What's coming after AC3200 Routers?

[thiggins]

How is it not true?? Then whats the point of having a 3200mbps routers?? Might as well just stay with 600mbps then.

The point of AC3200 is exactly to provide the ability to separate 5 GHz clients into two groups, slow and fast.

The catch is that for clients on the same radio, the same limitations still apply. Devices with lower maximum link rates limit total available bandwidth to less than the maximum capacity of the radio.

See How Much Throughput Can You Really Get From An AC Router?

 

[funkahdafi]

How is it not true?? Then whats the point of having a 3200mbps routers?? Might as well just stay with 600mbps then.

Because of the way Wifi works. I think in your response to the guy who asked the question you were not talking about AC3200 routers?

Anyways. No matter how much bandwidth you throw at a router, your overall throughput will always be defined by the slowest client on your router. Generally speaking. Wifi can only talk to one client at a time. Now if you have a slow client that is transmitting a big chunk of data, it consumes a lot of "airtime", in which no other client can transmit any data. So even if you throw one million gigabits on a router, as long as a slow 802.11a/g/n client is transmitting data, your fast client, even if it could use one million gigabit, can not transmit any data. It has to wait until the slow client has finished sending it's current "chunk". When slow client has finished it's chunk, it's the fast client's turn. Since it is much faster, it will send it's own chunk of data in very short time. However, then it's the slow client's turn again.

Read this, it explains it very well:

http://www.smallnetbuilder.com/wire...s-xstream-the-coming-battle-for-wi-fi-airtime

As I said in my first response, technologies such as Xstream and MU-MIMO will eventually deal with this. However, there is only one MU-MIMO router out at present (the Asus AC87) but not a single client that supports it. Xstream on the other hand does not need client support for it to work, but it's current implementation (there is only one router yet that does it, the Netgear R8000) is pretty sketchy at the moment.

All I was saying is that generally claiming that "The higher the mbps number, the more devices can share a higher speed over a WiFi." is not entirely accurate. I did not mean to offend.

 

[thiggins]

As I said in my first response, technologies such as Xstream and MU-MIMO will eventually deal with this.

MU-MIMO does not address the mixed client problem.

It provides a mechanism to use multiple RF chains more efficiently but does not treat them as separate radios.

MU-MIMO applies only for downlink (AP to STA) and works only with AC devices.

 

[funkahdafi]

MU-MIMO does not address the mixed client problem.

It provides a mechanism to use multiple RF chains more efficiently but does not treat them as separate radios.

MU-MIMO applies only for downlink (AP to STA) and works only with AC devices.

Yes, you are right. Xstream is the better, albeit more expensive approach.

Maybe next-gen routers will combine MU-MIMO and Xstream. Really interested to see what's coming after the AC3200 routers.

 

[IAAI]

i don't know what's next but i can imagine
routers would look like this

http://cdn.overclock.net/0/0a/350x700px-LL-0a361bca_router.png

 

[azazel1024]

Yes, you are right. Xstream is the better, albeit more expensive approach.

Maybe next-gen routers will combine MU-MIMO and Xstream. Really interested to see what's coming after the AC3200 routers.

Actually not necessarily. It is a big "depends".

The way Xstream default seems to be is that it assigns the first 3 clients to one radio and then the next 3 to the next radio. Or the alt seems to be to assign all 11n clients to one radio and all the 11ac clients to the other.

In either case, there are a lot of scenarios where MU:MIMO can work a lot more efficiently/provide more aggregate bandwidth. Of course it'll require devices supporting it.

I'll certainly grant a dual radio, MU:MIMO solution would be best, but I am sure I can think of even crazier setups where it would work even better. 160MHz, DFS support, dual radio and MU:MIMO with 8 spatial streams anyone? That is about the most overkill I can think of. Oh, and 10GBE ports to support the wireless bandwidth.

 

[thiggins]

The way Xstream default seems to be is that it assigns the first 3 clients to one radio and then the next 3 to the next radio. Or the alt seems to be to assign all 11n clients to one radio and all the 11ac clients to the other.

That is NETGEAR's implementation of XStream. The platform actually has more features, including dynamic radio assignment (5 GHz only).

Other router makers may choose to expose more features.

 

[funkahdafi]

The way Xstream default seems to be is that it assigns the first 3 clients to one radio and then the next 3 to the next radio. Or the alt seems to be to assign all 11n clients to one radio and all the 11ac clients to the other.

Yeah, unfortunately they left it up to the vendors how to actually implement it. Just shuffling the clients around on the radios regardless of their standard kind of defeats the purpose. Not sure why Netgear chose to go that route.

What they should be doing is assign 11n clients to one radio and 11ac clients to another, like you said.

