerated systems, the maximum achievable throughput is limited by the PCIe® bus. PCIe performance is influenced by many different aspects, such as motherboard, drivers, target platform, and transfer sizes. Run DMA tests upfront to measure the effective throughput of PCIe transfers and thereby determine the upper bound of the acceleration potential, such as the xbutil dma test.
Figure 1. Sample Result of dmatest on an Alveo
U200 Data Center Accelerator Card
An acceleration goal that exceeds this upper bound throughput cannot be met as the system is I/O bound. Similarly, when defining kernel performance and I/O requirements, keep this upper bound in mind.