General Bandwidth Guidelines; A Word About System Architecture And Performance - Matrox Mura IPX Series System Builder's Manual

Display wall
Hide thumbs Also See for Mura IPX Series:
Table of Contents

Advertisement

Now consider the example given below:
Assuming the transfer of standard HD streams (~500 MB/s), each Mura-IPX card is capable of transmitting its four input streams towards its
upstream PCIe switch. In a multi-level PCIe switch architecture, however, transferring streams from multiple cards through the same fabric
can cause bottlenecks resulting in stuttered playback and dropped frames. Consider the example above: The data path between input cards
#2 and #4 provide a total of 8 GB/s bandwidth towards system memory, and with each card requiring approximately 2 GB/s of bandwidth,
there is ample capacity to accommodate the data being transferred. However, if we consider the addition of capture cards #5-8, we see that if
we want to transfer 16 HD streams in addition to the four initial streams, we require a total of 12 GB/s of bandwidth. Depending on the
bandwidth available between the top level PCIe switch and the host (PCIe switch #7), a bottleneck causing reduced performance could be
present.

General bandwidth guidelines

It is virtually impossible to provide general guidelines for the installation of Mura cards in a PCIe-based system as there are many different
motherboards, and each client's Display Wall implementation is unique. Knowledge of the system architecture and the number and types of
inputs is required to optimally place capture cards in the system. By carefully calculating the required bandwidth and ensuring that no data
bottlenecks are present at any point in the system, the integrator can guarantee the optimal functioning of the Mura-based Display Wall.

A word about system architecture and performance

One factor that should be taken into account when using Matrox D-Series and Mura-IPX is that in order to improve performance, transfers
are performed using system memory, rather than peer-to-peer transfers. In other words, transferring graphical or video data from a capture
card to a display source involves first transferring the data to system memory and then from system memory to the display adapter. This is
done to address performance limitations imposed by the combination of capture and graphics cards.
84
Matrox Display Wall – System Builder's Guide

Advertisement

Table of Contents
loading

This manual is also suitable for:

Mura d-series

Table of Contents