Router Classification Thoughts - 2.4Ghz vs. 5GHz

[sfx2000]

TLDR; Recommend the site base the Rating Classes on the actual performance, and limit the non-standard extensions..

Just some thoughts... and how we can keep vendors honest...

802.11n supports three streams in most implementations - some vendors have put forth some non-standard extensions, for example TurboQAM in 2.4GHz, and ganging up 5GHz 3-stream Radio's to make an "AC3200" class device... BS... That's basically marketing... add up everything, including the non-standard crap, and you've got a really big number... and watch the devices fly off the shelves...

So.. Let's step back a bit...

A 3-steam 802.11n 2.4GHz radio that is "standards compliant" is an "N450" radio, and one that limits channel width to 20MHz (Hello Apple, this is for you!) is an N216 at best...

Going into 5GHz - it goes down the same path - we've got most vendors right now at 3-stream 802.11ac which is "AC1300" - some want to put two together - e.g. Broadcom X-Stream, but basically at the end of the day, from a client perspective, it's not AC2600, much less AC3200, it's AC1300 in the 5GHz band...

The 4-stream implementation from Quantenna is perhaps honest at Ac1600 speeds - 802.11ac MCS9 @ 80MHz channels, each stream being 433Mbits...

So what this means - dear marketeers...

Currently offered - an "AC3200" class device on the shelf is actually an N450/AC1300 (with dual radios, btw), or what is classified right now by SNB standards is an AC1750 device at best. Same goes with every "AC1900" class device, unless they, well, basically they don't, have a solution there - TurboQAM does not count in 2.4GHz

just saying... perhaps we rate them in band class - N450/AC1300 for example - if it's three stream across the bands, can't be any faster, eh? Add a stream to the 5GHz, then it becomes an N450/AC1600 class device.

At present - most of the clients devices you can buy - single stream 802.11n in 2.4GHz on the low-end (and very common actually), perhaps 2-stream there for N300 class - same goes with the 5Ghz band - N150, N300, N450, perhaps AC533/AC867 with AC1300 on the high end...

Reason why I bring this up - people look to the site for reviews and recommendations - and then when they buy that "AC3200" class router, and they hook up their low-end Lenovo/Dell/HP latops with a single stream 2.4Ghz 802.11n card, they are unhappy that they only get 65Mbps at best... how many times have we seen this as the collective "hive mind" and have to tell them - "that's the breaks, bro..."

sfx

 

[sm00thpapa]

Great write up SFX. Only networking educated people like us can understand this. I just watched a youtube video where a guy bought a Netgear X6 and ran a speed test on his Galaxy S5. He had a 50Mbps download speed and was only getting 15Mbps on his phone. Right there he said oh this router is crap and going back. I commented back and said no bro it's your phone it's most likely only a 1x1 chip in there. I hear people every day in Best Buy saying oh AC3200 is the fastest router on the market and the best routers out right now. I saw the Best Buy salesman sell some sweet older lady a Nighthawk X6 and she kept saying she only has an iPad and a cell phone in her home. Best Buy salesman robbed her blind just to make a sale since they all work on commission in Best Buy. Like I said before and still say an AC1900 router should be sufficient for any user right now. Just my opinion.

 

[L&LD]

Salesmen; we won't go there.

Just need to point out that comparing the little old lady with the ipad and cell phone is not the same as 'any user'.

Uses for higher than AC1900 class routers are real. Not all of us have little old lady workflows.

Great write up SFX. Only networking educated people like us can understand this. I just watched a youtube video where a guy bought a Netgear X6 and ran a speed test on his Galaxy S5. He had a 50Mbps download speed and was only getting 15Mbps on his phone. Right there he said oh this router is crap and going back. I commented back and said no bro it's your phone it's most likely only a 1x1 chip in there. I hear people every day in Best Buy saying oh AC3200 is the fastest router on the market and the best routers out right now. I saw the Best Buy salesman sell some sweet older lady a Nighthawk X6 and she kept saying she only has an iPad and a cell phone in her home. Best Buy salesman robbed her blind just to make a sale since they all work on commission in Best Buy. Like I said before and still say an AC1900 router should be sufficient for any user right now. Just my opinion.

