1 Gbps Internet - Possible over WiFi?

[RoryH]

triband means 3 bands. 3 channels means 3 channels. Triband means it has 3 radios, 3 channel means the radio uses 3 channels. The ac68U which is what i actually did my test with is capable of close to its theoratical rating (1300Mb/s on wifi AC) on synthetic torture tests but not practically. Each channel requires its own frequency space and provides bandwidth. Single channel wifi N gives 150Mb/s. Single channel wifi AC gives 433Mb/s. dual channel wifi N gives 300Mb/s. This has nothing to do with triband, its to do with the radio on the router itself. The more channels the bigger the frequency space needed (dual channel wifi N already uses most of the available 2.4Ghz spectrum which is crowded). 3 channel wifi AC uses 80Mhz of space in the 5Ghz band so its important that you find unused channels.

Regarding the theoretical max of 433 Mb/s per single channel AC, on the router charts filter on this website, is this regarded to as the 5 GHz Downlink Profile (this is for a single channel right?)?

If so, I would assume that the higher the number on those charts, the better it would be for achieving a higher download speed. And using my 3x3 adapater, it should be 3x the downlink profile?

 

[System Error Message]

Regarding the theoretical max of 433 Mb/s per single channel AC, on the router charts filter on this website, is this regarded to as the 5 GHz Downlink Profile (this is for a single channel right?)?

If so, I would assume that the higher the number on those charts, the better it would be for achieving a higher download speed. And using my 3x3 adapater, it should be 3x the downlink profile?

yup. 3 channel is 3x 433Mb/s giving almost 1300Mb/s theoratical.

 

[pete y testing]

Regarding the theoretical max of 433 Mb/s per single channel AC,

thats the max sync rate , and yes 3 x is what you get on a 3 stream router , on a 4 stream router you get 4 x , but again thats just max sync

So I don't understand when you say domestic routers don't achieve 1 gbps. Do you mean they don't hit the exact 1024mbps? It's fine I'm not aiming for the exact max.

correct , you may be fine with it but the fact remains users will not get the max speed they are paying for at this time on any domestic router , they may get close but no cigar

 

[sfx2000]

yup. 3 channel is 3x 433Mb/s giving almost 1300Mb/s theoratical.

That's the PHY layer link rate - the actual network rate counting overhead is less... AC867 generally will translate to about 500-550 Mbit/Sec at the IP layer...

Which goes back to my original statement - consider about 250 to 275 Mbit/Sec per spatial stream in use with 802.11ac - and that's not bad... a three stream radio connection should get one right around 750 to 850 Mbit/Sec best case...

To get a "gigabit" - one would have to have 4 spatial stream support for both the AP and the client station...

 

[pete y testing]

To get a "gigabit" - one would have to have 4 spatial stream support for both the AP and the client station...

even then with an asus rt-ac88u and asus pce-ac88 im getting about 100Mb/s on read but more like 70MB/s on write and that in the same room throughput wise , this is from a giga ethernet connected device to the 4 x 4 wifi connected device , it is however still bottlenecked by the wal to lan fir internet connectivity

 

[sfx2000]

even then with an asus rt-ac88u and asus pce-ac88 im getting about 100Mb/s on read but more like 70MB/s on write and that in the same room throughput wise , this is from a giga ethernet connected device to the 4 x 4 wifi connected device , it is however still bottlenecked by the wal to lan fir internet connectivity

Talking apples and oranges perhaps - my numbers assume gigabit connectivity between the iperf3 client and iperf3 server - so this takes WAN and routing out of the picture completely - we're measuring performance across the AP to the client...

Restating... and iperf3 is pretty much a good indication of potential performance - with a 2 spatial stream connection (AC867), one gets about 250Mbit/Sec per stream - do the numbers, and 1 gigabit/sec is realizable across the WLAN interface if both clients are 4 spatial streams.

(Routing between WAN/LAN is going to be impacted by the performance of the Router's SOC and SW stack, but OP was asking if gigabit WiFi is possible, and my numbers indicate that yes, it is...)

Upstream

- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bandwidth
[ 4] 0.00-10.00 sec 597 MBytes 501 Mbits/sec sender
[ 4] 0.00-10.00 sec 596 MBytes 500 Mbits/sec receiver

Downstream

- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bandwidth Retr
[ 4] 0.00-10.00 sec 635 MBytes 532 Mbits/sec 0 sender
[ 4] 0.00-10.00 sec 634 MBytes 532 Mbits/sec receiver

 

[sfx2000]

As and a point to compare - Gigabit Ethernet over Thunderbolt to 1GbE on the same MacBook Air...

