EP45-DS4 - Top PCIe slot not recognized by BIOS?

[mushbert]

Hello and merry Xmas/favoured equivalent holiday. I am rebuilding and expanding an old P45 machine, and have discovered a mysterious problem.

Let's start with the basics:

1. There are 2 PCIe slots (one x16, one x8) attached to the P45. I am using a GPU in the x16 slot and would prefer to avoid using the x8 slot (since it forces the x16 slot to run in x8 mode.)

2. There are 4 PCIe slots (one x4, three x1) connected through the ICH10R; when any x1 slot is in use, the x4 slot changes to x1 speed.

3. One of the x1 slots (PCIEX1_2) is blocked by the GPU cooler, leaving the x4 slot and two x1 slots.

4. I have three x1 cards I would like to use (Intel NIC, Creative X-Fi, Syba ASMedia AS1061-based SATA controller), so there are just enough slots accessible.

5. Only the SATA card will fit in the top slot (PCIEX1_1); due to their length the others are blocked by the P45's heatsink.


To make this easier to visualize, here's the layout diagram from the manual with the approximate sizes of the cards added:




The problem:

The SATA card will not work in the top slot (PCIEX1_1) or the next accessible one (PCIEX1_3.) In those slots, attached drives are not detected, the OROM splash-screen is not displayed, and Windows does not see the card at all.

However. When the card is in the top slot, it does have power and appears to be active. If I connect a 4-disk eSATA enclosure to the card, the enclosure's LED for disk 1 turns on (which indicates the card is using command-based port-multiplier mode like Windows' default drivers do - if FIS mode is enabled then all disk LEDs are lit.) Additionally, the card has an activity LED passthrough, so activity from connected drives are also shown by the case HDD LED - this is working normally, but it could be a passive circuit. The first time I booted Windows with the card in the top position, a "New Hardware Found..." notification starting with "Intel PCI Express" came up, so I believe the slot itself is recognized.


As far as troubleshooting, I can only think of two tests which don't require additional hardware or the removal of the P45 heatsink (which is a pain because of the heatpipes):

- Moving the GPU to the x8 slot so PCIEX1_2 is exposed for testing. This should show whether it's just the top slot, but won't identify if it's a BIOS issue or something else. Depending on which waterblock gets put on the GPU I may have access to this slot later.

- Trying an older BIOS. Undesirable as a solution because it means losing AHCI 1.20e (unless I manage to mod it back in - and I have NOT had good luck with modded BIOS on these boards.)


If anyone has seen this behaviour, knows if the PCIe slots have to be filled in some sort of order for them to work, or has any theories I'd love to hear them... Thanks for reading and all the best to you and yours,

mushymush!

[profJim]

Re: EP45-DS4 - Top PCIe slot not recognized by BIOS?

From page 10 in the EP45-DS4(P) user manual:

• (Note 3) If you are installing one PCI Express graphics card, be sure to install it in the PCIEX16_1 slot for optimum performance. When both of the PCIEX16_1 and PCIEX8_1 slots are installed with expansion cards, the PCIEX16_1 will operate at x8 mode.
  
• (Note 4) The PCI Express x4 slot shares bandwidth with the three PCI Express x1 slots. When it is populated with a x4 card, the three PCI Express x1 slots become unavailable.

Note 4 is somewhat vague and it looks like you have a choice of using the x4 slot or all three x1 slots, but not at the same time.
It looks like using the x8 slot will reduce the upper x16 slot to x8 bandwidth for your video card.

The EP45-DS4 board has bios versions from F5 - F10b.
The EP45-DS4P board has bios versions from F3 - F9a.

[mushbert]

Re: EP45-DS4 - Top PCIe slot not recognized by BIOS?

Note 4 is clearer once you see the Block Diagram from page 8 - sorry I didn't include it but the post was getting pretty long!



Also I just realized the SATA card is not recognized when it's in the 2nd available slot (PCIEX1_3) either and I'll correct this in the OP. The X-Fi card was working in this slot so that may point to a firmware cause. Will do more testing.

Oh and just to confirm, I have both boards but this is the DS4 and not the DS4P.

[profJim]

Re: EP45-DS4 - Top PCIe slot not recognized by BIOS?

I corrected the incorrect image link in your post so that the motherboard block diagram is displayed in your post.

I'm not sure what the x1 pci-e slot behavior/availability is if you install an x1 card in the bottom pciex4_1 x4 slot.

[mushbert]

Re: EP45-DS4 - Top PCIe slot not recognized by BIOS?

If both the diagram and Note 4 are correct then it indicates that the PCIe switch has only two settings - three lanes to the x4 slot, or one lane each to the three x1 slots (the remaining lane to the x4 slot is not switched). This would not surprise me since it's how the x8 and x16 slots work - even though you technically should be able to allocate say, fifteen lanes to the GPU and one lane to the x8 slot, it doesn't allow this. I have also used three PCIe cards in one of the DS4P boards (without using the x8 slot), not that it confirms anything. It is mostly the same board though, apart from the additional "Ultra TPM" circuitry and different BIOS versions.

[mushbert]

Re: EP45-DS4 - Top PCIe slot not recognized by BIOS?

Hello again. I continued testing using multiple firmware versions on both motherboard and SATA card, with no success. However the results confirmed that the presumed method of operation (see last post) is correct.

So:

The ASMedia card will only work in the P45 slots or the x4 ICH10R slot, while the other cards will work in any slot. This round of testing included another SATA card based on JMicron's JMB363, which had no problems in any of the slots. Apart from the obvious, the major difference the JMB363 card shares with the other working ones is that it is NOT a PCIe v2.x card, while the ASMedia one is. Since PCIe v2.x is supposed to be backwards compatible, it is not surprising that the ASMedia card works in the x4 slot (which, being connected through the ICH10R, is only PCIe v1.x.) This seems to indicate the problem must be due either to the way the x1 slots are wired, or an incompatibility between the SATA card and the PCIe switch Gigabyte used.

I hope either Syba or Gigabyte will have a solution (or at least an explanation) for this, but it seems unlikely. So it seems my options are fairly limited, and all the ones that aren't compromises cost money:

1. Use the card in the x8 slot.
2. Use one less drive than needed.
3. Buy a PCIe riser/extension cable to allow one of the longer cards to use the available slot.
4. Buy another SATA card, likely based on the SiI3132 chip since it's the only other one I know of that supports FIS-based switching.
5. Buy another cooling solution for my GPU, which will allow access to the blocked x1 slot.

If anyone has other suggestions I'm all ears. :|

[mushbert]

In which a conclusion was reached

After a long hiatus I finally got around to working on this. The issue was mostly resolved using a SiI3132-based card in the desired slot. I say mostly, because at first there seemed to be an issue where the card was visible but attached drives were not, but this seems to have gone away.

There is at least one other glitch present. With the ICH10R's SATA ports in AHCI or RAID modes and one or more drives attached to them, the PCI-E card's Option ROM displays an error message instead of detecting the attached drives: "Warning:Have Option ROM can not be invoke (Vendor ID:1095h, Deivce ID:3132h)" (sic). However, as this reduces the boot time by a few seconds and the connected drives are available in Windows, this is acceptable for me. It also adds support for optical drives, which the ASM1061 lacks.