S3420GPLX PCI Issues

[idata]

Dear Experts,

I recently constructed a system based on the S3420GPLX motherboard and encountered a couple problems relating to the PCI & PCI-E slots. First, the system configuration:

Current System Configuration:

Intel S3420GPLX MotherboardIntel Xeon X3470 ProcessorIntel SC5650UPNA Pedestal Server CaseKingston KVR1066D3Q8R7SK2/16GI Memory Kit (x2)Point Grey FWB-PCIE-02 Dual Bus 1394b PCI-E Card (x5)HIS H545H512P Radeon HD 5450 PCI Video CardWestern Digital WD1001FALS 1TB Hard Drive

Motherboard Setup:

Slot 1 (32-bit PCI): See Problem # 1, HIS H545H512PSlot 2 (PCI-E Gen1 x1): Point Grey FWB-PCIE-02Slot 3 (PCI-E Gen1 x4): Point Grey FWB-PCIE-02Slot 4 (PCI-E Gen2 x4): Point Grey FWB-PCIE-02Slot 5 (PCI-E Gen2 x8): See Problem # 2, Point Grey FWB-PCIE-02Slot 6 (PCI-E Gen2 x8): Point Grey FWB-PCIE-02 DIMM B1: Populated, 8GBDIMM B2: Populated, 8GBDIMM B3: EmptyDIMM A1: Populated, 8GBDIMM A2: Populated, 8GBDIMM A3: EmptySystem BIOS: R0048 (latest version)BMC Firmware: v01.23 (latest version)FRUSDR Utility: 18 (latest version)

Background: The system is intended to interface with up to 10 high-resolution, high-bandwidth IEEE1394b video cameras from Point Grey Research and perform real-time image analysis. Each video camera is capable of fully utilizing an entire 800Mbps IEEE1394b bus, hence the use of multiple dual-bus IEEE1394b PCI-E adapters. A PCI video card was added due to the poor general performance and limited resolution of the onboard video solution.

Problem # 1: The motherboard will not POST with the HIS H545H512P PCI video card installed. The startup process halts before the display is enabled, preventing access to the BIOS and preventing OS boot. The diagnostic LED code is 0x50 "Enumerating PCI Buses". This happens regardless of whether the monitor is connected to the on-board video or the PCI video card. Removing the PCI video card allows the computer to boot normally; however, the performance of the onboard video is inadequate for the intended application.

This particular PCI video card has be successfully tested in another machine without issue. I have tried disabling the onboard video in the BIOS, as well trying the dual monitor video BIOS option, without success. I also tried using a higher wattage power supply in case the standard 400W was insufficient; this did not work either. Removing all other add-on cards and peripherals does not change anything. Are there any known compatibility issues with the S3420GPLX motherboard and PCI video cards? What are my options / potential workarounds for this issue? Please note that a PCI-E video card cannot be used since all PCI-E slots will eventually be occupied with dual-bus IEEE1394b adapters.

Problem # 2: Slot 5 (PCI-E Gen2 x8) does not appear to be functional on this motherboard and will not detect the installed Point Grey FWB-PCIE-02 card. The system still boots normally in this case, the card is simply not detected. The other four PCI-E slots appear to function normally and will properly detect the Point Grey FWB-PCIE-02 cards. All FWB-PCIE-02 cards have been tested and verified to work correctly in another system. Are there any BIOS settings or configuration changes that could be made to enable use of all 5 PCI-E slots? Is it possible/likely that the motherboard is defective?

Thank you,

Robert

[DSilv11]

WOW!! All the specifics in one posting This is going to take a bit to digest.

At first glance, I don't see anything obvious, but do suspect this could be a much deeper dive that usually occurs in the forums.

Are you working with the local distrubutor / field application team on this?

Well, here goes.

Problem # 1:

Have you tried any other video cards?

I am not aware of any PCI related video card issues, but when I check the Test Hardware & OS list, I see that all the tested cards are PCIe.

You may want to set the Assert NMI on PERR / SERR under Server Management to disable before installing the video card and see if it make any difference. (I don't think it will , but some video cards like to through PERR at power up.)

Problem # 2

Can you tried install just 1 card in SLOT 5?

