Can't Get Gigabit Throughput From Your Router?

[thiggins]

I've seen numerous posts saying SNB's wired routing throughput test results aren't matching up to real world results with "Gigabit" fiber connections and I'd like to know why.

So if you're someone who is getting below 500 Mbps throughput from a router we say should deliver > 900 Mbps, I'd like to hear from you.

What I want ot know is the following:

• The router make and model
• WAN connection type
• What you are using to test performance (Speedtest, timed download, etc.)
  
• What changes have you made from defaults for wired routing features? I don't care about wireless settings.

Any other information you'd like to provide is welcome.

Thanks,

 

[sfx2000]

I'm wondering if some of the performance challenges with the Gigabit connections might be due to PPPOE, which can have some performance issues with off the shelf routers in the consumer space...

 

[MichaelCG]

From the consumer side of things, a lot of complaints are probably coming from the use of speedtest.net and other similar sites that just can't provide the bandwidth to max out connections. So many users will complain their connection/router is slow, but quite often it is the rest of the path through their ISP to the far end server that can't deliver at those speeds.

I know the generic speedtest.net site cannot come anywhere near close to hitting gig speeds on my GF connection. I have three options to actually come close to max'ing out my link.
1.) Use local systems for testing
- moving large files or using iperf
2.) Use the GF speedtest site (speedtest.googlefiber.net)
- Google has on local on-network servers in various markets
3.) Use multiple data streams
- download 10 copies of large ISO images across multiple download sites

The moment you start adding in latency and/or packet loss, the ability to handle single streams at those rates goes down quickly.

 

[thiggins]

Thanks, Michael.

Typically, people report 900 Mbps+ throughput connected directly to the ISP modem, then much lower throughput when the router in question is put inline. That would not be a problem with the speedtest method.

 

[Deepcuts]

I actually experienced the opposite.
SNB tests were showing RT-N66U and RT-AC68U doing less throughput then I was seeing while using speedtest.net and testing my own outside servers.
Since then, I gave up using Asus routers, but I had several friends of mine with the same Asus routers getting less throughput then myself. Turns out they were using "crappy" machines behind the router.
Once they got some decent hardware, their results were on par with mine.

 

[RMerlin]

Thanks, Michael.

Typically, people report 900 Mbps+ throughput connected directly to the ISP modem, then much lower throughput when the router in question is put inline. That would not be a problem with the speedtest method.

Some users are enabling features that disable their router's so-called hardware acceleration, causing the NAT/packet processing to be done purely on the CPU. Many routers won't clearly notify when that is the case. That will have a major impact on performance.

There's also still a lot of people using garbage-bin quality Ethernet cables, bought at the local convenience store. Gigabit is very sensitive to cable quality, so when they remove the router from the equation, it means they also remove the cable that is used either between the modem and the router, or the one between the router and the computer. Checking if there is any increase in retransmitted packets might be a good idea then. I was hit by this myself years ago, as the cable I self-crimped used cheaper Cat5-rated RJ45 connectors. Re-crimping with connectors rated for Cat 6 fixed my packet re transmission issues.

 

[MartinW]

Hi Tim,

I was actually one of those users

Im on a 1000/100 connection via provider Bahnhof / carrier IP-ONLY. Measurements below is with as much out of the box settings as possible, I have only changed any (if needed) to get the connection.

Measuring against Swedish test site bredbandskollen.se, I am based i Sweden, I have managed the following measurements WAN-> LAN, all at night at 03.00-04.00 the ensure minimum performance hit from net congestion.

DLINK DIR-879 960Mbit (Very good and better than your published result of 941)

Netgear Nighthawk(R) X4S R7800 700Mbit (Other sites indicate a max of approx 730Mbit (http://www.tweaktown.com/reviews/76...2600-mu-mimo-160mhz-router-review/index4.html), your measurements = 941Mbit)

ASUS RT66U approx 800Mbit ( I can not find your measurement, I do not have this in my possession anymore, so I can not verify)

Edgerouter X 936Mbit (I do not think you have tested this, please note that you will only get 430Mbit if you do not enable NAT hwassist)

Hope this helps

/Martin

 

[thiggins]

Thanks, Martin. I explored this a bit more.
https://www.smallnetbuilder.com/lan...-to-find-the-best-router-for-gigabit-internet