Does An AC Router Improve N Device Performance?

[L&LD]

Tim,

Article link:

http://www.smallnetbuilder.com/wire...oes-an-ac-router-improve-n-device-performance

Thanks for this update on this topic.

When I was recommending these (Asus RT-AC66U's, RT-AC56U's and currently RT-AC86U's) routers to my customers, I would add a cautionary 'let's test to see if we see an improvement' fall back clause to my recommendation - precisely because of your original article!

http://www.smallnetbuilder.com/wire...-new-router-really-improve-performance-part-2


I'm glad to see that my recommendations were not without merit. With a router just being a specialized computer, I tend to buy the best I can within my wallets means. This has served well for the last few years.


Hope you enjoyed your crow dinner and thanks again for another well thought out and well written article once again.

 

[thiggins]

I'm still picking feathers out of my teeth...

I wish I could explain why AC routers have these effects. The RT-N66U / RT-AC66U difference in 2.4 GHz is particularly puzzling. Same processor and even same 2.4 GHz radio.

 

[RMerlin]

I'm still picking feathers out of my teeth...

I wish I could explain why AC routers have these effects. The RT-N66U / RT-AC66U difference in 2.4 GHz is particularly puzzling. Same processor and even same 2.4 GHz radio.

Two other factors that can influence performance on top of the chip itself:

1) The PA used. No idea if it's the same for the two routers
2) The CLM profile Asus used when they compiled the driver. Here's an example of what kind of low-level configuration is involved in these CLM profiles:

https://raw.githubusercontent.com/R.../release/src-rt-6.x/wl/clm/src/wlc_clm_data.c

Ultimately, this is what will configure the output power under specific scenarios. I have no idea if Asus used the same profile for both models. So the difference might possibly be coming from this profile rather than the fact it's an AC router.

There are a few additional, low-level settings configured by the CFE and the firmware related to power management. I have no idea how these interact with the whole setup.

 

[Razor512]

I wonder if the wireless drivers in the router have anything to do with it.

From my experience with some 3rd party firmware, I have seen different drivers have very large impacts on the wifi performance.

For example, when testing different builds of DD-WRT for the R6300 v2, Some would function normally while others will have an issue where wifi will be running at around 300mbit/s of actual throughput, and then for a 2-3 seconds, jump to over 500mbit/s, and then drop back down to around 300mbit/s.

It also seems that some wifi routers will drop the connection rate for no reason while some of the newer ones will be less overzealous with dropping the connection rate. For example, an N300 client connected to a WNR3500L V2 will be more likely to drop the connection rate while doing a throughput benchmark than with the same client, connected to the R6300 V2 (at least when using my Atheros AR5B97 wifi adapter

I still have no idea why it happens but even when close, when your wifi client is connected to certain wifi routers, during a benchmark under ideal conditions, the connection just won't maintain the highest connection rate for the duration of a test.

Overall, it just seems that some newer routers are better able to hold a higher connection rate without being so quick to drop the rate.

 

[azazel1024]

This has been my experience as well with testing an AC1750 router and extensive tests in a bunch of locations compared to, what I consider a very good, N600 router. Granted, I have an Intel 7260ac 11ac compatible wifi adpater in my laptop, but the 2.4GHz tests should still carry over exactly.

I find that in my high signal strength locations, I get around 15% better throughput, but in my medium and especially low signal strength locations, I get between 50-100% better performance on the 2.4GHz band down and about 3/4ths of that improvement in uplink.

Example, 2.4GHz 40Mhz in my medium/low signal strength location, I'd get ~4.5MB/sec down with the N600 router, with the AC1750 in the same location I'd get 8-8.5MB/sec. In my very low signal strength location I'd get 2.3MB/sec down, with the AC1750, 3.5MB/sec down.

At a guess, maybe it is other component selection? Whether better amplifiers, better signal processing, or what not, but there is deffinitely a noticable increase in performance just with 11n only testing.

Even testing with my tablet, which is rather limited by its dual band 150Mbps 11n adapter, performance at distance is up on the 2.4GHz and 5GHz bands by a good 30-70% at medium/long distance. Close in, indistinguishable (its an SDIO adpater, limited to 100Mbps minus overhead on the internal bus, which in practice seems to mean around 75-80Mbps absolute max performance).