Can you try a different PCIe (NIC or VIdeo) card in SLOT 5 if you have one?

Deep diving:

With all the cards installed (except the video) run the SIT (System Information Tool) tool from EFI and attach the logs

http://downloadcenter.intel.com/Detail_Desc.aspx?agr=Y&DwnldID=19657&lang=eng&OSVersion=&DownloadType=Utilities%2C%20Tools%20and%20Examples http://downloadcenter.intel.com/Detail_Desc.aspx?agr=Y&DwnldID=19657&lang=eng&OSVersion=&DownloadType=Utilities%2C%20Tools%20and%20Examples

2nd : Do a dump of the PCI configuration space which can also be done in EFI

Boot to the internal EFI shell with a USB key install to collect the data

switch to the USD drive (FS0: most likly)

You should get a prompt FS0:>

Type the following commands which will save and display a list of all the PCI devices in the system

FS0:> PCI >> out.txt

FS0:> PCI -b

Then for each device found dump the full PCI config space using the command FS0:>pci BUS DEV FUNC where BUS DEV FUNC are teh numbers displaied when you used teh PCI-b command.

i.e.

FS0:> PCI 00 00 00 >> out.txt

FS0:> PCI 00 01 00 >> out.txt

FS0:> PCI 00 03 00 >> out.txt

(all the address in the first PCI listing)

FS0:> PCI 09 00 00

Then attach

[idata]

Doc_SilverCreek wrote:

WOW!! All the specifics in one posting This is going to take a bit to digest.

At first glance, I don't see anything obvious, but do suspect this could be a much deeper dive that usually occurs in the forums.

Are you working with the local distrubutor / field application team on this?

Well, here goes.

Problem # 1:

Have you tried any other video cards?

I am not aware of any PCI related video card issues, but when I check the Test Hardware & OS list, I see that all the tested cards are PCIe.

You may want to set the Assert NMI on PERR / SERR under Server Management to disable before installing the video card and see if it make any difference. (I don't think it will , but some video cards like to through PERR at power up.)

Problem # 2

Can you tried install just 1 card in SLOT 5?

Can you try a different PCIe (NIC or VIdeo) card in SLOT 5 if you have one?

Deep diving:

With all the cards installed (except the video) run the SIT (System Information Tool) tool from EFI and attach the logs

http://downloadcenter.intel.com/Detail_Desc.aspx?agr=Y&DwnldID=19657&lang=eng&OSVersion=&DownloadType=Utilities%2C%20Tools%20and%20Examples http://downloadcenter.intel.com/Detail_Desc.aspx?agr=Y&DwnldID=19657&lang=eng&OSVersion=&DownloadType=Utilities%2C%20Tools%20and%20Examples

2nd : Do a dump of the PCI configuration space which can also be done in EFI

Boot to the internal EFI shell with a USB key install to collect the data

switch to the USD drive (FS0: most likly)

You should get a prompt FS0:>

Type the following commands which will save and display a list of all the PCI devices in the system

FS0:> PCI >> out.txt

FS0:> PCI -b

Then for each device found dump the full PCI config space using the command FS0:>pci BUS DEV FUNC where BUS DEV FUNC are teh numbers displaied when you used teh PCI-b command.

i.e.

FS0:> PCI 00 00 00 >> out.txt

FS0:> PCI 00 01 00 >> out.txt

FS0:> PCI 00 03 00 >> out.txt

(all the address in the first PCI listing)

FS0:> PCI 09 00 00

Then attach

Problem 1:

As you suspected, disabling assert NMI on PERR & SERR didn't make a difference. Unfortunately, I currently do not have any other PCI video cards handy. I did however test a PCI network adapter in slot 1 and it appeared to work properly. If you suspect that it may just be a compatibility issue with this particular PCI video card, then I could buy a different one and try it.

Problem 2:

A single dual-bus 1394b card (FWB-PCIE-02) in slot 5 does not appear to be detected. Other PCI-E cards do appear to work however. Interestingly, this included a single-bus version of the Point Grey 1394b card (FWB-PCIE-01). Because of this, I decided to capture logs for several different hardware configurations to try and narrow down the cause. Configurations for two of the tests:

Test 1:

Slot 6: FWB-PCIE-02Slot 5: FWB-PCIE-02Slot 4: FWB-PCIE-02Slot 3: FWB-PCIE-02Slot 2: FWB-PCIE-01Note: Card in slot 5 does not appear to be detected at all. All other cards are detected and seem to work as expected.

