1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

Typical CPU usage for Wan->Lan throughput?

Discussion in 'Tomato Firmware' started by cloneman, Jun 21, 2014.

  1. cloneman

    cloneman Networkin' Nut Member

    I know this question comes up a lot but I have yet to know how to tell if my config is optimal or being bottlenecked by something. (shibby 117)

    I'm downloading plain http @ 30mbit/s over ethernet and my 600Mhz RT-N66U is at 0.4 ~ 0.5 CPU load. QoS on/off doesn't seem to matter - the only thing I have enabled is bandwith monitoring.

    This seems rather high for the Asus, IIRC I got the same load numbers or better as on my E4200v1.

    1. Should be relying on the webui to measure load?
    2. What's the typical load I should be getting on a RTN-66U for x amount of wan-to-lan mbits? (assuming no magic CTF or fast nat)
  2. remlei

    remlei Networkin' Nut Member

    CPU load gets no where 0.00-0.01 with LAN-WLAN traffic in my case using my Asus N12HP with Shibby v120 firmware. CPU load gets around 0.10-0.20 when there is WAN related traffic at the speed of 50mbit/sec. Most of the time it just sits in 0.00.

    the only thing I notice is that SIRQ gets around 30% load with around 120mbit/sec transfer rate.
  3. cloneman

    cloneman Networkin' Nut Member

    Thanks for the comment...

    Yeah something is up. Your router is only 300Mhz yet the CPU load is much less than mine. I will try upgrading to 120 and erasing nvram

    EDIT: Updated/Wiped to Shibby 121 Today. It's behaving as expected now; 30mbit wan-to-lan produces negligible CPU usage (0.04, 22% sirq)
    Last edited: Jul 13, 2014
  4. rs232

    rs232 Network Guru Member

    If I download something via e.g. torrent at 5 MB/sec or higher my Asus RT-N16 is completely unresponsive until the download is completed. I have a 120Mb connection for reference.

  5. shibby20

    shibby20 Network Guru Member

    Maximum Wan->Lan throughput:

    RT-N16 without bcm_nat ~100-120Mbps
    RT-N16 with bcm_nat ~ 170Mbps
    RT-N66U/AC66u without bcm_nat ~240Mbps
    RT-N66U/AC66u with bcm_nat ~320Mbps
    RT-N18u/AC56u/AC68u without CTF ~330Mbps
    RT-N18u/AC56u/AC68u with CTF ~600Mbps
    R7000 without CTF ~410Mbps
    R7000 with CTF ~950Mbps

    but if you load CPU by any others services then throughput will be much lower. USB support is the most
    aggravating CPU service. And it`s not samba/ftp/torrest only. Also if you have syslog/bw monitor/etc stored in USB device, then it also load CPU.
    Morac and cloneman like this.
  6. ycodryn

    ycodryn Reformed Router Member

    Hi. I have installed tomato-RT-N66U_RT-AC6x--121-VPN-64K and I want to enable bcm_nat but I`m getting
    module bcm_nat not found in modules.dep. I have to flash AIO to get that? Thank you.
  7. shibby20

    shibby20 Network Guru Member

    correct. Only K26 and K26RT-N support bcm_nat. K26RT-AC is not compatible with this module.
  8. ierwin

    ierwin Serious Server Member

    Sorry for reply an old post.
    Shibby you said "BCM4708 based devices CTF" means Tomato's CTF module backs in Tomato-ARM or you just means this is a test result for devices running on their stock firmware?
  9. shibby20

    shibby20 Network Guru Member

    Those results was on Tomato. Tomato-arm supports CTF
    Nullity and ierwin like this.
  10. rs232

    rs232 Network Guru Member

    Does CTF take the same approach as bcm_nat (e.g. bypass control to improve speed)?
    Or asked in other words does CTF disable QoS/filtering of traffic?
  11. piyokos

    piyokos Serious Server Member

    Any chance of BCM_NAT for K26RT-AC builds in the future?

    I've been running my E4200 for half a decade with Tomato firmwares but my cable company offers 300/20 now, with BCM_NAT enabled I can only get ~180-220Mbps, and that's with an overclock to 533MHz. I figure if I upgrade I might as well go for an AC but the ARM models are pricey (and vertically mounted - for cooling I guess, but the look drives me crazy) and K26RT-AC without BCM_NAT doesn't look like it'd perform any better.

Share This Page