It will be interesting to see how Asus do it on their upcoming AC3200 machine.

 

[thiggins]

Yeah, unfortunately they left it up to the vendors how to actually implement it.

That is done with virtually all end-product implementations of chip vendor technology. The chip guys come up with the reference design (hardware and software) and the product companies decide which parts of it to expose to customers.

Just shuffling the clients around on the radios regardless of their standard kind of defeats the purpose. Not sure why Netgear chose to go that route.

What they should be doing is assign 11n clients to one radio and 11ac clients to another, like you said.

Please (re)read the Smart Connect part of the review. The R8000 is doing precisely that.

 

[funkahdafi]

Ok. I thought I read on a thread about the R8000 here in the forum that the device would put the first three devices, regardless of their standard, on the first radio, and the next three on the other. That's what one user reported anyways.

 

[thiggins]

Ok. I thought I read on a thread about the R8000 here in the forum that the device would put the first three devices, regardless of their standard, on the first radio, and the next three on the other. That's what one user reported anyways.

NETGEAR's implementation is more sophisticated than that.

The reviewer's guide says:

For a new incoming wireless client:

1. Check for load balancing: If the difference between the number of connected clients on each band is >= 3, we connect the new client to the band with less number of clients connected.

Otherwise,
2. Check Client Speed: If the client is an 11a, 1x1 11n, or 2x2 11n, we connect the new client to band 4.

Otherwise,
3. Check Signal Strength: We check the signal strength of the new client, if the signal strength is low, we connect the client to band4, otherwise we connect the client to band1.

Band 1 refers to channel number: Channel 36 to channel 48
Band 4 refers to channel number: Channel 148 to channel 161

For example: Typically if load balancing does not kick in:
- 802.11ac and 3x3 802.11n devices at short range will be on Band 1
- All other WiFi devices including 802.11ac and 3x3 802.11n devices at long range will be in Band 4

 

[KGB7]

That is a huge dissapontment.

I know 2.4GHz band had limitations with mixed b/g/n clients. But after a decade of development, we are still dealing with same issue in 5ghz band with n/AC clients.

Might as well just use a bunch of cheap 5ghz APs.

 

[funkahdafi]

This:

For example: Typically if load balancing does not kick in:
- 802.11ac and 3x3 802.11n devices at short range will be on Band 1
- All other WiFi devices including 802.11ac and 3x3 802.11n devices at long range will be in Band 4

just sucks. Mixing 11n and 11ac clients on the same radio defeats the purpose of Xstream.

 

[sm00thpapa]

This:



just sucks. Mixing 11n and 11ac clients on the same radio defeats the purpose of Xstream.

Agreed! Put all N on 1 band and all AC on 1 band. And if I was the FCC I would get rid of ABG devices once and for all. If Microsoft can end Windows XP so can the 2.4GHZ band. That's just my opinion cause I know some people will have an issue with what I just said.

 

[azazel1024]

Agreed! Put all N on 1 band and all AC on 1 band. And if I was the FCC I would get rid of ABG devices once and for all. If Microsoft can end Windows XP so can the 2.4GHZ band. That's just my opinion cause I know some people will have an issue with what I just said.

So...get rid of the 2.4GHz band? Or do you mean just get rid of a/b/g clients? For the later...well, pretty rare to find a new a/b/g client these days, so at some point, yes, there won't be any new ones made at all.

For getting rid of the 2.4GHz band...that is like saying because you don't like mini-vans and you live in the city and they cause a problem for parking, we should get rid of all mini-vans.

Lots of people NEED 2.4GHz and lots of people have no issues with 2.4GHz and have no congestion.

Lets not talk about wireless eugenics thank you very much.

 

[htismaqe]

Methinks you've stumbled upon the next great money grab.

Let's abolish the 2.4Ghz band and force everything to 5Ghz.

Then the consumer wireless companies can sell 2 or even 3 devices for every one they sell now because you'd have to have additional APs to cover anything larger than a freaking loft apartment.

 

[azazel1024]

Methinks you've stumbled upon the next great money grab.

Let's abolish the 2.4Ghz band and force everything to 5Ghz.

Then the consumer wireless companies can sell 2 or even 3 devices for every one they sell now because you'd have to have additional APs to cover anything larger than a freaking loft apartment.

To be fair, the AC1750 router I have now, sitting in my basement, at the corner of my house, can cover the rest of my house with 5GHz, with the exception of a bit of my dinning room/living room and my playroom. The playroom in part because it is slightly too far away, but also because there is a 4ft thick cinderblock and stone chimney blocking half the room from the router.

If the router could be centrally located (can't really), it probably could cover the entire house, maybe with the exception of some of that playroom because of the chimney. Its a 2,500sq-ft rancher (including basement).

