PXI

cancel
Showing results for 
Search instead for 
Did you mean: 

Transfer Rate decrease when the MXI-4 interface is installed.

      My original configuration was a PXI-1056 chassis with a pxi-8196 controller running LabviewRT.  With this configuration, reading the data from a third party digitizer card took 130 usecs. The configuration was changed to use a second chassis with a MXI-4 interface to the origianl chassis. The controller was moved to the second chassis, a PXI-1031, and the MXI-4 interfaces installed. Now, with this new configuation the read time has increased to 190 usecs.
 
     Shouldn't the MXI be able to transfer the data just as fast as the PXI bus? So adding the interface shouldn't slow things down. The MXI interface is copper not fiber optic.
 
     Does the PXI-1031 have a slower bus than the PXI-1056?
 
     Is there overhead associated with the MXI interface and this is what is to be expected?
 
     Would using a fiber optic interface improve the read time?
 
Dave
 
 
0 Kudos
Message 1 of 6
(3,659 Views)
Hi Dave,

Both the PXI bus (which is part of the PCI bus) and MXI-4 have a theoritcal max rate of 132 MB/s.  However, the actual sustained rates are different.  With an embedded controller, you can transfer data at a sustained rate of 110 MByte/s.  With MXI-4, the sustained rates are 78 MBytes/s.  Both Copper and Fiber Optic will have the same sustained throughput rates.  MXI Express, which utilizes PCIe technology, can have 110 MBytes/s if you are interfacing with a PXI chassis.  It would be higher with a PXIe chassis.

Hope this helps!

Adam W
Applications Engineering
National Instruments

0 Kudos
Message 2 of 6
(3,644 Views)

Hi Adam,

Thanks for responding. From what you wrote I should expect slower rate through the MXI-4, 110 MB/s vs 78 MB/s. Which is an aproximate difference of 25%. With my configuration the transfer rate difference is about 50%, 30MB/s vs 20 MB/s.  The size of the block is only 4000 bytes. Which I am sure is the reason for the slower transfer rate to start with. But what could account for the big decrease when the MXI is used? Is the MXI even less effecient at smaller block size? Or is there something else.

I just wanted to make sure that the results I am seeing are what I should expect. And data transfers through the MXI-4 are slower and can be 50% slower than the same tranfer over the PXI bus.

One another point. The 130 usec transfer time is the typical value but occasionally the transfer time spikes to 250+usecs.

Dave

 

0 Kudos
Message 3 of 6
(3,638 Views)
Hi Dave,

I don't think that the difference between the two rates will necessarily be linear; that is, if you are not transferring at the max sustained rate, I don't think we can assume the difference between MXI and PXI will be 25% every time.  I am curious, though, how you are transferring your data.  You said you are doing a block of 4000 bytes.  How is that block broken up?  How big are the chunks of data you are transferring?

Thanks Dave!

Adam
0 Kudos
Message 4 of 6
(3,628 Views)

I am confused about my system and it's PXI bus speed. It seems to me that the bus is operating at only 33 MHz. The bus analyzer indicates that the bus is operating at a 33 MHz rate.

And the Keithley pxi extension can be set to run at 66 MHz  but when it is the digitizer data is corrupted. Does this indicate that the Aqiris digizer can only operate at 33 MHz?

Should the bus be able to operate at 66 MHz? How is it determined what speed the bus operates at? Is it determined by the ability of the "slowest" card in the system?

It seems to me that even if the pxi extension increases the transfer time from the digitzer, if the bus could be set to operate at 66 MHz the transfer time would be reduced.

 

Configuration

PXI-1031

        pxi-8196 controller

        NI mxi-4 or Keithley KPXI-CON-PXIREM  pxi extension

         Vangaurd PXI bus analyzer

 

PXI - 1056

        NI mxi-4 or Keithley KPXI-CON-PXIREM  pxi extension

        Aqiris DC440 digitizer

0 Kudos
Message 5 of 6
(3,537 Views)
dcwill,
The 32bit PCI bus can only operate at 33MHz as seen in the table located near the bottom of this page http://www.howstuffworks.com/pci.htm

I also ran the calculations on the drop in performance:
78mB/s = 0.0128205s/mB

110mB/s=0.0090909s/mB

0.0090909s/mB / 0.0128205s/mB = 0.70909
130uSec / 190uSec                      = 0.68421

Thus, the difference in time is very proportional to the difference in transfer rates.


Message Edited by John E on 06-08-2007 03:56 PM

John E.
Applications Engineering
National Instruments
0 Kudos
Message 6 of 6
(3,526 Views)