Test 2:

Slot 6: FWB-PCIE-02Slot 5: FWB-PCIE-01Slot 4: FWB-PCIE-02Slot 3: FWB-PCIE-02Slot 2: FWB-PCIE-02All cards appear to be detected and function correctly.

I have attached the logs from the above tests.

The main difference that I notice between the FWB-PCIE-01 and the FWB-PCIE-02 cards is that the FWB-PCIE-02 includes several PCI/PCI bridge devices in addition to the Serial Bus Controllers. Is it possible that these PCI/PCI bridges are causing some kind of compatibility issue that is limited to slot 5?

Thanks,

Robert

[DSilv11]

Just a note to tell you I am looking at the logs.

And a reminder that as a user to user forum, not all "answers" are correct.

i.e a pci device is never inilizied by BMC. The BMC is a seperate device. The video devicer is inside the same chip, but it on a pci\pcie bus and is controlled by the BIOS & processor during DC power on. Based on your and sagayo issues, it does not seem to be inilizing correctly.

More when I get through the logs and PCI space dumps.

[DSilv11]

Hmm, Something strange with using a PCI video card for sure, but I did get one to work.

Disabled the on-board video and installed a PCIe and a PCI video card.

PCIe card came up primary and nothing on PCI card.

Loaded XP32

Device manager sees both video card, but the PCI card is banged out and can't start.

Load all the drivers (PCI card still banged out)

Rebooted - still banged out

Shutdown and removed PCIe video card.

Booted system. PCI card displaied POST and booted into OS without any issue.

AC cycled system and it still is working.

PCI card was an old S3 card and the PCIe card was ATI RV370 Radeon X300SE.

I would guess based on these symptoms that BIOS is having a configuration issue setting up CMOS but once it finds and configures the cmos\ PCI card, it is happy. This needs some more investigation.

[DSilv11]

demottrc wrote:

Is it possible that these PCI/PCI bridges are causing some kind of compatibility issue that is limited to slot 5?

Very Possible

The mother board uses a PCI-e switch to expand a X8 pcie lane from the processor to slot 4 (x4) & slot 5 (x8) and the FWB-PCIE-02 is using a PCIe switch to split a x1 PCIe port out to 2 IEEE devices. Under the PCIe specification it is premissiable to swizzle the lanes to assist in routing of the traces.

However, if when all is said and done, some vendors PCIe switches have issues if lane 0 does not get connected to lane 0 when going through two switches. I don't see anything on the mother board side that would causing an issue on a quick look it is all stright through, but the card may be swizzling or if you are using a riser in the slot it might swizzle and the two switches may not be able to train the link.

Are you using any riser cards or are you directly plugged into the mother board?

[idata]

Doc_SilverCreek wrote:

Hmm, Something strange with using a PCI video card for sure, but I did get one to work.

Disabled the on-board video and installed a PCIe and a PCI video card.

PCIe card came up primary and nothing on PCI card.

Loaded XP32

Device manager sees both video card, but the PCI card is banged out and can't start.

Load all the drivers (PCI card still banged out)

Rebooted - still banged out

Shutdown and removed PCIe video card.

Booted system. PCI card displaied POST and booted into OS without any issue.

AC cycled system and it still is working.

PCI card was an old S3 card and the PCIe card was ATI RV370 Radeon X300SE.

I would guess based on these symptoms that BIOS is having a configuration issue setting up CMOS but once it finds and configures the cmos\ PCI card, it is happy. This needs some more investigation.

Hmm, interesting. That definitely shows that it should be possible to get a PCI video card to work. Unfortunately, that method didn't seem to work for me. Disabling the onboard video and using a PCI-E card works fine by itself; but as soon as I add the H545H512P PCI card, the system refuses to POST (same LED error code: 0x50). But since it worked in your case, I will try getting some other PCI video cards to test with.

