PCI-5640R Impaired Transmit data path

[avenn]

I am using the PCI-5640R with the "default personality driver" to send and receive IF signals.

Everything had been working great until I recently moved the card into a new machine at which point the transmit data path was always returning with:

"Error -301903 occurred at PCI-5640R API.lvlib: _ni5640R Check Generation Underflow.vi"

This basically means that the card was not being supplied data quickly enough.

To take my software of the equation I replicated the problem with the NI5640R demo applications called "Stream To Disk.vi" and "Stream from Disk.vi".  "Stream To Disk.vi" works fine, but the "Stream from Disk.vi" always returns with the underflow error suggesting that the TX data path is somehow impaired.  I know that the host can transfer tx data to the card because demo applications that do not require real time streaming such as "Single Tone Generation.vi" work fine.

  The only obvious difference between my 2 machines is that the working one has a PCI slot and the one with the impaired TX data path has  a PCI-X slot on a riser card (which from what I've read should be backwards compatible with PCI).

Both machines are running 64bit windows 7 with Labview 2010(32bit) and NI5640 driver version 1.5.

Here are the specs for the new machine that is having problems:

  Intel Xeon X5680 3.33GHz
  Windows 7 64 bit
  Model: SuperMicro 6016T-UF
  Motherboard: X8DTU-F  http://www.supermicro.com/products/motherboard/QPI/5500/X8DTU-F.cfm
  PCI-X Riser Card: RSC-R1UU-AXE8

I have also attached a NI IO trace that was taken while running the "Stream From Disk.vi" example. ( I had to add the .log extension for the forum to take it)

--Andrew

[avenn]

I am using the PCI-5640R with the "default personality driver" to send and receive IF signals.

Everything had been working great until I recently moved the card into a new machine at which point the transmit data path was always returning with:

"Error -301903 occurred at PCI-5640R API.lvlib: _ni5640R Check Generation Underflow.vi"

This basically means that the card was not being supplied data quickly enough.

To take my software of the equation I replicated the problem with the NI5640R demo applications called "Stream To Disk.vi" and "Stream from Disk.vi".  "Stream To Disk.vi" works fine, but the "Stream from Disk.vi" always returns with the underflow error suggesting that the TX data path is somehow impaired.  I know that the host can transfer tx data to the card because demo applications that do not require real time streaming such as "Single Tone Generation.vi" work fine.

  The only obvious difference between my 2 machines is that the working one has a PCI slot and the one with the impaired TX data path has  a PCI-X slot on a riser card (which from what I've read should be backwards compatible with PCI).

Both machines are running 64bit windows 7 with Labview 2010(32bit) and NI5640 driver version 1.5.

Here are the specs for the new machine that is having problems:

  Intel Xeon X5680 3.33GHz
  Windows 7 64 bit
  Model: SuperMicro 6016T-UF
  Motherboard: X8DTU-F  http://www.supermicro.com/products/motherboard/QPI/5500/X8DTU-F.cfm
  PCI-X Riser Card: RSC-R1UU-AXE8

I have also attached a NI IO trace that was taken while running the "Stream From Disk.vi" example. ( I had to add the .log extension for the forum to take it)

--Andrew

[Peter_W]

Hi Andrew,

Thank you for reposting your question here, it should be addressed in a timely manner.

For reference, here is the original post from the Real Time forum.

Regards,

Peter W.

[JVP]

Hello Andrew,

I apologize for the delayed response. Thank you for the IO Trace. I was hoping to get more information from the trace, however from what I saw there weren't any errors. You are correct about the PCI-X being compatible to PCI. However, there are instances where the PCI will perform slower in a PCI-X slot (this is from what I have read). I just wanted to inform you that I am looking into this and going to see if I can reproduce this issue on my end. I will keep you updated as I progress.

Best Regards,

Jignesh P

[JVP]

Hello Andrew,

I would like for you to upgrade to NI-5640r 1.6 driver version, which can be found here. Can you also tell me what version of NI RIO you have installed? After you install the latest driver version do you still get the same error?

Best Regards,

Jignesh P

[avenn]

 We are actually doing this on 2 machines in parallel and I upgraded one of them to use the 1.6 version of the driver and I get identical behavior.

Our NIRIO version is 4.0 according to the Measurement & Automation Explorer.

--Andrew

[JVP]

Hello Andrew,

Can you tell me what NI RIO driver version you have installed? I know you mentioned that this is only happening on the system with the PCI-X, just want to confirm that.

I am trying to setup a system here (software wise) to look exactly like yours.

Best Regards,

Jignesh P

[JVP]

Hi Andrew,

Thanks for the update, I saw the edit after I had replied.

Best Regards,

Jignesh P

[avenn]

We are using NI-RIO version 4.0.

--Andrew

[JVP]

Hello Andrew,

I was able to test the PCI-5640r in a PCI-X slot with one of our rack mount controllers (RMC). Which worked just fine.However, in your case that's a little different. I have an older Pentium 4 which is native PCI RMC. The reason why you arerunning into issue with the PCI-X slot is because:

The newer RMC are native to PCIe system, they generally have a PCIe/PCI bridge which goes through the south bridge(ICH) instead of the north bridge (IOH). As a system gets more and more complex, the PCI card gets further and furtheraway from memory controller. If you had an older system you would be fine with a PCI card in a PCI-X slot. Because thenewer motherboards are PCIe based, and there isn't a direct bath for the PCI device to the DMA/memory controller, that is why you are seeing the issue.

The only suggestion I can make is that you try using an older RMC or older PCI native device.

Best Regards,

Jignesh P