How Well Do AC Routers Handle Mixed Networks?

[theoak]

With regards to your second generation iPad ...

My third generation iPad (the first with the retina display) can only connect at a max of 65 Mbps on N 5 Ghz. I imagine your second generation would have been the same.

I think 4th generation iPad introduced true 150 Mbps connection rates. The iPad Air introduced true 300 Mbps connection rates. Of course we are talking N in both cases here.

Edit:

Oh ... considering you were using a NETGEAR router ... you could use the genie app's Network Map to determine link speed rates. At least to my laptop what I see via the genie has always been what Windows OS reports as the connection rate. (Why NETGEAR does not include this in the router GUI is beyond me.)

 

[thiggins]

Thanks for the tip on the Genie app. I'll try it.

I don't know why Apple would limit 5 GHz to 20 MHz link rate, as they do 2.4 GHz.
A 1x1 N radio should be able to support 150 Mbps link rate.

 

[offroad]

Just trying to understand N versus Mixed modes and how to get the most speed. Does iPad 2 lock into N when it connects or does it use some mixed mode? Am a beginner to the different transmission speeds and how to optimize it all.

 

[thiggins]

Just trying to understand N versus Mixed modes and how to get the most speed. Does iPad 2 lock into N when it connects or does it use some mixed mode? Am a beginner to the different transmission speeds and how to optimize it all.

The iPad gives you no control over the mode it uses. When used with an N or AC router, it will connect at N rates. Only AC clients can connect at AC rates.

If an iPad is used with an older G router, it will connect at G rates. See How Fast Can Your Wi-Fi Go?

 

[L&LD]

Tim Higgins,

Just a quick thank you for your great website and specifically this latest article, excellent!

I am about to deploy a small/med wireless network and am leaning towards AC class routers myself - this article gives me more reasons that I'm on the right track with my thinking.


What would give me more confidence is a second look at the 1900 AC class showdown between the Netgear R7000 and the Asus RT-AC68U, both with current firmware.

Any plans for this second look in the near future? Or are the upcoming Wave 2 AC routers more interesting than anything we have available now?


Thank you in advance for any answers you can provide.

 

[thiggins]

What concerns do you expect a second review are going to address?

The only additional coverage I plan is to see how the three AC1900 routers perform with TurboQAM (600 Mbps in 2.4 GHz).

I did run a check on RT-AC68U USB file sharing performance. New firmware has not fixed the AC68U USB 3.0 performance problems. The retest I ran using 3-0-0-4-374-371 showed same performance using USB 2.0 or USB 3.0 port.

It looks like the way ASUS "fixed" the problem I had with my standard USB 3.0 test drive disconnecting was to limit the USB 3.0 port to 2.0 speed, same as enable the Reducing USB 3.0 interference setting.

 

[L&LD]

Thanks for the quick repy!

I am not really interested in the USB performance of routers - more for the range/throughput of any router I'm considering - especially for the 2.4GHz range (the 5GHz range seems easily bettered with the newest routers - especially compared to the RT-N66U).

Seems like the latest (501) beta firmware for the RT-AC68U is getting pretty stable for most people; was hoping for your great (objective) testing to see if the results actually are better than the first 'preview' you did with the original shipping firmware.


Also, the 600Mbps 'perk' of these 1900 AC class routers is not what is intriguing to me: WAN/LAN, LAN/WAN and Total Simultaneous are more important to me.


Basically: do the current AC class routers replace the RT-N66U (with RMerlin's SDK5 latest firmware) in each performance aspect? Or is that product still somewhere in the hopefully near future?

 

[sfx2000]

Thanks for the tip on the Genie app. I'll try it.

I don't know why Apple would limit 5 GHz to 20 MHz link rate, as they do 2.4 GHz.
A 1x1 N radio should be able to support 150 Mbps link rate.

I think it really came down to the user experience - Only recently have mobile CPU's had the horsepower and bandwidth to really use more that a single stream on 802.11n (in 2.4 or 5Ghz).

With that consideration, one would be burning up power with little benefit.

Not just iDevices, pretty much any mobile/handheld.

 

[thiggins]

With that consideration, one would be burning up power with little benefit.

What power is being burned? Wireless manufacturers have long made the case that higher wireless link rates help devices save power.

The argument is higher link rate means they can move the same amount of bits in less time, saving airtime. Less airtime means the radio sleeps more, saving power.

 

[L&LD]

sfx2000, I could agree with that in theory: but surely the idevice Tim tested had more than enough horsepower for 40MHz link rate (not the same as dual stream, correct?).

 

[sfx2000]

What concerns do you expect a second review are going to address?

The only additional coverage I plan is to see how the three AC1900 routers perform with TurboQAM (600 Mbps in 2.4 GHz).

I did run a check on RT-AC68U USB file sharing performance. New firmware has not fixed the AC68U USB 3.0 performance problems. The retest I ran using 3-0-0-4-374-371 showed same performance using USB 2.0 or USB 3.0 port.

It looks like the way ASUS "fixed" the problem I had with my standard USB 3.0 test drive disconnecting was to limit the USB 3.0 port to 2.0 speed, same as enable the Reducing USB 3.0 interference setting.

Thoughts would be that USB3.0 is not the limiting factor - problem most likely rests with CPU performance, kernel tuning and the network stack. USB 3.0 is fairly CPU intensive on any platform.