Thanks,

Robert

[idata]

Doc_SilverCreek wrote:

demottrc wrote:

Is it possible that these PCI/PCI bridges are causing some kind of compatibility issue that is limited to slot 5?

Very Possible

The mother board uses a PCI-e switch to expand a X8 pcie lane from the processor to slot 4 (x4) & slot 5 (x8) and the FWB-PCIE-02 is using a PCIe switch to split a x1 PCIe port out to 2 IEEE devices. Under the PCIe specification it is premissiable to swizzle the lanes to assist in routing of the traces.

However, if when all is said and done, some vendors PCIe switches have issues if lane 0 does not get connected to lane 0 when going through two switches. I don't see anything on the mother board side that would causing an issue on a quick look it is all stright through, but the card may be swizzling or if you are using a riser in the slot it might swizzle and the two switches may not be able to train the link.

Are you using any riser cards or are you directly plugged into the mother board?

All cards are plugged directly into the motherboard. So I guess it is probably an issue with the PCI-E switch used on the FWB-PCIE-02 cards then? It is interesting that it works in Port 4, but not Port 5 though.

I will try finding some alternate multi-bus 1394b cards. Is it likely that other cards using PCI-E switches will have the same issue? Or do you think its just an isolated incompatibility?

Thanks again,

Robert

[idata]

Heh, you got me on the BMC/PCI/Video thing. One more good reason why I'm not an engineer...

I wonder if this is because of a limited amount of PCI addressible memory space that is allocated towards the PCI slot and/or the video subsystem? I know I'm grasping quite a bit here but there's no place in BIOS setup to reserve any memory space for video. Maybe if the PCI video card was old enough and not resource-intensive, it would initialize on its own.

I'm curious, what memory range is allocated toward the S3 card in XP? It may possible that there may not be enough addressible space for all of the features on the OP's card.

[idata]

I would tend to think that this is just an isolated incompatibility. I use a Highpoint RocketRAID 640 card that has a PLX 8609 8-lane PCIe switch. I tried to use two of these cards in different combinations between slots 4-6, but I always got an "OUT OF MEMORY" error when the card's BIOS screen appeared. I could only get one to work, but it did work reliably in each of the three slots, as long as I didn't have another card that had the same exact PCIe switch.

Then I added an Intel PRO/1000 PT Quad-port LP PCIe 8X NIC, which itself has another PCIe switch on it. While it did work along with the RocketRAID card, it only worked reliably if I put the RocketRAID in slot 6 and the Intel quad NIC in slot 4. The system was very unstable in any other combination, but I never took the time to find out why (I needed my server to work ASAP)

I think it's hit-or-miss, but this is probably a good reason why none of my cards are on the compatiblity list!

[DSilv11]

Well, Made it through the PCI space dumps and found an interesting thing (which means I am going to ask for more data.)

PCI Segment 00 Bus 02 Device 04 Func 00 [EFI 0002040000] is the link between the mother board and your card.

Without getting too deep, (or maybe just a little too deep) this line:

00000050: 40 00 01 18 00 00 00 00-00 00 40 00 00 00 00 00 *@.........@.....*

indicates some interesting things.

The last 40 indicates that a card is present in the slot 5

The 1 in the 01 indicates that the device is a PCIe Gen1 device.

The 0 indicates it is not working.

The 18 indicates a Link training error.

The question is what link error and why.

To get this I need to get even deeper into the PCI configuration space and get a capture of the PCIe Advance Error Registers.

However, I don't know anyway to get these from EFI.

There are tools for DOS, Windows and Linux that can dump these (what OS are you using?),

but before going into those, the real question is do you want to bother?

I am not seeing any issues on the mother board side using other generic PCIe bridge cards which makes me suspect that the link error is a timing issue on the card's PCIe to PCI bridge. With the extended error registers, I should be able glean some additional information which might help isolate the culprit.

I did some fact googling and found several 1394b 3 & 4 port cards which might meet your needs more swiftly and even leave you a PCIe slot for video also.

[idata]

Doc_SilverCreek wrote:

Well, Made it through the PCI space dumps and found an interesting thing (which means I am going to ask for more data.)

PCI Segment 00 Bus 02 Device 04 Func 00 [EFI 0002040000] is the link between the mother board and your card.