 

[thiggins]

AC3200 routers are tested as dual-band routers for the charts. The same AC1900 class test client is used and only the high-band 5 GHz radio is used since the standard 5 GHz test channel is 153.

AC3200 routers get classed differently in the charts so that readers can find them.

To your point about most people having 1x1 or 2x2 clients, I will admit that is a "fault" with the charts. They do show performance higher than most buyers will ever get.

But I don't think testing with lower-order clients is the answer. Buyer education is. And that has always been SmallNetBuilder's focus and will continue to be.

 

[L&LD]

AC3200 routers are tested as dual-band routers for the charts. The same AC1900 class test client is used and only the high-band 5 GHz radio is used since the standard 5 GHz test channel is 153.

AC3200 routers get classed differently in the charts so that readers can find them.

To your point about most people having 1x1 or 2x2 clients, I will admit that is a "fault" with the charts. They do show performance higher than most buyers will ever get.

But I don't think testing with lower-order clients is the answer. Buyer education is. And that has always been SmallNetBuilder's focus and will continue to be.

I agree with buyer education being the answer. No matter how fool proof a rating system is built, nature will always find a way to make a better 'fool' to break it.

 

[azazel1024]

Having the AC3200 classed differently I think is part of the way to go.

I think buyer education is as well.

I don't entirely agree with some of the other points though SFX2000. TurboQAM can be used by a number of clients (most of the newer Broadcom and Qualcomm 11ac client chipsets support 256QAM in 2.4GHz AFAIK). Of all of the extensions to standards, 256QAM in 2.4GHz is probably one of the most widely adopted non-standard standards that has come before. AFAIK, the ONLY company not currently supporting it is Intel. Marvell, QCA and Broadcom are all supporting it on some level, most of them widely with their current basestations and clients. At least in the consumer space (I can't speak to enterprise, but since this site and most consumers aren't going to be buying enterprise gear for their home or small network...)

On the AC3200...yeah, the naming is crap. How would you best name it though? Tri-radio N450/AC1300? You start getting in to rather cumbersome naming conventions. As far as a lone client matters, it is no better than a regular N450/AC1300 router, but if you have a bunch of clients, the second 5GHz radio has a lot of benefits over and above a typical N450/AC1300 router. So it isn't really in the same class.

It comes back to buyer education. It doesn't matter how you label a product, the buyer has to understand that what they get depends on both their router AND on their client. I don't really expect router manufacturers to emphasis that. First of all, they are trying to sell product. So long as they aren't lying or truely obfuscating the heck out of things, I think they are on the up-and-up.

 

[sfx2000]

Well - it's a good discussion... I've raised points, mostly in rant mode, but there's been considerate discourse on the other side of it... all appreciated...

1) Going back to ratings - if a vendor brings a real improvement - e.g. Quantenna with their 4*4 implementation - this should merit rank more perhaps than another vendor tossing together a dual radio 3*3 implementation, and calling it "smart connect" - basically, from an engineering perspective, it really doesn't wash, but the chipset vendor likes this as it allows them to sell twice as many chipsets to folks that only look at SpecSheets...

oh well... and then we can damn them with implementation details perhaps

2) TurboQAM, 802.11ac Features in 2.4GHz - sorry, I have to take sides with the standards bodies here - and I applaud Intel and Apple, who have deployed a significant amount of clients, with not supporting this crap... I've see clients and adjacent AP's crash when seeing these attributes in the beacon frames... The only client vendor that I've seen with widespread support for TurboQAM is RealTek to be honest... TurboQAM in 2.4GHz needs to stop >-now-<, and any vendor that supports it, well, they should get docked for it being non-standard.

Standards are there for a reason - to guarantee interoperability, and we have 15 years worth of legacy to consider with 802.11 b/g/n... clients, AP's, and adjacent networks... some might not care, but I do...

I guess at the end of the day - the current Rating Scale, in some ways, has hit a hard wall, and it's not in favor of the customers/users... it's all in favor of the vendors at this point, and the "buyer education" reasoning doesn't wash post-sales... it just causes a very negative view of the vendors' promises...

sfx

 

[azazel1024]

I haven't played with 256QAM 2.4GHz basestations so far, but I haven't heard of issues with it. Is there a particular reason why it would cause problems? I mean, 11ac means that 5GHz 11n clients are going to see 256QAM in the beacon frames, but they don't support that modulation rate.

