Will A New AC Router Bring You Wi-Fi Joy?

[htismaqe]

Down the near road there is no question these new routers are going to out preform the older 1900 class routers it's just a matter of firmware the hardware is already there much faster processer and much more memory. Once the firmware is nailed down people wont be buying 3 year old routers thinking they are better then the new tech routers. And normally most vendors stop caring about routers after about three years and support gets less and less or stopped all together. I must say though Asus is better at long term support then most. Would i buy a bran new 68U in very late 2015 NO it just dont make since. Just my opinion. Not to mention the AC-3100 and 88U work great and will only get better. So for those that have the means then go for it.

The thing is, the prices of those AC1900 routers are going to start to come down. The price of the R7000 has dropped by over 25% in the last 6 months and will continue to drop after the holidays.

It's all about cost/benefit. You're banking on future capabilities, which is fine. The problem is that these "beasts" don't currently justify the cost because the benefits just aren't there - in fact, as @thiggins showed, some of them actually performed worse than the top of the line AC1900 router.

Also, I just don't see the use cases. As was noted, there aren't any MU-MIMO clients really and Tri-Band seems to be a nitch right now rather than a prevalent use case. Again, those features seem to be "future-ready" but not really "now important".

 

[RMerlin]

That said, MU-MIMO is going to need all the CPU it can get, once devices really come on line and start mixing with SU, N and legacy devices. The beamforming calculations are a b*tch and moreso for MU.

That might explain why their modern wireless SoCs such as the BCM4366 have their own CPU now.

 

[thiggins]

That might explain why their modern wireless SoCs such as the BCM4366 have their own CPU now.

Yup. Although you would know better than I how much processing is offloaded to them.
In the briefing I received on the original xStream architecture, Broadcom said everything was done in the Wi-Fi SoC. All the main processor had to do was hand off TCP/IP packets. Dunno if that actually happens.

 

[sfx2000]

That said, MU-MIMO is going to need all the CPU it can get, once devices really come on line and start mixing with SU, N and legacy devices. The beamforming calculations are a b*tch and moreso for MU.

Indeed... the heavy lifting for MU calculations/decisions will be in the WiFi chipsets themselves, as that is where all the 11ac work is done...

 

[RMerlin]

Yup. Although you would know better than I how much processing is offloaded to them.
In the briefing I received on the original xStream architecture, Broadcom said everything was done in the Wi-Fi SoC. All the main processor had to do was hand off TCP/IP packets. Dunno if that actually happens.

These newer Broadcom routers load an actual firmware that runs on these wifi CPUs. Asus's beta 378_xxxx firmware would log the firmware load operation at boot time, done by the dhd (dongle host driver). I assume that means that most of the work previously done by the wireless driver on the main CPU is now run by that firmware on the offload CPUs. I assume that would include any calculation related to beamforming, (MU)-MIMO, WPA2, etc...

This is at least what I can theorize from what I can observe, as I don't have access to the driver source code to confirm if it's indeed the way it works. But it would all add up.

 

[sfx2000]

I assume that means that most of the work previously done by the wireless driver on the main CPU is now run by that firmware on the offload CPUs. I assume that would include any calculation related to beamforming, (MU)-MIMO, WPA2, etc...

This is at least what I can theorize from what I can observe, as I don't have access to the driver source code to confirm if it's indeed the way it works. But it would all add up.

You are indeed correct...

By the time it gets to the router SoC, it's ethernet - all the heavy Wireless lifting is done by the radios...

 

[faur]

...so...finally...it will or not bring joy?![emoji6]


Sent from my iPad using Tapatalk
Mikrotik(for routing) + AC88U(for 5Ghz) + AC68U(for 2.4Ghz) + Cisco switch(for NAS link aggregation) + AC87U (second AP)

 

[Jeffb]

Asus AC530o has indeed brought me joy. Replaced 2 year old AC68U and coverage is better than expected......covers the whole house from my basement install to the 2nd floor!

 

[System Error Message]

A newer router can provide better coverage but thats also because they come with either better amps or chips or antennas. The AC3200 has better antennas than the AC68U which is why the 5300 has better coverage than the AC68U because even the AC3200 already has better coverage. Take the antennas off the 5300 and put them on the AC68U and you will see better coverage.

