Netgate Discussion Forum
    • Categories
    • Recent
    • Tags
    • Popular
    • Users
    • Search
    • Register
    • Login

    1000mbps Router

    Scheduled Pinned Locked Moved Hardware
    23 Posts 10 Posters 21.9k Views
    Loading More Posts
    • Oldest to Newest
    • Newest to Oldest
    • Most Votes
    Reply
    • Reply as topic
    Log in to reply
    This topic has been deleted. Only users with topic management privileges can see it.
    • N
      nexusN
      last edited by

      Sorry for bumping the post, but I have one important point to clarify,
      Can you confirm that E-350 can be an option for full utilization of 1Gbps NAT?

      On feedbacking the incapability of Atom,
      some on the other forums replied that it is Atom's slow DMI on all peripherals,
      and they think that E-350 will be enough.
      If so, with its TDP E-350 can also be considered for building the router.

      Thank you.

      1 Reply Last reply Reply Quote 0
      • C
        clarknova
        last edited by

        I expect any socket 1155 CPU will do 1000 mbps on pfsense. Get a good DC-DC PSU, such as the picoPSU and your idle power consumption could be around 20W (I built a Core i3 system with a 3.5" hdd that idled around 17W measured at the wall). Good luck meeting your budget.

        I'm not sure about the Brazos parts. I think 1000 Mbps would be pushing it, although you might get close with fastforwarding enabled. I saw somewhere around 600 Mbps in iperf on a D510 using that option.

        db

        1 Reply Last reply Reply Quote 0
        • S
          slth
          last edited by

          You don't even need a PicoPSU to achieve those figures

          What would be the best way to benchmark the maximum troughput? I could set up a cross cable to another gigabit host and run iperf, but this method wouldn't represent the actual load on a system that has to route 1000mbps of traffic.

          1 Reply Last reply Reply Quote 0
          • C
            clarknova
            last edited by

            iperf should give you some good figures to work with. You can adjust packet size, TCP window, parallel streams, etc to simulate different loads. I wouldn't use pfsense itself as an iperf endpoint though. If you want to test pfsense's routing throughput you'll want to connect a couple hosts on different interfaces to do the testing.

            Somebody on dslreports.com just mentioned ixload. It looks interesting but I haven't tried it yet.

            db

            1 Reply Last reply Reply Quote 0
            • N
              nexusN
              last edited by

              @slth:

              You don't even need a PicoPSU to achieve those figures

              What would be the best way to benchmark the maximum troughput? I could set up a cross cable to another gigabit host and run iperf, but this method wouldn't represent the actual load on a system that has to route 1000mbps of traffic.

              Thank you for sharing,
              no doubt a good news to hear ;D
              Then I will go ahead with the G530 as my router, idling at 20W will be good enough.

              1 Reply Last reply Reply Quote 0
              • N
                nexusN
                last edited by

                @nexusN:

                Sorry for bumping the post, but I have one important point to clarify,
                Can you confirm that E-350 can be an option for full utilization of 1Gbps NAT?

                On feedbacking the incapability of Atom,
                some on the other forums replied that it is Atom's slow DMI on all peripherals,
                and they think that E-350 will be enough.
                If so, with its TDP E-350 can also be considered for building the router.

                Thank you.

                Sorry for putting this old post to the front, I still want to know if E-350 is a potential option for 1Gbps router build.
                I know Atom D525 can only handle upto ~550Mbps, but how about E-350?
                My estimation is as below:
                http://www.anandtech.com/bench/Product/328?vs=110&i=25.26.27.28.29.30.31.32.33.34.35.36.37.38.39.40.41.42.43.45.46
                D510=1.6GHz, D525=1.83GHz, D2700=2.13GHz,
                the score in Cinebench for D2700 should be ~1.33*709=Cinebench:942.97, D525=Cinebench:810
                while for routing it should be largely a single thread issue, E-350 maybe a better choice on it. I am not sure if HT would help.
                D525@Cinebench:810=550Mbps, D2700@Cinebench:942.97 should be around 640Mbps, then E-350@Cinebench:1174 should be around 797Mbps, ignoring the effect of HT.

                Anyone would be sharing on this?
                Thanks in advance.

                1 Reply Last reply Reply Quote 0
                • W
                  wallabybob
                  last edited by

                  The name Cinebench suggests to me its a benchmark involving playing movie files. Such a benchmark is likely to be able to take advantage of the enhanced graphics capabilities of the E-350 over the Atom based systems. Hence I would be very cautious about extrapolating Cinebench performance to routing. (I don't know of any routing software using graphics hardware for routing.)

                  1 Reply Last reply Reply Quote 0
                  • pttP
                    ptt Rebel Alliance
                    last edited by

                    According to the info posted in Cinebench web, it also test CPU

                    CINEBENCH is the perfect tool to compare CPU and graphics performance across various systems and platforms (Windows and Mac OS X). And best of all: It's completely free.

                    http://www.maxon.net/en/downloads/cinebench/technical-information.html

                    Maybe it can be valid if you run the Benchmark in Advanced mode and select "Single Core" CPU benchmark, due the fact that PF will only use a single core.

                    1 Reply Last reply Reply Quote 0
                    • M
                      mattlach
                      last edited by

                      I can't speak for the CPU use for that kind of load, but I will say this.

                      Don't plan on using on board ethernet.

                      Either get two separate intel PCIe server nic's or one of their dual port models.

                      Integrated NIC's - especially Realtek - are terrible for this kind of load.

                      1 Reply Last reply Reply Quote 0
                      • N
                        nexusN
                        last edited by

                        @mattlach:

                        I can't speak for the CPU use for that kind of load, but I will say this.

                        Don't plan on using on board ethernet.

                        Either get two separate intel PCIe server nic's or one of their dual port models.

                        Integrated NIC's - especially Realtek - are terrible for this kind of load.

                        For NIC,
                        I am glad to say Intel Onboard 82579V + Intel Pro1000 CT w/ 82574L are fine to use in the above case on testing.

                        1 Reply Last reply Reply Quote 0
                        • First post
                          Last post
                        Copyright 2025 Rubicon Communications LLC (Netgate). All rights reserved.