Without getting too deep, (or maybe just a little too deep) this line:

00000050: 40 00 01 18 00 00 00 00-00 00 40 00 00 00 00 00 *@.........@.....*

indicates some interesting things.

The last 40 indicates that a card is present in the slot 5

The 1 in the 01 indicates that the device is a PCIe Gen1 device.

The 0 indicates it is not working.

The 18 indicates a Link training error.

The question is what link error and why.

Ah, that is interesting. So the motherboard is actually detecting a card in slot 5, but a reliable link is not being established.

Doc_SilverCreek wrote:

To get this I need to get even deeper into the PCI configuration space and get a capture of the PCIe Advance Error Registers.

However, I don't know anyway to get these from EFI.

There are tools for DOS, Windows and Linux that can dump these (what OS are you using?),

but before going into those, the real question is do you want to bother?

The current operating system is Windows 7 Professional 64-bit. If it is not too difficult, I wouldn't mind dumping the Advanced Error Registers. What software would I need to do this?

Doc_SilverCreek wrote:

I am not seeing any issues on the mother board side using other generic PCIe bridge cards which makes me suspect that the link error is a timing issue on the card's PCIe to PCI bridge. With the extended error registers, I should be able glean some additional information which might help isolate the culprit.

I did some fact googling and found several 1394b 3 & 4 port cards which might meet your needs more swiftly and even leave you a PCIe slot for video also.

Since it does appear to be an isolated compatibility issue, I have also been looking for alternate multi-channel 1394b controller cards. I have located a couple potential candidates, and am waiting to hear back from the camera manufacturer regarding compatibility with their cameras and software. Possible cards being considered are: the FireBoard800-e Pro Dual Bus card from Unibrain and the PCIe-FIW64 Quad Bus card from Adlink. If you happen to know of any other multi-channel 1394b cards, I would definitely appreciate your recommendations. It is important to note that due to the bandwidth requirements of the cameras being used (~75MB/s), each camera will need its own dedicated 1394b bus. This means that the 1394b cards must have multiple OHCI chips, presumably connected via a PCI-E switch.

Thanks again,

Robert

[idata]

sugayo wrote:

Heh, you got me on the BMC/PCI/Video thing. One more good reason why I'm not an engineer...

I wonder if this is because of a limited amount of PCI addressible memory space that is allocated towards the PCI slot and/or the video subsystem? I know I'm grasping quite a bit here but there's no place in BIOS setup to reserve any memory space for video. Maybe if the PCI video card was old enough and not resource-intensive, it would initialize on its own.

I'm curious, what memory range is allocated toward the S3 card in XP? It may possible that there may not be enough addressible space for all of the features on the OP's card.

Oh, I hadn't considered the required memory space of the video card to be a possible cause, but that could definitely be a factor. The card I was using had 512MB of video memory, which is way more than necessary for this application. I will look into trying a card with less video memory. Any idea what a safe upper bound may be?

Thanks,

Robert

[DSilv11]

In windows my preferred free tool is R\W everything.

http://rweverything.phpnet.us/ http://rweverything.phpnet.us/ (I know nothing about the vendor, but the tool is very handy & very dangerous tool as it will let you write to register best left alone unless you know what your doing)

In this case you just need to install by running the setupRW and it will start up in the default PCI space mode.

Click on the save all icon and you can save your entire PCI configuration space to a .rw file.

(You can also view it in notepad if you want to see what is in it.)

This file is a bit longer than the last. (about 250 pages)

I have not worked wityh any ieee cards for quite awhile so really do not have a recommendation.

The http://www.unibrain.com/Products/p1394/fb800e_dual.htm http://www.unibrain.com/Products/p1394/fb800e_dual.htm indicates 160 Mbytes/sec total bandwidth throughput and runs PCIe X1

The http://www.adlinktech.com/PD/web/PD_detail.php?pr2008&cKind&pid=798&seq=&id=&sid http://www.adlinktech.com/PD/web/PD_detail.php?pr2008&cKind&pid=798&seq=&id=&sid= lists data transfer rates up to 3.2 Gb/s, which is about twice as fast running PCie x4.

