From EdacWiki
Jump to: navigation, search

If you're wondering WhyAmIgettingPciErrors, rule out mechanical stress on the PCI connectors.

Mechanical stress can skew the electrical signals on the PCI bus by increasing contact force on some of the lines while reducing it on the others.

Your chassis may have been poorly manufactured or damaged in shipping or installation, such that the slot ends are offset from their normal position. If you need to push the card sideways in order to lock it into position on the back of the chassis, try releasing the lock and if the card is heavy, orient the chassis so the card sits vertically in the slot.

Another possibility with riser cards is that the retention bracket may be stressing the riser to mainboard connector. This was the case with certain steel brackets included in some Supermicro 1RU server chassis. The brackets were slightly curved from the process of rolling the metal edges, exerting a slight sideways force on the PCI connector. This happened in many Beowulf cluster nodes where originally the Myrinet cards were suspected of having high failure rates and many were replaced, but eventually it was discovered that removing the metal riser retention bracket reliably stopped the errors and the few of the failures were due to the cards.

Personal tools