Even with the newer chipsets out there - for example, Broadcom's BCM4707/4708/4709 family, this is a dual-core Cortex-A9 processor @ 1GHz, and most of the AC1900 class devices are memory limited - 256MB is typical, and looking at memory parts, probably on a 16-bit bus at that.

Also consider that most board support packages are built with GCC, and not ADS (ARM's commercial compiler), there's a lot of room for improvement. GCC isn't bad, but it's not as optimized as ADS is...

One is not going to get the same level of performance out of dual-core ARM that one would expect with a modern x86 processor, not yet at least...

 

[sfx2000]

What power is being burned? Wireless manufacturers have long made the case that higher wireless link rates help devices save power.

The argument is higher link rate means they can move the same amount of bits in less time, saving airtime. Less airtime means the radio sleeps more, saving power.

With the current generations of CPU's - e.g.. Apple A6 (and later), Qualcomm's Krait (Snapdragon), Samsung Exynos, and perhaps nVidia Tegra 3, performance is there on the various interfaces, including memory bus, etc... they can make use of the bursting at high speed - the earlier chips, memory performance really wasn't there to make the most of it.

 

[sfx2000]

I'll add a second note for handhelds... most of the chipsets used in this space are multiple mode - e.g. they support WiFi as well as Bluetooth - and then it's a tradeoff - die complexity vs. power - and if the chipset is a thin MAC, you've got additional complexity between having to handle the 802.11 stack as well as the bluetooth stack.

Complexity vs. power is also pretty much the reason why you see single stream only support on handhelds - the main power consumer in 802.11n is decoding the streams - 2 stream takes about 2.5 times the computational power - decode two streams and combine.

Again, goes back to how much CPU horsepower the host has, and what the user experience is expected to be - the bias here is towards the user interface and compute requirements there, not towards making the most of the wireless interface.

sfx

 

[L&LD]

sfx2000, I think you may have missed my post (#10) in this thread.

40MHz bandwidth operation is not the same as two streams (which requires two antennas), correct?

 

[offroad]

The iPad gives you no control over the mode it uses. When used with an N or AC router, it will connect at N rates. Only AC clients can connect at AC rates.

If an iPad is used with an older G router, it will connect at G rates. See How Fast Can Your Wi-Fi Go?

Am most interested in what happens when a client that can only communicate on G speed connects to a router set up as an N speed router. Does that cause all clients to downgrade to G speed? Is your wifi only as fast as your slowest client?

 

[thiggins]

Am most interested in what happens when a client that can only communicate on G speed connects to a router set up as an N speed router. Does that cause all clients to downgrade to G speed? Is your wifi only as fast as your slowest client?

The mechanism is similar to that illustrated for the AC router. G devices are slower than N devices, so require more air time to move the same amount of data.

If both N and G are trying to move data as fast as possible, the effect is that described in Add, Don't Replace When Upgrading to 802.11n.

If G devices are doing email and browsing, the effect will be less. The slowdown occurs only when G and N devices are trying to move data at the same time.

 

[sfx2000]

Hi Tim,

Quick observation - I see in your AC class router tests, something missed is 802.11n/802.11a coexistence.

Would be interesting to see how 802.11ac performs in this environment, where other 5Ghz clients/AP's are adjacent.

One of the features of 802.11ac is dynamic channel allocation and clear channel assessment - 11ac is supposed to be downright neighborly compared to what's been observed in the b/g/n space - many lessons learned from 802.11n with b/g coexistence, and some delta's between legacy 802.11b vs. 801.11a.

The expectation here is that 802.11a/802.11n should work relatively unimpaired when in vicinity of an 11ac AP - in 11n mode, mostly likely one might see two-steam support, as most 11ac chipsets only support two spatial streams at the moment.

So the setup would be - 11ac client with iperf or a Chariot stream running (downlink/uplink/bidirectional), and then bring on the 11a or 11n with a similar stream - the second test case would be similar, except start the 11a/11n streams first.

One would think that the 11ac box would adjust the MCS accordingly while the legacy 11a/11n clients are active, and then step back up to wider channels when the legacy clients are idle.

With the Octobox, it seems like this could be easily scripted.

 

[thiggins]

An 802.11a client is one of the clients used and I ran a test doing specifically what I think you are asking for on page 3 of the article.
I have no way to monitor MCS or link rate during testing. Only RSSI on uplink tests only.

 

[sfx2000]

sfx2000, I could agree with that in theory: but surely the idevice Tim tested had more than enough horsepower for 40MHz link rate (not the same as dual stream, correct?).

Not an "Apple" limitation, it's a Broadcom chipset limitation.

Did some digging - the BCM4329 chipset used in iPad/iPad2/iPad3 is limited to MCS7 - which means no 40MHz wide channel support. This also goes for iPhone 4/4S - which is not relevant as they don't support 5Ghz in any event.

iPad 4th Gen/iPad Mini/MiniRetina/iPad Air and all iPhone 5's (5/5s/5c) support wide channels in 5Ghz for single stream.

Interesting note - Samsung's Galaxy S4 is an 802.11ac single stream device... BCM4335.

sfx

 

[L&LD]

sfx2000, thank you for the response!

Not to be argumentative, but I would still call that an Apple limitation... they are, after all, spec'ing the specific hardware going into their products.