Putting coverage aside until internal bottlenecks are sorted out there is no point in going for a newer wifi router like the AC5300 unless your entire network is only wifi. And this is because internally single GMII variant links are still being used to ethernet. There isnt yet a chip manufacturer that has changed this with switches. TILERA doesnt use switches in their SoC so they have a direct wire from the interface chip to CPU which is one reason why it copes well with very high traffic and configurations but mikrotik adds a switch chip to the CCR1009 because it doesnt have enough interfaces but the switch chip only has a 1Gb/s link to the CPU despite providing 4 or 5 ports and this happens with every single switch embedded in every router, at least mikrotik and ubiquiti explains this in their datasheet.

Commercially you would have a 24 port switch with 2 SFP+ or 48 port switch with 4 SFP+ which helps to reduce bottlenecks also i have seen establishments that use normal switches with only gigabit ethernet just combine all the ports together in 2 halves as to eliminate bottlenecks. So even if broadcom has an integrated switch in the SoC, the switch is actually 8 ports but 4 ports used for LAN, one to CPU and 3 normally unused when they could've had 4 ports to CPU and 4 to LAN but this is not the case and it is the same with any router SoC with an integrated switch. In the case of the AC88U it has 4 ports on switch, 1 to CPU, 1 to realtek Switch and 2 unused GMII links. Even though the realtek specifies a dual link only a single link is used. An unusual example would be the RB450G and RB850gx2 which both have a switch chip but the switch chip has 2 links to the CPU which involve eth0 in which eth0 can be unlinked from the switch to use the other available link straight to the CPU. Openwrt has a very good explanation for the RB450G and information on some consumer routers involving switching.

Until the internal bottlenecks are solved we are not going to see improvements in speeds despite the advancements in wifi. Gaming still goes best with gigabit ethernet, NAS is either used on the router or used on gigabit ethernet or multiple gigabit ethernet or 10Gb/s ports. SFP still hasnt made its way to consumer routers yet despite the growth of FTTH and considering the prices of these routers it is very disappointing that you are only paying for 800mhz more in total CPU frequency, faster wifi, a few more chips and nothing else that it now costs the same as a PC.

When you hit a link bottleneck performance decreases even more than using slower interfaces because of having to resend packets.

 

[Jeffb]

My network is all wi-fi. "Take the antennas off the 5300 and put them on the AC68U and you will see better coverage", just out of curiousity, how practical is that? I cannot just "add" 4 more antennas to the existing antennas on the Ac68U! I would also venture that better CPU and design have helped with coverage and throughput, as you point out, "because they come with either better amps or chips or antennas"......no matter how you cut it, in my application, the 5300 is much better...could I have "gotten by" with the AC3200? Maybe, but I chose, not to!
Unequivocally, for me, in my application, the 5300 works much better and I am satisfied. Whether anyone else should spend the money is determined by them, their situation and the research, tests and reviews that they choose.

 

[Deleted member 22229]

There is very little difference in the antennas if anything the radiating element is a little higher in the plastic on the newer antennas basically there are all just simple dipoles no magic there..

 

[sfx2000]

I was expecting better results with the crop of MU-MIMO devices, even in legacy 802.11n (2.4GHz) and SU-MIMO in 802.11ac Wave 1 (5GHz)...

For 2.4GHz, even in Legacy Modes (e.g. nothing above 802.11n standards 64-QAM), one would expect a 4*4:4 radio to offer a bit more range based on analog diversity (4th radio adds a bit, about 2-3 dB, but compared to a 3-stream N450 radio it's about 16 percent), and the coding gains on MIMO - 4RF chains allow for interesting coding chains at the MAC level for redundancy and error recovery with STBC's and LDPC's...

For 5GHz - Wave1 11ac, plus a little bit more - most 11ac AC1900 class devices are 3*3:3 (the WRT1900's are 4*4:3), so again, extra coding gain thru MIMO coding, along with Rx/Tx diversity...

MU schedulers not excepted, even SU-MIMO across current 11n/11ac clients should have been a bit better compared to the current outstanding group of AC1900 class devices...

Odd... but when we look at Tim's results on the VeriWave workbench...

It seems like the chipset vendors have a bit of work to do - even outside of MU-MIMO for Wave2 chipsets.

sfx

 

[sfx2000]

Until the internal bottlenecks are solved we are not going to see improvements in speeds despite the advancements in wifi. Gaming still goes best with gigabit ethernet, NAS is either used on the router or used on gigabit ethernet or multiple gigabit ethernet or 10Gb/s ports.

