z820 928-Fatal PCIe error

[AlfoldTV]

We have a HP Z820 workstation that displays this error multiple times every day: "928 fatal PCIe error".

We've already updated the BIOS to 3.91 Rev.A but it hasn't solved the problem.

We've also tried downgrading the BIOS to 3.88 Rev.A (sp71580.exe) but at the end of the flashing process it gives an error and won't finish. Also tried to do this via USB stick, in the BIOS, but no success.

Is it possible to downgrade the BIOS of a HP Z820? If it's possible, then how? 

[BambiBoomZ]

AlfordTV,

I think this error is due to an incorrect PCIe version setting on a particular PCIe slot.  For example, if you are using a PCIe 3.0 GPU or device on a PCIe 2.0 lot, that slot needs to be reset to PCIe 2.0.   

Check the z820 slot specifications by slot number and see if any slot is populated with a card that does not match the slot speed configuration and then modify the setting for that slot to match the card.

Cheers,

BambiBoomZ

[Cmcdona]

I have the same issue (and it can take a couple of reboots to get it back).

Latest BIOS Nvidia with a GTX 1070  (PCIe 3.0 x16 card) in slot 2 (which is the slot with the 928 error)

[BambiBoomZ]

Cmcdona,

What are the BIOS settings for Slot 2?

I've seen a number of posts that suggest fairly common issues with older series systems when using a Pascal GPU.  These are always followed by the suggestion to update the BIOS and chipset driver to the latest.

This is probably not relevant to the issue, but the listing of the PCIe slot configuration in the z820 specs is a bit odd:

2 PCI Express Gen3 x16 slots

1 PCI Express Gen3 x16 slot- Available ONLY when 2nd CPU is installed.

1 PCI Express Gen3 x8 slot- with x16 connector- Available ONLY when 2nd CPU is installed.

1 PCI Express Gen3 x4 slot- with x8 connector

1 PCI Express Gen2 x4 slot- with x8 connector

1 PCI 32bit/33MHz slot

1 Mechanical- oly slot, supporting cards which mount only to the I/O bulkhead and not the motherboard (half length, full height)

The PCIe x8 connectors are open ended, allowing a PCIe x16 card to be seated in the slot.

As each Xeon E5 supports 40 PCIe lanes in the z820, it seems to me that two GPU's could run at x16 with a single CPU.

Cheers,

BambiBoom

[stroese]

Hi!

We are experiencing the same problem. The HP-Z820 workstation reboots when the PCIe link goes down. I've tried changing the PCIe link speed in the BIOS to Gen 2 (our PCIe target is Gen 2) but without any change. I also tried downgrading the BIOS, but only v3.90 can be installed, which has the same issue. v3.8x can't be installed any more.

This reboot upon PCIe link down is new with using a more recent BIOS version, btw. Since few weeks ago, we have been using a much earlier BIOS version (can't remeber the exact version, but very likely something <= v3.8x). And with this BIOS version no such reboot happed upon surprise PCIe link down.

Did anyone of you get this issue fixed somehow? Does anyone have some other ideas what to test or how to downgrade to an earlier BIOS version?

Many thanks in advance,

Stefan

[Z440Roger]

Stefan, i would have set the BIOS to Automatic for the PCI link speed and then read post #3 here: https://h30434.www3.hp.com/t5/Business-PCs-Workstations-and-Point-of-Sale-Systems/928-fatal-pcie-err...

Often reseating the card can be enough. If that not helps, move the card to the second 16x slot if you have two processors.