PXI 5441 hardware timed carrier frequency sweep

[labJunky]

hi,

I looked on the ni website and have found a benchmark for streaming to a PXI 5421 Arbitrary waveform generator using a MXI-express controller.

http://zone.ni.com/devzone/cda/tut/p/id/6266#toc0

It states that streaming from disk at a rate of 114 MB/s is achievable, though it does not say whether this rate is the maximum or sustainable though-put.

This seems similar to the configuration I have, a MXI express link to a PXI 5441 card (which incidentally states in the data-sheet that 100 MB/s streaming is possible).

At the moment it seems that my streaming benchmark is not limited by the hard-drive reading speed, but by something else. The hard-drive reading rate benchmark was about 120 MB/s, but streaming from the disk or the RAM memory to the card only gave a benchmark of 50 MB/s.

Can anyone tell me how to improve this? Or how it was possible the 'you guys at NI' can get 114 MB/s using a similar setup?

Thanks,
Paul

[ryannorthington]

labjunky,

How are you gathering your benchmark for the 50 MB/s value?  Are you using the examples found on ni.com/streaming?

NI-FGEN Stream from Disk Using Win32 File IO 

From here download and unzip the examples.  The first thing you are going to want to do is create a file full of test data.

1. To do this go to NI-FGEN (This is the folder that was downloaded from the web) >> FileIOBenchmarkVIs >> Win32 Write to File Speed Test.vi. 
2. Make sure you jot down where you created the .bin file. 
3. Now go back to your downloaded folder and open NI-FGEN >> ApplicationExamples >> NI-FGEN >> More >> niFGEN Stream From Disk Maximum Transfer Rate.vi
4. Specify your test file and run the example.  The Data Transfer Rate measurement will be displayed at the bottom.

If you are not using these streaming examples can you give us a little more insight as to what you are using to take benchmarks? 

You are going to take a performance hit for every leg of the system you encounter (PXI bus, MXI card/cable, hard drive, etc).

Maybe you can post the code you are using to determine the throughput of the system.

[labJunky]

Hi Ryan,

thanks for your reply.

Yes I have been using the benchmarking tools found on the ni website, from the Streaming section. This is where I got he repeatable benchmark value of 50 MB/s. On a separate test using the Vi that tests the read speed of the hard drive (Win32 Read From File Speed Test.vi) I got 120 MB/s, but as you know this vi does not stream to the PXI card, it only reads the data from the disk into memory. Then I used the Win32 Read Write Block Size Test.vi to find the read/write block size giving the maximum speed. (The optimal block size was found to be 4194304.)

As you mentioned in your post I should expect to reduce the streaming performance for every section of my system between the memory on the computer (be it RAM of hard drive) to the pxi card itself. However, the benchmark on the National Instruments website ( http://zone.ni.com/devzone/cda/tut/p/id/6266#toc0 ) for a MXIe system and PXI 5421 card is stated to be 114 MB/s. How is this possible? What was the system?

My PC has the following specifications:

Dell Precision 390, Intel Core 2 Duo E6400, 2.13 GHz, 2 MB Cache, 1066 MHz FSB, Intel 975X Express chipset, 2 GB of RAM at 667 MHz, and 3 hard drives, one for the system, and the other for streaming from a RAID 0 configuration of two WD2500YS hard-drives. (The RAID controller is integrated on the Motherboard.)

It has three PCIe slots. A x16, x8 (wired x4), and a x1 slot in which the NI PCIe-8361 card sits. This is cabled via a MXIe cable, to a PXI 8360 in slot 1 of the PXI 1036 chassis. The PXI 5441 card sits in slot 6 of that chassis.

From this spec' is it possible to work out what the maximum streaming rate should be?

Thanks for your help,
Paul.

[ryannorthington]

labJunky,

Thanks for all the great information!  The information on the website appears to be inaccurate. I am unsure as to what the issue is but we are working to correct this error and will update the webpage with the correct benchmarking data. With regards to your issue, I do apologize for this error. Based on the data we have seen the rates you are seeing are close to expected if not slightly better.  If you would like to discuss other hardware options for increasing your throughput then your best bet would be to contact NI support.

[jaced]

Hi,

I have a few questions. What is the part number/ serial number of the PCIe card that you are using? Am I correct in thinking that you are using the niFGEN Stream From Memory Maximum Transfer Rate.vi to perform your stream from memory benchmark? We are currently investigating the root cause of this behavior and this information will be very helpful in clarifying the issue.  

[labJunky]

hi,

I had a look for the serial number and part number of the PCIe-8361. On the registration page of the ni website, I found the serial number is: Q10307404 (NI PXI-PCIe 8361, 3m cable...). Also I looked at the board itself since I was not sure where to find the part number. On the backside I found three white stickers. With the following numbers:
PCIe-8361, E696D1, and 191376C-02. Also printed on the circuit board was:
UL940-0 191378B-01.

If they are not the correct numbers could you tell me where I might find them? I no longer have the boxes the cards came in.

paul.

[labJunky]

also forgot to mention that you are correct. I am using the niFGEN Stream From Memory Maximum Transfer Rate.vi.

[jaced]

Hi Paul,

Thank you for all the information. I would like to take this issue offline if that is okay. With your permission I can have our web team look up your contact info in our database and either give you a call or send an e-mail, which ever you prefer and we can discuss possible reasons for this behavior and solutions. If you are comfortable with posting your contact info that would be fine as well but I completely understand if you do not wish to.

[labJunky]

hi there, please email me at paul.condylis@iesl.forth.gr

[labJunky]

hi,

just to finish. NI contacted me, sent me a new PCIe 8361 card (for free), which I installed and got 104 MB/s streaming from Ram and 102 MB/s from disk.

The new card and the old card (which I have to send back) are completely different. The components are different and the new one is about a third the size of the one I had. The model number of the board is 195315C-01L, PCIe 8361 low profile. I think the different from the board number of the old card is that it ends in B, and this one ends in C.

paul