Upstream
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bandwidth
[ 4] 0.00-10.00 sec 1.08 GBytes 932 Mbits/sec sender
[ 4] 0.00-10.00 sec 1.08 GBytes 931 Mbits/sec receiver

Downstream
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bandwidth Retr
[ 4] 0.00-10.00 sec 1.08 GBytes 931 Mbits/sec 0 sender
[ 4] 0.00-10.00 sec 1.08 GBytes 931 Mbits/sec receiver

 

[pete y testing]

we're measuring performance across the AP to the client...

yup and thats where i, measuring the throughput from wifi 4 x 4 client to ethernet nas using simple windows transfer

 

[sfx2000]

And at some point - speedtest get's somewhat imprecise... same connection - same host - speedtest mini...

 

[sfx2000]

even then with an asus rt-ac88u and asus pce-ac88 im getting about 100Mb/s on read but more like 70MB/s on write and that in the same room throughput wise , this is from a giga ethernet connected device to the 4 x 4 wifi connected device , it is however still bottlenecked by the wal to lan fir internet connectivity

The problem with doing SMB transfers is filesystem overhead - and that isn't a good measurement for the link - iperf3 just pushes packets, so it's a better measurement.

Same goes with FTP - which is what many 2G/3G providers used to use (esp with CDMA, as their CDG3 carrier acceptance test plan explicitly used FTP - I co-wrote a whitepaper discussing why this was a bad idea)

 

[pete y testing]

and that isn't a good measurement for the link

but it is a good measurement of real world expectation and throughput under what most would consider normal practice and usage

until recently a few manufactures where basing their throughput tests on ftp which uses a different protocol and thus differing results to what most would consider normal transfer measuring , yes i agree smb doesnt show what the link is possible of at its best , but it does show what any user can expect with the same setup and conditions , eg as in "i just got the x router and it only does this but i have seen reports x fast , file transfers imho are the best way to give a real world example of what can be expected

 

[sfx2000]

but it is a good measurement of real world expectation and throughput under what most would consider normal practice and usage

I think we need to respectfully disagree here...

SMB is a repeatable benchmark across similar platforms - but it doesn't actually measure thruput across the link itself...

What if one end is a Raspberry Pi and the other end is a Xeon Server running Windows Server 2016?

Then your benchmark is relative to your use case - and it doesn't actually measure the link state performance...

In that same use case - iperf3 will still accurately measure performance across the link - taking away application and file system IO differences.

 

[L&LD]

Definitely sounds like a greater device. Thanks for the post. I wont be having the issue of range though, as its a relatively small flat.

Given that the difference in price between the 68U and AC3100 is 140, do you really think I would see that much of a boost between the two sitting at 5-7 meters away on the 5Ghz band to justify the price? I know you might probably not have the answer to that cause of the varying factors. I guess I will have to evaluate my budget.

For me, the difference to get the RT-AC3100 is around $400 here. And it is 'worth it' to me, given even the small differences I saw as outlined in the link I provided. There is only one thing stopping me (if only the cash was available).

I too live in a smaller home, but highest performance is what I chase (with the few dollars I have).

As already noted, the RT-AC3100 with it's 4x4:4 antennae/streams, fast dual core processor and double the ram and NVRAM of any other Asus router currently available is the best chance to achieve as close as possible to what you were asking for in the first post/title thread.

Yes, you can save a little money (a few trips to the local fast food joint). But what's the fun in that?

Glad I could provide a little info for you to chew on.

 

[sfx2000]

SMB is a repeatable benchmark across similar platforms - but it doesn't actually measure thruput across the link itself...

What if one end is a Raspberry Pi and the other end is a Xeon Server running Windows Server 2016?

Then your benchmark is relative to your use case - and it doesn't actually measure the link state performance...

And for that use case - you're measuring the application and not the link itself...

If one wants to measure SMB performance, then your use case is valid, but in this thread - is WiFi capable of a 1Gb/Sec link, it's not...

 

[sfx2000]

As already noted, the RT-AC3100 with it's 4x4:4 antennae/streams, fast dual core processor and double the ram and NVRAM of any other Asus router currently available is the best chance to achieve as close as possible to what you were asking for in the first post/title thread.

I think in the current Asus 2016 line-up - the RT-AC3100 is a good bet - better than the other units...

Still the AC1900 class is the sweet spot for WiFi performance, and there, within the Asus camp, the RT-AC68 series is a good value... it's problems are well known and worked around, stable 3rd party FW for those that want to go down that path, and a wealth of community insight and knowledge around that product.

 

[L&LD]

I think we need to respectfully disagree here...

SMB is a repeatable benchmark across similar platforms - but it doesn't actually measure thruput across the link itself...