Maybe what we need is a new 2.4GHz standard above and beyond 11n to make "mandatory" some of the improvements we've seen in 5GHz. For example mandating EBF support, mandating 256QAM support, mandating LDP error checking (as opposed to FEC), mandating MU:MIMO support, etc.

 

[thiggins]

I guess at the end of the day - the current Rating Scale, in some ways, has hit a hard wall, and it's not in favor of the customers/users... it's all in favor of the vendors at this point, and the "buyer education" reasoning doesn't wash post-sales... it just causes a very negative view of the vendors' promises...

Our testing is not intended to be vendor biased at all. But since it uses a higher class STA than most devices that are actually used with these products, you are correct that it can set unrealistic performance expectations for people who don't read the test method details.

I'll look into putting a disclaimer in the appropriate charts.

You're also right that education after a purchase isn't helpful. That's why everyone should be reading SmallNetBuilder when doing their pre-purchase research.

 

[sfx2000]

I haven't played with 256QAM 2.4GHz basestations so far, but I haven't heard of issues with it. Is there a particular reason why it would cause problems? I mean, 11ac means that 5GHz 11n clients are going to see 256QAM in the beacon frames, but they don't support that modulation rate.

Maybe what we need is a new 2.4GHz standard above and beyond 11n to make "mandatory" some of the improvements we've seen in 5GHz. For example mandating EBF support, mandating 256QAM support, mandating LDP error checking (as opposed to FEC), mandating MU:MIMO support, etc.

I've had a chance to play around a little bit with TurboQAM, and it's diminishing returns in 2.4GHz in many environments... and when using 20Mhz channels, it's a small step (3 stream's it's 260Mbit vs. 216Mbit for example). It's a modest improvement, but in even suburban environments, the noise floor is high enough that the chips fall back to 64QAM pretty quickly..

I think the challenge with bringing some of the other features down is the 15 years worth of legacy in the 2.4GHz space - LDPC is optional (and more common these days), STBC's are more common - I would like to see certain things removed from 11n, like greenfield mode, and better mixed-mode support, as 11ac has shown clearly that it works well in a mixed 11a/n/ac environment - removing Asymmetric MCS would be nice, but I think that would hurt more than help in 11n

MU-MIMO and EBF, in the 2.4GHz space, I think for the most part, isn't going to offer much... again, it's about the old-stuff and interoperability more than anything else.

 

[sfx2000]

Our testing is not intended to be vendor biased at all. But since it uses a higher class STA than most devices that are actually used with these products, you are correct that it can set unrealistic performance expectations for people who don't read the test method details.

I'll look into putting a disclaimer in the appropriate charts.

You're also right that education after a purchase isn't helpful. That's why everyone should be reading SmallNetBuilder when doing their pre-purchase research.

It's a good idea... but let me posit a real world scenario...

Case in point - someone buys a brand new MacBook Pro 2015 edition - Broadwell Processor, fast PCI-E SSD, 3-stream 802.11ac capabilities (N216/AC1300), and then they go out and buy the lastest/greatest Netgear or Asus AC3200 Router/AP... Bob's a smart guy, and he's trying to get the best of everything...

And then they're upset because the fastest the connection is... tada... AC1300... which is as fast as any single radio, we all know, an AC3200 can go... it's topped out, it cannot go any faster...

So the guy has burned 150 bucks over what any AC1750 class router can do... irregardless of the 'classification' - the ASUS AC3200 is at best an AC1750, or better put, N450/AC1300 class router. The AC-87U is honestly an N450/Ac1733 class, not an AC2400 class device.

I won't count TurboQAM/QAM256 in 2.4GHz as N600, as it's non-standard, however, if someone were to release a 4 stream 802.11n in 2.4GHz, that would legitimately be an N600 class capable radio.

That's why I say that the current classification system is flawed... it's misleading at best, and plays to the marketing folks, not to the consumers...

Sorry to be pedantic about things, but honestly, this "arms race" is hurting the customers at the end of the day.

sfx - a grumpy old man, and I've earned the right to be that