In reply to a post by rotor2k:

Those numbers are always raw throughput. As soon as you tack on PPPoE, the numbers plummet.

Hmm, good reason to use Sky FTTC rather than BTwholesale FTTC.

In reply to a post by babis3g:

the 2130n was originaly up to 500mb but has been updated to 800mb for sure 100% even is at their official web saying this

I saw the draytek.com.tw site reporting the 2130 as designed for new "ultra-fast" broadband. Shame its not got the feature set of the 28xx range.

yeah the 28xx range stayed behind but then these models are for load balance really or handle 2 connection same time
To me it seems their chippest can not hundle more for what job all ready they doing
Hopefully will be an update up to 120mb like 2920 and 2820 models(got by time higher speed from the original out of factory)
I have the 2850n and a VM connection with 100mb and i am lucky if will hit 90mb very rear times but not compalins it handles 2 connections well (adsl & VM) with many useful features & options

But the 2130 seems more power full

Edited by deleted (Fri 18-Jan-13 17:33:57)

any idea how this could be done on a asuswrt or tomotausb router?

Maybe the experts can advise more about its chippest if is strong enough because with this review they opened its case (2130 but bear in mind is about 2 years old review so may some updates have taken place)

http://www.smallnetbuilder.com/wireless/wireless-rev...

Test speed
http://www.smallnetbuilder.com/wireless/wireless-rev...

I have also emailed the main offices in Taiwan and see what they saying because i am interesting about it

Edited by deleted (Mon 21-Jan-13 17:28:28)

In reply to a post by rotor2k:

In reply to a post by David_W:

FreeBSD - and therefore pfSense - doesn't currently support RFC 4638, so is limited to a 1492 byte MTU on PPPoE. I've scoped out what needs doing to add this support, and hope to get round to implementing it and ultimately submitting patches, but I have no timescale for this work as I'm very busy at the moment.

From your sig, do you work for Zen? It looks like AA ISP have implemented their own custom fix for 1500 MTU on routers that don't support RFC 4638. Do you know anything about this?

http://wiki.aa.org.uk/index.php/RouterOS#1500_MTU_ov...

Oh, and PS, I tried to sign up with Zen, but they weren't interested (FTTP).

I don't work for Zen - the details in my signature relate to my connection at our home office (several family members work from home).


The RouterOS link you give is to a kludge - apparently you can force RouterOS to configure a non-RFC compliant MTU and can tell the AAISP end you're going to do so. The correct and supported way to negotiate a 1500 byte MTU is to use RFC 4638.


I'm gradually putting into place the elements needed to add RFC 4638 to pfSense. At the moment I'm slowly setting up the necessary development environment and debugging setup - I got part of this setup working last night. It doesn't help that I'm having to work around which of my machines have jumbo capable network interfaces, though I have now got a jumbo capable interface (an Intel ET dual port server NIC) in my pfSense box.

It's likely to be months before I have it something ready to share with my current workload. If I get it working, I'll submit the kernel patch to FreeBSD, the mpd patch to mpd, and the overall patches to pfSense.


I'm not sure about Zen's policy on FTTP. At the moment, FTTP is a niche product, though it will become more mainstream as BT's commercial deployment speeds up and as FTTP on demand becomes available.

Edited by deleted (Mon 21-Jan-13 14:14:57)

Just a quick update. I'm returning the RT-N66U -- it definitely isn't capable of handling 300 Mbps of PPPoE.

For future generations, here is the summary: I have a 300+ Mbps Internet connection, supplied over fibre (Fibre to the Premises), and requiring PPPoE. I purchased an Asus RT-N66U, thinking it would have ample power to support 300 Mbps, but unfortunately it appears that PPPoE is inefficiently implemented in the Linux kernel, and as a result CPU usage skyrockets to 100%, giving a maximum download speed of around 200 Mbps.

I looked at many options, engaged in discussions on several forums and IRC channels, and these are my findings:

- Asus RT-N66U running stock firmware (I tested two versions: 260, and beta 321). Maximum download speed of around 200 Mbps.

- The RT-N66U running Shibby. It was slower (around 180 Mbps).

- My existing Draytek 2820Vn (which served me well for 4+ years on ADSL), which maxes out at around 70 Mbps. It was never intended for these sorts of speeds. I am now using it as a wireless Access Point.

- I wouldn't mind another Draytek, but the 2130n (the fastest in its price bracket) is almost certainly not capable of achieving these speeds. Draytek Taiwan is reasonably sure it would max out at around 200 Mbps with PPPoE (from a an email someone on the Draytek forum sent to Draytek).

- A vendor (thanks Nick at LinITX!) did a test for me of a RouterBoard RB2011 which has a 600 MHz processor. Even overclocked to 750 MHz (this is a feature of the Mikrotik RouterOS) it was still only capable of 280 Mbps, and that was without any firewall rules (i.e. not apt for real-world use).

- A 4+ year-old Atom motherboard I have lying around (Intel D945GCLF2 with Atom 330, a dual-core 1.6GHz) running RouterOS. It breezed through the test: 310 Mbps at 20% CPU, including NAT and firewall rules enabled.

- Currently as my router I'm running RouterOS on a VM that runs on a Xeon E3-1220 (a quad-core 3.1 GHz) -- and of course it absolutely flies. CPU averages at 0.32%, and hitting the full 300 Mbps sends the VM CPU to a whopping 7.13% (and the VM only has 2 virtual CPUs out of the 4 physical cores available, so this is roughly equivalent to 3.5% of the processor's capability). For those of you about to point how insecure this is, I am passing through a NIC to the VM (using VT-d) for the WAN connection.

This really highlights to me how different processors handle different workloads. The MIPS-type processors used in routers may be good at some things, but are terrible at others; or maybe it is just that the Linux implementations in use aren't optimised. Hopefully as the broadband market continues to produce faster products, manufacturers will produce more powerful routers.

Another fascinating result is that the Atom motherboard idles at 24W, the Xeon idles at 30W, and the RT-N66U idles at 21W. So there is clearly no correlation between the amount of power consumed (at idle), and the raw processing power available in a CPU (I know that the wireless cards in the Asus require a few Watts to power as well). These are all measured using my UK equivalent of a kill-a-watt.

My current plan is to purchase an Intel DN2800MT motherboard (with an additional NIC), which idles at less than 10W.

I'd like to thank everyone for their input, and hopefully these test results will be useful for someone in the future.

Very informative post. Thanks.

Very useful post, glad you found a good solution in the end, I just want to point out though that Routeros is a Linux implementation, packet routing and inspection is very cpu intensive so the key is a fast CPU with multiple cores.

A CCR1036-12G-4S arrived yesterday so I'll be testing that out, 36 cores @ 1.2ghz should be more than enough I imagine!

Hmmm that's a nice piece of kit. Expensive though!