With some careful testing and playing with it, my range HAS increased too, but that is hardered to be apples to apples. Using the exact same antennas from my N600 on my AC1750 router (both have dettachable external, but 3dBi in the case of the AC1750 and 5dBi in the case of the N600), I find that I can actually extend the connection a little further past where the N600 would drop 5GHz. It is still slow, but at my dinning room table 50-55ft from my router location, up a floor and through 2-3 walls I could not get a connection on 5GHz with my N600 router. With the AC1750 and its 3dBi antennas, I could make a connection, but it was BAD trying to do anything. With the 5dBi antennas, it is still slow, but folder loads over SMB are no longer painfully slow, webpages load noticably faster, etc. It's only about 3MB/sec down and 1.3MB/sec up, but it HAS a connection and a workable one on 5GHz where the N600 router couldn't even connection. Even using the shipping antennas I could get a connection (but I'd only get around 1MB/sec down and a couple of hundred KB/sec up).

2.4GHz is harder to test, but that also seems to be actually slightly greater range. Stepping in to the steps to my garage and sitting on them, I'd lose the 2.4GHz connection to my router (exact opposite side of the house diagonally, roughly 75ft from the router, through a floor SUPER obliquely and 3-5 walls depending on how you measure it). InSSIDer could see it, but windows wouldn't connect (this is the steps attached to the room where my "very low" signal strength testing is done). With the AC1750, windows will connection, but it is effectively unusable with the shipping 3dBi antennas. With the 5dBi antennas off the N600, it is barely usable, but very slow. Around 1MB/sec down and 100-200KB/sec up, with transfers halting for many seconds at a time.

So, I'd say range does seem to have increased too, if not hugely, by a very small amount.

Performance at range, huge difference.

 

[TonyH]

Hi,
Note, -AC routers have heavier AC adapter. IMO, PA has to handle wider b/w and needs more juice....

 

[L&LD]

I'm still picking feathers out of my teeth...

I wish I could explain why AC routers have these effects. The RT-N66U / RT-AC66U difference in 2.4 GHz is particularly puzzling. Same processor and even same 2.4 GHz radio.

I would attribute it to just better design; you know, better attention to the smallest details throughout the radio chain. From the antenna to the microcode in the drivers.

After all, they have learned a thing or two since designing and playing with the RT-N66U level hardware over 3 years ago.

 

[Razor512]

For me, when I upgraded from the WRT54GL to the WNR3500l v2, I got a decent wifi performance and range boost I then modded it with some external antennas and got a decent range and performance boost. Next, moving from the 3500Lv2, to the WNDR4700, offered a huge range boost, in addition to better speed across the range.

I live in a congested wifi environment.

With my WNR3500l, if I start a file transfer or benchmark, my speeds will quickly reach around 25mbit/s, but then over the course of around 30-40 seconds, drop to around 5mbit/s (really why does it have to do that, for the first 10-15 seconds, it was doing 25mbit/s with no corruption or errors to the data that I was copying.

with the newer router, it will quickly ramp up to a similar speed, but only drop to around 22mbit/s

Router makers need to add some user accessible settings where they can have some control over how much of a jerk the wifi adapter will be while you are transferring data.

 

[ChristineBCW]

Thanks for this review. Of course, "Why didn't you include the other 28 brands?!!" won't even be mentioned here. cough cough...

I wonder if there weren't blackened-eyes back at Asus HQ when they hand out team-awards for delivered product - "Mine's better'n yers... nyah nyah nyah..."

I'd love to hear the loser's versions of the Why's, and if they're spiking that other team's punch this Christmas.

 

[htismaqe]

Performance at range, huge difference.

Absolutely agree.

 

[sfx2000]

I'm still picking feathers out of my teeth...

I wish I could explain why AC routers have these effects. The RT-N66U / RT-AC66U difference in 2.4 GHz is particularly puzzling. Same processor and even same 2.4 GHz radio.

Better and newer chipsets, updated drivers, more optimization on the networking stack. Much more recent Board Support Packages (BSP's) with later linux kernels.

It's the trickle down effect of making 802.11ac work in 5Ghz, and as such, it's a selective reduction of certain 802.11n options - it's a much cleaner stack.

Good stuff!

 

[L&LD]

Thanks for this review. Of course, "Why didn't you include the other 28 brands?!!" won't even be mentioned here. cough cough...

What other brands? There are other brands? No way have I seen another brand. What brands? 28? Where???


Lol...

 

[apcsb]

It seems the post should've been called 'Can AC router do better N?'.
What these tests prove is that newer generation AP hardware is better than the previous-gen. Good news, the manufacturers may have opted for cutting costs instead, yet this has nothing to do with .11ac vs .11n. There's just better system rate-over-range (i.e. SNR) and, potentially, driver improvements.

The only way to conduct a valid test would be to take an .11ac router that allows falling back to 'strict .11n' behaviour (simply disabling MCS8/9 is not enough) and actually testing against the same client in both modes. Needless to say, no 2.4GHz tests

 

[L&LD]

It seems the post should've been called 'Can AC router do better N?'.
What these tests prove is that newer generation AP hardware is better than the previous-gen. Good news, the manufacturers may have opted for cutting costs instead, yet this has nothing to do with .11ac vs .11n. There's just better system rate-over-range (i.e. SNR) and, potentially, driver improvements.