[idata]

Doc_SilverCreek wrote:

In windows my preferred free tool is R\W everything.

http://rweverything.phpnet.us/ http://rweverything.phpnet.us/ (I know nothing about the vendor, but the tool is very handy & very dangerous tool as it will let you write to register best left alone unless you know what your doing)

In this case you just need to install by running the setupRW and it will start up in the default PCI space mode.

Click on the save all icon and you can save your entire PCI configuration space to a .rw file.

(You can also view it in notepad if you want to see what is in it.)

This file is a bit longer than the last. (about 250 pages)

I have attached the log file from RW-Everything.

Hardware configuration was setup to be the same as in the previous Test 1:

Slot 6: FWB-PCIE-02

Slot 5: FWB-PCIE-02Slot 4: FWB-PCIE-02Slot 3: FWB-PCIE-02Slot 2: FWB-PCIE-01

Slot 1: Empty

Thanks,

Robert

[idata]

Hello Robert,

I have successfully deployed this board for a custom storage server, but I have experienced similar issues with the PCI/PCIe subsystem. Hopefully my experience will help you troubleshoot your problems!

First, for problem # 1, I found that the PCI slot is not initialized when the BMC is booted, which happens as soon as power is applied to the board. The PCI slot is initialized when the BMC detects an ACPI PWR_ON event (pressing the power button), and the CPU and the PCH begins the actual POST sequence, well after the video subsystem is already initialized by the BMC. What I think is happening is that the video subsystem on the BMC does not check for a PCI-based video card in order to determine if the on-board video needs to be initialized for the BMC video redirect feature. Enabling the dual video feature in BIOS only forces the BMC to check the mechanical 16X (electrically 8X) Gen2 PCIe slot for the presence of a video card so that the BMC video redirect feature would still work. However, since the video card itself receives power when power is first applied (prior to actually hitting the power button for cold boot), the video BIOS expects a POST command from the system BIOS to tell the video card that it is ready to process the video card's own POST sequence. This never happens because the BMC is not checking for a video BIOS on the PCI slot, but since the card is already asserting a POST-Ready command on the bus, the BMC freezes because the command is unexpected and the exception is left unhandled. Again, this is what I suspect that is happening, but I could not be completely certain that this is the case. You would need to understand the difference between a BMC boot-up sequence and a BIOS boot-up sequence to really understand why your video is not working as expected. In my application, I use the on-board video function because I have no need for a high-performance desktop graphics solution on a entry-level server grade platform... The short answer to problem # 1 is that there is no solution for what you want to do. This motherboard will not work for your purpose with the hardware that you mentioned, and I don't think anyone makes a combo video/1394b card...

For problem # 2, your 1394b cards are Gen1 cards - they are not compatible with this board's Gen2 slots. Even though they seem to be function in slot 4 and 6, I can attest that you will experience intermittent data loss and performance issues if you continue to use Gen1 cards in the Gen2 slots. I started my project with Gen1 SATA RAID cards, and although I could build my RAID arrays, they always had intermittent communication issues, causing the cards to rescan for drives every 30 seconds or so depending on how much traffic was on that particular PCIe lane. With 20 hard drives online that had to be rescanned 5 drives at a time, this was a very time-consuming process, during which the OS (tested with both Linux and Server 2008) was unresponsive. My only resolve was to swap the cards in Gen2 slots with Gen2 cards, and I have not had any issues or performance problems since.

My only reliable reference is the S3420GP Technical Product Specification document on the Intel support site. Take a peek at it if you haven't already, because it answered a lot of my questions as to why I couldn't just simply plug in a PCIe card and get it to work as I would expect it to. However, I am curious as to why you are using entry-level server grade hardware when your application clearly calls for a high-performance workstation? Just seems like a mismatch of hardware to what you need your system to do.

Good luck with your project! Hope this helps!

Best Regards,

Oly

[idata]

sugayo wrote:

Hello Robert,

I have successfully deployed this board for a custom storage server, but I have experienced similar issues with the PCI/PCIe subsystem. Hopefully my experience will help you troubleshoot your problems!

