Does the PXIe 6375 require a 4 lane controller?

[oevidneg]

Does the PXIe 6375 require a 4 lane controller for everything other than single sample captures? I developed an real-time application using a 4 lane controller and it worked successfully.  When I substituted a single lane controller, it is only able to handle simple requests (reset, self test, return a single sample).  For everything else, I get a hardware failure message.  I suspect it is related to DMA/throughput.  Does anyone have insight into this?

[ZYOng]

What model of controllers are you using for single-lane and 4-lane?

[oevidneg]

I have a PXIe-8360 (single lane) and a PXIe-8370 (4 lane).

[ZYOng]

I don't think it is relevant to the lane numbers. Since you are using two different models of remote controller modules, I assume you have two different host desktops. Are they the same model?

Using the MXI system is tricky as NI does not have control over the BIOS on the host PC. You might need to go over all the steps in MXI-Express Compatibility and Connectivity Troubleshooting Guide

[oevidneg]

Actually, this was using the same computer.  I swapped out the controller in the PXI chassis and I have both interfaces installed in my computer chassis.

[ZYOng]

I want to point out that MXI enumeration is tricky and different models of NI modules behave differently.

Try swapping the PCI host adapter on the host desktop. Some desktop might not have DMA access on all PCIe slots.

[Robert__R]

Hi oevidneg,

Those remote controllers are pretty similar (except for number of lanes on the cable), so it's likely a throughput issue. As a side note, it's not strictly a "number of lanes" thing, except that the x1 controller (in this case) has 1/4 the bandwidth of the x4. A x1 controller at a faster rate (which NI doesn't offer) could/would also work.

You could look at the sample rate you're trying to achieve and calculate the bandwidth requirements to verify. For example, 2 bytes per sample * 1 MSample/sec * 8 channels = 16 MB/s. That should be achievable pretty easily. But running at 10 MSamples/sec would put it at 160 MB/s, which is more of a grey zone. If it's a reasonably efficient stream then it will work great, but there are many ways to increase overhead and break it.

- Robert

[oevidneg]

Hi Robert - the test case I was running was very simple.  I was capturing 32 8-byte floating point samples from a couple of channels at 10 Khz.  Somewhere between 10-20 Mhz. bandwidth. That's why I was speculating whether it was DMA related functionality or something of the kind as related to the controller capabilities. I have to look into ZYOng's comments.