This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

AM1808 Ethernet performance

Other Parts Discussed in Thread: AM1808

I want to check the performance benchmark of Ethernet of the AM1808 EVM according to this page, http://processors.wiki.ti.com/index.php/DaVinci_PSP_03.20.00.13_Device_Driver_Features_and_Performance_Guide#Performance_and_Benchmarks_2

I have installed the iperf tool (iperf-2.0.4) in my linux PC and the AM1808 EVM, with the EVM using NFS on the linux PC,
but then the performance is much lower than specified in the page,

in Linux PC, I type "iperf -s" and get
------------------------------------------------------------
Server listening on TCP port 5001
TCP window size: 85.3 KByte (default)
------------------------------------------------------------
[  4] local 210.17.139.219 port 5001 connected with 210.17.139.217 port 60751
------------------------------------------------------------
Client connecting to 210.17.139.217, TCP port 5001
TCP window size: 32.0 KByte (WARNING: requested 16.0 KByte)
------------------------------------------------------------
[  6] local 210.17.139.219 port 39770 connected with 210.17.139.217 port 5001
Waiting for server threads to complete. Interrupt again to force quit.
[ ID] Interval       Transfer     Bandwidth
[  6]  0.0-60.0 sec    267 MBytes  37.3 Mbits/sec
[ ID] Interval       Transfer     Bandwidth
[  4]  0.0-60.0 sec    197 MBytes  27.5 Mbits/sec

then in EVM, I type "iperf -c 210.17.139.219 -w 16k -d -t60" and get
------------------------------------------------------------
Server listening on TCP port 5001
TCP window size: 32.0 KByte (WARNING: requested 16.0 KByte)
------------------------------------------------------------
------------------------------------------------------------
Client connecting to 210.17.139.219, TCP port 5001
TCP window size: 32.0 KByte (WARNING: requested 16.0 KByte)
------------------------------------------------------------
[  3] local 210.17.139.217 port 60751 connected with 210.17.139.219 port 5001
[  5] local 210.17.139.217 port 5001 connected with 210.17.139.219 port 39770
[ ID] Interval       Transfer     Bandwidth
[  5]  0.0-60.0 sec    267 MBytes  37.3 Mbits/sec
[ ID] Interval       Transfer     Bandwidth
[  3]  0.0-60.0 sec    197 MBytes  27.5 Mbits/sec

So the trasnfer rate is just 27.5Mbps, right? much lower than the 66.7Mpbs mentioned in the page.
Have I done something wrong?

Besides that, how to test the RMII PHY?

Thanks for any help

  • I think the result given by TI is tested in the LAN, but it seems that your test enviroment is in the WAN.right?

  • I have changed the ip to the LAN ip now, but still get the same performance

    in EVM, I type "iperf -c 192.168.64.219 -w 16k -d -t60" and get
    ------------------------------------------------------------
    Client connecting to 192.168.64.219, TCP port 5001
    TCP window size: 32.0 KByte (WARNING: requested 16.0 KByte)
    ------------------------------------------------------------
    ------------------------------------------------------------
    Server listening on TCP port 5001
    TCP window size: 32.0 KByte (WARNING: requested 16.0 KByte)
    ------------------------------------------------------------
    [  3] local 192.168.64.217 port 40212 connected with 192.168.64.219 port 5001
    [  5] local 192.168.64.217 port 5001 connected with 192.168.64.219 port 56143
    [ ID] Interval       Transfer     Bandwidth
    [  5]  0.0-60.0 sec    267 MBytes  37.4 Mbits/sec
    [ ID] Interval       Transfer     Bandwidth
    [  3]  0.0-60.0 sec    198 MBytes  27.6 Mbits/sec

    in linux host, I type "iperf -s -w 16k"
    ------------------------------------------------------------
    Server listening on TCP port 5001
    TCP window size: 32.0 KByte (WARNING: requested 16.0 KByte)
    ------------------------------------------------------------
    [  4] local 192.168.64.219 port 5001 connected with 192.168.64.217 port 40212
    ------------------------------------------------------------
    Client connecting to 192.168.64.217, TCP port 5001
    TCP window size: 32.0 KByte (WARNING: requested 16.0 KByte)
    ------------------------------------------------------------
    [  6] local 192.168.64.219 port 56143 connected with 192.168.64.217 port 5001
    Waiting for server threads to complete. Interrupt again to force quit.
    [ ID] Interval       Transfer     Bandwidth
    [  6]  0.0-60.0 sec    267 MBytes  37.4 Mbits/sec
    [ ID] Interval       Transfer     Bandwidth
    [  4]  0.0-60.0 sec    198 MBytes  27.6 Mbits/sec 

    thanks

     

  • sorry to mislead you~~~~.

    TI's result come from Rx+Tx, so your ethernet performance is 37.4 + 27.6=65 Mbits/sec .

    Good Job!

  • Oh, if that's the case, then the perfromance seems ok.

    But TI's performance page says "Data captured here is for "iperf" in client mode.",
    while in the iperf test result, although i use client in EVM, it says
    ------------------------------------------------------------
    Client connecting to 192.168.64.219, TCP port 5001
    TCP window size: 32.0 KByte (WARNING: requested 16.0 KByte)
    ------------------------------------------------------------
    ------------------------------------------------------------
    Server listening on TCP port 5001
    TCP window size: 32.0 KByte (WARNING: requested 16.0 KByte)
    ------------------------------------------------------------
    [  3] local 192.168.64.217 port 40212 connected with 192.168.64.219 port 5001
    [  5] local 192.168.64.217 port 5001 connected with 192.168.64.219 port 56143
    [ ID] Interval       Transfer     Bandwidth
    [  5]  0.0-60.0 sec    267 MBytes  37.4 Mbits/sec
    [ ID] Interval       Transfer     Bandwidth
    [  3]  0.0-60.0 sec    198 MBytes  27.6 Mbits/sec

    So the part in red color is also counted as the performance? it is really confusing me.

    One more question, in the client side, [  3]  and [  5], which one is transmitting data, which one is receiving data? 

    Thanks again.

  • I was considering using this CPU for an IP camera with 10MP+ resolution. We have our own image processing chip and compressor, so no reason to try to use the other offerings that help with compression.

    But the ethernet performance can't support that res.  Even H.264 will saturate and be forced to raise the QFACTOR.

    Otherwise, the chip is a great deal!  The price is what attracted me. Is there one with higher performance ethernet, or perhaps a static ram buffer can be added to improve performance, as is possible with some stand alone ethernet chips?

    I suspect this chip tops out at 5MP transmission, especially if the camera supplies multiple streams in both h.264 and jpeg.