First, for problem # 1, I found that the PCI slot is not initialized when the BMC is booted, which happens as soon as power is applied to the board. The PCI slot is initialized when the BMC detects an ACPI PWR_ON event (pressing the power button), and the CPU and the PCH begins the actual POST sequence, well after the video subsystem is already initialized by the BMC. What I think is happening is that the video subsystem on the BMC does not check for a PCI-based video card in order to determine if the on-board video needs to be initialized for the BMC video redirect feature. Enabling the dual video feature in BIOS only forces the BMC to check the mechanical 16X (electrically 8X) Gen2 PCIe slot for the presence of a video card so that the BMC video redirect feature would still work. However, since the video card itself receives power when power is first applied (prior to actually hitting the power button for cold boot), the video BIOS expects a POST command from the system BIOS to tell the video card that it is ready to process the video card's own POST sequence. This never happens because the BMC is not checking for a video BIOS on the PCI slot, but since the card is already asserting a POST-Ready command on the bus, the BMC freezes because the command is unexpected and the exception is left unhandled. Again, this is what I suspect that is happening, but I could not be completely certain that this is the case. You would need to understand the difference between a BMC boot-up sequence and a BIOS boot-up sequence to really understand why your video is not working as expected. In my application, I use the on-board video function because I have no need for a high-performance desktop graphics solution on a entry-level server grade platform... The short answer to problem # 1 is that there is no solution for what you want to do. This motherboard will not work for your purpose with the hardware that you mentioned, and I don't think anyone makes a combo video/1394b card...

For problem # 2, your 1394b cards are Gen1 cards - they are not compatible with this board's Gen2 slots. Even though they seem to be function in slot 4 and 6, I can attest that you will experience intermittent data loss and performance issues if you continue to use Gen1 cards in the Gen2 slots. I started my project with Gen1 SATA RAID cards, and although I could build my RAID arrays, they always had intermittent communication issues, causing the cards to rescan for drives every 30 seconds or so depending on how much traffic was on that particular PCIe lane. With 20 hard drives online that had to be rescanned 5 drives at a time, this was a very time-consuming process, during which the OS (tested with both Linux and Server 2008) was unresponsive. My only resolve was to swap the cards in Gen2 slots with Gen2 cards, and I have not had any issues or performance problems since.

My only reliable reference is the S3420GP Technical Product Specification document on the Intel support site. Take a peek at it if you haven't already, because it answered a lot of my questions as to why I couldn't just simply plug in a PCIe card and get it to work as I would expect it to. However, I am curious as to why you are using entry-level server grade hardware when your application clearly calls for a high-performance workstation? Just seems like a mismatch of hardware to what you need your system to do.

Good luck with your project! Hope this helps!

Best Regards,

Oly

Problem 1:

Ah, so you are saying that it is impossible to use a video card in the S3420GPLX PCI slot. If so, then that is disappointing. We don't strictly need a high performance video card in this application. In fact, once deployed we probably won't even need one at all. But during development it would be helpful to see the video stream(s) in real-time. The onboard video doesn't seem to provide even basic 2D acceleration though, limiting the displayed frame-rate to below 5Hz. Although not ideal, we could use a PCI-E video card provided we still have four working PCI-E slots available for 1394b controller cards during development. Do you think a video card would work in Slot 5? Or is Slot 6 the only video capable slot on this motherboard?

Problem 2:

Is the Gen1 card, Gen2 slot incompatibility a universal thing? Or is it just limited to this motherboard? I always thought that PCI-E Gen2 was supposed to be backwards compatible with Gen1 cards.

I took a look at the S3420GP TPS document. The block diagram for the S3420GPLX indicates that Slots 4 & 5 are connected to the CPU via a Gen2 PCI-E switch. Is this the source of the compatibility issue? I'm curious why Slot 4 seems to work fine though…

The decision to use this particular motherboard was due to the combination of having 5 PCI-E slots and supporting 32GB of memory. Other workstation level boards were considered, but most either supported less memory, or did not have sufficient PCI-E slots available. I was not aware of any compatibility issues inherent in using entry-level server hardware, so it seemed to be the most cost effective solution.

The workstation/desktop boards that I looked at also appear to either use a mix of Gen1/Gen2 slots or be entirely Gen2 based. If we do have to abandon the S3420GPLX, wouldn't we still have the same problem if we used the PCI-E Gen1 1394b cards?

Thanks,

Robert