Most consumer Router/AP's are getting CPU bound with the applications we're putting upon them - it's a hardware and a software limitation...

Hardware - Consider that most 802.11ac AC1900 class routers are Cortex-A9 dual-cores, which is comparable to a smartphone from 2-3 years ago - slightly higher clock rates, but still a slow memory bus (compared to later ARM's), and general task loads that we're asking those two cores to do within the current SW context... going to ARMv7 was a logical progression from the old MIPS cores...

On a SW perspective - I think many consumer grade vendors (outside of Apple and Cradlepoint, which use *BSD) owe a deep appreciation to the open souricng of the Linksys WRT54G GPL dumps - Tomato, Shibby, most OEM builds, even the OEM dev packages are based off that dump, the WRT54G DNA is there - and it's getting old and brittle.. but it's been easy...

 

[System Error Message]

Most consumer Router/AP's are getting CPU bound with the applications we're putting upon them - it's a hardware and a software limitation...

Hardware - Consider that most 802.11ac AC1900 class routers are Cortex-A9 dual-cores, which is comparable to a smartphone from 2-3 years ago - slightly higher clock rates, but still a slow memory bus (compared to later ARM's), and general task loads that we're asking those two cores to do within the current SW context... going to ARMv7 was a logical progression from the old MIPS cores...

On a SW perspective - I think many consumer grade vendors (outside of Apple and Cradlepoint, which use *BSD) owe a deep appreciation to the open souricng of the Linksys WRT54G GPL dumps - Tomato, Shibby, most OEM builds, even the OEM dev packages are based off that dump, the WRT54G DNA is there - and it's getting old and brittle.. but it's been easy...

Actually phones had higher clock speeds. dual core ARM A9 started at 1Ghz and clocked up much easily on phones and unlike the ARM A15 used a lot less power but with the way routers are design the ARM A9 isnt a logical path because if you measure the power usage of your router you will find that it uses a lot of power all the time so might as well use the A15. Back than you had a choice between ARM A8, A9 and A15 for reference designs that were common. A15s were used in servers as well.

 

[scoopnog]

The other comments about different STAs still apply.

That they were different clients was clear in the documentation. What is less clear to me as a consumer is how much credence I should give to the router benchmark rankings if router results can change so much with different clients

Imagine multitudes of others here including yourself are cognizant of this, so it is what it is I guess.

 

[RMerlin]

That they were different clients was clear in the documentation. What is less clear to me as a consumer is how much credence I should give to the router benchmark rankings if router results can change so much with different clients

Imagine multitudes of others here including yourself are cognizant of this, so it is what it is I guess.

That's why these should only be used as general guides, not as definitive information as to what performance you will experience with your particular clients. As 802.11 keeps increasing in complexity, so will interoperability issues. Look at the Quantenna chipset for a good example - works fine with some clients, but constant disconnections with Apple-based clients, and now some Android clients also exhibit a major power drain. Always worked flawlessly with my Intel 7260AC.

 

[thiggins]

What is less clear to me as a consumer is how much credence I should give to the router benchmark rankings if router results can change so much with different clients

Wireless benchmarks are useful only as indication of comparative performance. This is why I make no changes in the testbed including updating firmware on the R7000, or even Windows OS.

No test method can provide an accurate indication of the performance you will experience in your wireless environment with your clients and traffic.

 

[Deleted member 22229]

No real world test here but i was at my neighbors across the street and with my phone i could easily connect to my 3100 and achieved about a 60 mbps download get 90 mbps inside my house so still very usable to say the least. This was done using the 5 GHz band i dont use 2.4 any longer.

 

[sm00thpapa]

No real world test here but i was at my neighbors across the street and with my phone i could easily connect to my 3100 and achieved about a 60 mbps download get 90 mbps inside my house so still very usable to say the least. This was done using the 5 GHz band i dont use 2.4 any longer.

Can you give an estimate on the distance it was to your neighbors house? My 5GHz band drops significantly on my AC88U when I'm out on my patio about 30 feet away and going through a wall and glass french doors.

 

[thiggins]

Kal-el, sounds like a real world test to me. Thanks for sharing