What if one end is a Raspberry Pi and the other end is a Xeon Server running Windows Server 2016?

Then your benchmark is relative to your use case - and it doesn't actually measure the link state performance...

In that same use case - iperf3 will still accurately measure performance across the link - taking away application and file system IO differences.

but it is a good measurement of real world expectation and throughput under what most would consider normal practice and usage

until recently a few manufactures where basing their throughput tests on ftp which uses a different protocol and thus differing results to what most would consider normal transfer measuring , yes i agree smb doesnt show what the link is possible of at its best , but it does show what any user can expect with the same setup and conditions , eg as in "i just got the x router and it only does this but i have seen reports x fast , file transfers imho are the best way to give a real world example of what can be expected

sfx2000, I would respectfully disagree with you on this. Firmly on pete y testing's side on this one.

Real world performance numbers are the only important aspect to measure. Sure, it is up to the reader to decide if they are applicable to them and their situation, but it is the more 'real' number to me.

I don't really care if the theory indicates that a certain absolute metric is possible. Real world results take the theory, the overhead and everything else in this 'real world' we live in and give us a number that is not over stated.

SMB may be a repeatable benchmark for users across the world. That doesn't help anyone though trying to ascertain if a certain speed can be reached in their normal workflows, nor how it will compare with wired GbE connections either (i.e. the overhead of WiFi).

Not everyone knows the nitty gritty of networking (wired/wireless) like you do. A real world number takes 'everything' into consideration. No guessing needed.

 

[sfx2000]

sfx2000, I would respectfully disagree on pete y testing's side on this one.

Real world performance numbers are the only important aspect to measure. Sure, it is up to the reader to decide if they are applicable to them and their situation, but it is the more 'real' number to me.

I don't really care if the theory indicates that a certain absolute metric is possible. Real world results take the theory, the overhead and everything else in this 'real world' we live in and give us a number that is not over stated.

Change the verbiage to "respectfully" and I might agree with you, or at least not take umbrage with your comment.

In my many roles within the wireless industry - this one is a sore point - and that call performance - and it's lay people within uninformed opinions that state subjective performance and not consistent and repeatable/measurable performance that made me go down that path...

Take the variables out - and measure the true statistics... those are facts that can be proven, over and over again - otherwise, it's subjective opinion, and cannot be reproduced or duplicated in a consistent manner..

 

[L&LD]

Change the verbiage to "respectfully" and I might agree with you, or at least not take umbrage with your comment.

In my many roles within the wireless industry - this one is a sore point - and that call performance - and it's lay people within uninformed opinions that state subjective performance and not consistent and repeatable/measurable performance that made me go down that path...

Take the variables out - and measure the true statistics... those are facts that can be proven, over and over again - otherwise, it's subjective opinion, and cannot be reproduced or duplicated in a consistent manner..

sfx2000, my apologies for the small 'dis' (I have since edited/corrected my post above). Can I blame auto correct (please)?

I'm not against measurable and repeatable performance 'statistics'. But they are not the only metric worth discussing.

Particularly if a person has taken pains to fully outline their 'real world' setup so that others can duplicate his/her findings too.

 

[sfx2000]

sfx2000, my apologies for the small 'dis' (I have since edited/corrected my post above). Can I blame auto correct (please)?

I'm not against measurable and repeatable performance 'statistics'. But they are not the only metric worth discussing.

all good - autocoorect gets the best of many

 

[RoryH]

Does anyone have an opinion on the TP-Link C3150? I've been having a look at that and the Asus 88U for my 1 gigabit connection.

I wont be getting a 4x4 adapter yet, so I will just be trying to get the best and fastest possible connection over my 3x3 network adapter.

The C3150 has top scores across the downlink charts on both the 2.4 and 5 GHz. It also matched and did slightly better than the 88U on the review on this site.

Also:

Test Description TP-LINK Archer C3150
WAN - LAN TCP (Mbps) 941
LAN - WAN TCP (Mbps) 941

Test Description ASUS RT-AC88U
WAN-LAN (Mbps) 802
LAN-WAN (Mbps) 791

Does this mean that the 88U can only just about max out at 800 Mbps wired? That would be a bit surprising since my cheap ZTE supplied router can hit higher (870).

The only thing is, I was reading PC Mag's review here (http://uk.pcmag.com/tp-link-ac3150-...-wireless-mu-mimo-gigabit-router-archer-c3150) and the 88U clearly does amazingly better.

Does this mean that the charts on SNB are not a reliable gauge for actual home use? But how did the TP Link do so well here?

Lastly, going back to my original question, what do you advise would be the fastest router / best option for achieving the highest possible wireless speed (with a 3x3 network adapter)?