The only way to conduct a valid test would be to take an .11ac router that allows falling back to 'strict .11n' behaviour (simply disabling MCS8/9 is not enough) and actually testing against the same client in both modes. Needless to say, no 2.4GHz tests

No, the post is the name of the Article Tim Higgins wrote.

And the tests performed are valid and useful.

Did you read the article?

 

[azazel1024]

It seems the post should've been called 'Can AC router do better N?'.
What these tests prove is that newer generation AP hardware is better than the previous-gen. Good news, the manufacturers may have opted for cutting costs instead, yet this has nothing to do with .11ac vs .11n. There's just better system rate-over-range (i.e. SNR) and, potentially, driver improvements.

The only way to conduct a valid test would be to take an .11ac router that allows falling back to 'strict .11n' behaviour (simply disabling MCS8/9 is not enough) and actually testing against the same client in both modes. Needless to say, no 2.4GHz tests

My Intel 7260ac allows changing to HT mode from VHT (IE 11n from 11ac). In that case 5GHz shows roughly similar improvements to the 2.4GHz band between my Archer C8 and WDR3600. Roughly up to 80% increase in throughput at medium to long range (compared with 110-200+% improvement when switching it over to VHT/11ac mode for 5GHz).

Also I've tested 11n only clients on 5GHz too with similar showings in 5GHz improvement as 2.4GHz improvement (give or take a bit).

There a clear jump in 11n only performance for newer 11ac routers (or so it appears) compared to even very good 11n only routers. Its much more noticable at medium to long range, but there also appears to be a modest boost in performance at short range too.

 

[sfx2000]

The Broadcom 4360 in my Macbook Air (configured as 2 stream) does very well in 11n for 5GHz and 2.4GHz from a client perspective.

I think it comes down to cleaner firmware in the chipsets, more compute horsepower in the chipsets (4360 can do 3-stream in 2.4GHz and 5GHz for 11n/ac), and better drivers.

Most 11ac AP's also have better router system-on-chips, along with more memory (usually these days it's 256-512MB).

(also, from what I've seen in packet traces, 11ac chipsets streamline and disable a lot of HT (11n) features that can impact performance, so their 11n implementation is cleaner there as well).

 

[AdvHomeServer]

I can answer 'YES, performance improves a lot' but only in a left handed way.

I have 2 Netgear routers, R6300v1 and R6300v2. Both have DD-WRT loaded. The R6300V2 is a 5GHz client bridge elsewhere in the house. Media devices are attached via wires. (The 2.4ghz on the media bridge is used for media related purposes using a separate SSID.)

Both 5GHz are configured as N-only because I have no ac client devices in the household.

DD-WRT reports a 877.5 Mbit connection speed between the two routers on 5GHz. A test on the bridge using wires and LanSpeedTest gave me a speed of nearly 300Mbits on the wireless bridge. This would give all 3 attached 10/100 media devices nearly simultaneous full band capability, if needed.

I was surprised. Remember, DD-WRT is set to N-only on 5GHz. (It's possible N-Only really means N+AC in DD-WRT, but it still made my day.)

 

[azazel1024]

Sure about that? Because if it is reporting a link speed of 877.5Mbps, that isn't 11n, that is 11ac.

 

[AdvHomeServer]

Sure about that? Because if it is reporting a link speed of 877.5Mbps, that isn't 11n, that is 11ac.

That's my point. It's kind of a hybrid. Maybe N only really means N+AC with DD-WRT. Maybe there's some kind of weird loophole. 5GHz wireless-N devices work fine but at expected slower speeds. The 2nd router bridged at AC speeds flies.

It negated my need for AC wireless laptop cards for now since I can get AC speeds for some things and still get 5GHz N for others. I thought I was going to have to upgrade all clients to AC to make 5GHz on my AC router useful for anything other than a bridge. Now I can wait, which is good because i have HP laptops that may need whtelisted cards and they may not exist.

 

[sfx2000]

That's my point. It's kind of a hybrid. Maybe N only really means N+AC with DD-WRT. Maybe there's some kind of weird loophole. 5GHz wireless-N devices work fine but at expected slower speeds. The 2nd router bridged at AC speeds flies.

It negated my need for AC wireless laptop cards for now since I can get AC speeds for some things and still get 5GHz N for others. I thought I was going to have to upgrade all clients to AC to make 5GHz on my AC router useful for anything other than a bridge. Now I can wait, which is good because i have HP laptops that may need whtelisted cards and they may not exist.

Huh???

None of this makes sense man...