RT-N66 with Merlin 380.57 firmware max throughput

[ALW]

Hi there, first post here so be gentle

I've been noticing some speed issues recently whereby the full 75Mbps of my broadband connection was being throttled by the above router, to around 61Mbps. A quick test plugging a PC into the VDSL modem, bypassing the Asus proved this.

A little digging on here made me think that enabling HW acceleration might be worth a try. After enabling this by disabling per-IP monitoring and QoS the full bandwidth is available again.

This left me wondering whether in the real practical world, when multiple devices are using bandwidth, as opposed to just one, is this difference an overall limitation to the network, i.e. if I had two devices, demanding let's say 35Mbps each, would the router deliver this, or would the above cap apply due to internal hardware performance issues?

I realise I could do an experiment but I thought someone one here might have the answer; internally within the router where is the limitation, is it just a CPU / processing limitation, or is it a per-port limitation?

I suppose this is all leading to questions about max speeds vs. enabling QoS to prioritise bandwidth for certain services. I'm toying with the idea of upgrading to a newer Asus, particularly after setting up a friend's RT-AC87 recently, but this elderly device is still amazingly good and by far the best router I've used (particularly with Merlin's awesome firmware).

Thoughts?

Andy.

 

[L&LD]

First rule of thumb is that you can never ever have enough processor power.

Second is that each and every feature you enable will be a hit to performance. Including some not so obvious 'features' (such as enabling 6 additional ssid's, for example).

Third is that multiple clients will multiply the negative effects above (always).

In my case, I (finally) jumped to an RT-AC68U router (original 800MHz dual processor, not the latest revisions with dual 1GHz processor) from an RT-N66U and was pleasantly surprised that while not earth shattering, the benefits are noticeable in my environment. And my ISP levels are significantly below yours too.

http://www.snbforums.com/threads/is-the-rt-ac68u-a-good-replacement-for-a-rt-n66u.30303/#post-237039


After getting a good baseline feel for the last month (approx.) with this new hardware, I decided to see if dslreports or my normal usage of my network would show any improvement by enabling Adaptive QoS.

http://www.dslreports.com/speedtest


At first, I left Adaptive QoS on it's 'auto' settings and yes, the improvements were noticeable on the dslreports stats page (the 'my results' tab, after the test has finished). I then tried the manual settings and the suggested 80% input of my upload and download speeds.

Finally, I decided to make my actual upload/download speeds be the '80%' targets instead (I input 14/70 for my actual 11/56 U/D speeds in the gui). This is the setup I'm running on now and it has given me the best overall performance and responsiveness for my uses.

I was already getting A, A, A for my Bufferbloat, Quality and Speed ratings, but I now get A, A+, A with the above options.

When I did a true limit of 80% of my ISP speeds those were A+,A+,A, but I do not feel any effective difference in latency, but I am enjoying faster download and upload speeds with my 25% exaggerated speeds that I input in the QoS Adaptive (manual) gui. Depending on the time period I test, I can hit close to 58Mbps down and 12.5Mbps up for long periods of time.

If you have the money to buy a new router, the upgrade will probably be worth it for you too.

The newest revision RT-AC68U's are shipping with an even more powerful 1GHz (dual) processor. Or, if your budget allows; the RT-AC3100, RT-AC88U or the RT-AC5300 all sport 1.4GHz (dual) processors to make your network as optimized as possible.

Staying with the RT-N66U is not a bad decision either at this time. Newer Asus models should be introduced or become available soon. But in any event, an AC class router with (preferably) dual 1GHz processors or faster is undoubtedly needed to get the best network performance for anyone with a true high speed ISP of at least 50/10 Mbps or more.

 

[ALW]

...http://www.dslreports.com/speedtest...

Thanks for taking the time to write such an amazingly detailed post. I wasn't aware of the dslreports speedtest so will have a play with that and understand the different results it provides!

 

[DroidST]

I then tried the manual settings and the suggested 80% input of my upload and download speeds.

Finally, I decided to make my actual upload/download speeds be the '80%' targets instead (I input 14/70 for my actual 11/56 U/D speeds in the gui). This is the setup I'm running on now and it has given me the best overall performance and responsiveness for my uses.

I was already getting A, A, A for my Bufferbloat, Quality and Speed ratings, but I now get A, A+, A with the above options.

When I did a true limit of 80% of my ISP speeds those were A+,A+,A, but I do not feel any effective difference in latency, but I am enjoying faster download and upload speeds with my 25% exaggerated speeds that I input in the QoS Adaptive (manual) gui. Depending on the time period I test, I can hit close to 58Mbps down and 12.5Mbps up for long periods of time.

@L&LD So, is your actual broadband U/D 11/56 or 14/70?

It looks like your actual is 11/56 and you entered higher numbers for manual under Adaptive QoS. I thought we should enter 90-93% (per @RMerlin ) of our actual so in your case, 10/51

My advertised Comcast U/D is 10/105 but provisioned (and at what I actually achieve 12.5/128) at 12.5/131.25. Comcast delivers ~25% more than advertised rate.

 

[L&LD]

I pay for 50/10 but get significantly higher.

 

[DroidST]

I see, like what Comcast does. So you did enter in QoS manual 14/70.

What actual performance over U/D 14/70 did you get (I assume you checked with QoS disabled)?

 

[mstombs]

My N66 can speedtest at 220Mbps with the Broadcom CTF enabled and real ISP connection, and no QOS IP accounting etc, without CTF max speed was 120-150 depending on options and firmware (Tomato etc). This is with wan type dhcp - I assume pppoe has higher overhead? My current connection is very asymmetric 10/150 so should probably be using QOS which would limit the download - so will upgrade to a dual core ARM router at some point.