Though the garage would be screwed no matter what (unless you placed the router in the playroom next to the garage, but then you couldn't cover the rest of the house).

2.4GHz really can cover the entire house, even with the router on the far side of the house (though again, garage is screwed, but with central location it wouldn't be). I can still get ~3MB/sec sitting in the playroom from the router ~60ft, one floor, a couple of interior 2x4" walls and the 4ft chimney in the way, away from my laptop.

Of course being able to cover it and covering it well are two difference things, which is why I have an AP sitting in the playroom next to the TV, which covers the playroom, garage and the living room/dinning room on the other side of the wall. So minimum speeds in the house on my laptop are 10MB/sec in the worst location and generally >20MB/sec (if not >>>20MB/sec). My tablet isn't as fast as it peaks at 10MB/sec in the best location, but it also doesn't drop below 5MB/sec in the worst location.

For really great 5GHz coverage...I kind of do need a second AP inside the house for 5GHz only. I am kind of hoping someone will come out with a halfway decent wall wart style access point that is AC1200, instead of just these AC750 ones. I'd love to get one, disable 2.4GHz on it and try to somewhat centrally locate it on the main level. Though I am hoping a slight repositioning of my current AP might help out with things, that or switching to an AC1750/1200 router with beamforming and good external antennas.

My N600 has excellent 2.4GHz range and pretty decent 5GHz range, but 5GHz performance does drop of resonably quickly. My AC1750 has very slightly better 2.4GHz and 5GHz range, but 5GHz performance stays VERY strong up until you are edging up on maximum range, instead of starting to drop off quickly like the N600 (even if you don't lose the connection entirely on either one till almost the same location. The N600 you might get 3MB/sec near max range, where as the AC1750 you might get 10MB/sec, move a little further and both basically disappear. Move a little closer and the N600 goes to maybe 5-6MB/sec, but the AC1750 goes to 20-25MB/sec).

 

[thiggins]

This just sucks. Mixing 11n and 11ac clients on the same radio defeats the purpose of Xstream.

That's not true.

802.11ac uses many of the same throughput enhancement techniques that were introduced in 802.1n. It achieves higher throughput primarily by adding 256-QAM modulation, more spatial streams (from bonding more channels) and MAC enhancements that enable packing more bits into a frame. So N and AC devices are much better partners in mixed environments because they can communicate the overhead parts of their frames at the same faster rate.

The main issue is difference in speed. Either flavor of device with a lower signal level will require more airtime to move the same amount of data. That is why XStream assigns devices to radios based on airtime use, not technology.

Put another way, the airtime requirement of 3x3 N and 1x1 AC STAs at highest supported rates (450 Mbps and 433 Mbps) is about the same.

NETGEAR's current assignment rules don't represent all that XStream is capable of. It's just their interpretation of what works best for the mix of devices right now.

Will someone else do channel assignment a different way? Can NETGEAR evolve its channel assignment rules? Yes and yes.

 

[sfx2000]

That's not true.

802.11ac uses many of the same throughput enhancement techniques that were introduced in 802.1n. It achieves higher throughput primarily by adding 256-QAM modulation, more spatial streams (from bonding more channels) and MAC enhancements that enable packing more bits into a frame. So N and AC devices are much better partners in mixed environments because they can communicate the overhead parts of their frames at the same faster rate.

The main issue is difference in speed. Either flavor of device with a lower signal level will require more airtime to move the same amount of data. That is why XStream assigns devices to radios based on airtime use, not technology.

Put another way, the airtime requirement of 3x3 N and 1x1 AC STAs at highest supported rates (450 Mbps and 433 Mbps) is about the same.

NETGEAR's current assignment rules don't represent all that XStream is capable of. It's just their interpretation of what works best for the mix of devices right now.

Will someone else do channel assignment a different way? Can NETGEAR evolve its channel assignment rules? Yes and yes.

This is a great comment - with 5Ghz, things are going a lot better than the 802.11n deployments in the 2.4GHz space - 802.11ac fixed a lot of issues with 11n and the various options it had...

From first hand experience - 11ac AP's, regardless of chipsets, tends to be very well behaved in a mixed mode environment of 11a/11n/11ac clients across single/two/three streams on the client side. There is a bit of a performance hit when running mixed mode clients, but not as severe as it was/is in 2.4GHz space.

This isn't really a netgear issue - they just define requirements and put plastic things into cardboard boxes to put on store shelves for people to buy... and those requirements have more than just a technical basis, there's the business considerations - hence we sometimes get half-baked solutions on the cutting edge as we have right now.

The real effort is on the part of the chipset vendors - Broadcom/QCAtheros/Marvell/Ralink/Mediatek...

sfx