Not getting full bandwith of internet connection on Wireless N

[killeraxemannic]

I have done a few speed tests lately after upgrading my internet connection to the Comcast extreme 105 on my wireless and I have found that I am not getting my full download speed over wireless on my RT-N65R. I have 2 wireless devices, a Laptop and a Desktop and neither of them can get my full 120 mb/s down I get on my hardwired desktop on 2.4 or 5ghz wireless. On 2.4 they will get about 45 mb/s down and on 5ghz they will get about 65 mb/s. I have had the laptop within a few feet of the router when testing so I know that distance is not an issue. My network is also the only 5ghz network visible so there is no interference. I also have an older Dlink Dir 655 2.4ghz router I tried plugging in and I got about the same results on 2.4 ghz as with my Asus Router.

I am beginning to wonder if the real maximum speed of wireless N is around 65 mb/s? That seems strange as both my desktop and laptop have cards rated at 450/300 5/2.4ghz cards in them and are definitely connecting at those data rates when connected to the router according to windows. I would love to try AC but my budget doesn't support it right now.

 

[stevech]

I have done a few speed tests lately after upgrading my internet connection to the Comcast extreme 105 on my wireless and I have found that I am not getting my full download speed over wireless on my RT-N65R. I have 2 wireless devices, a Laptop and a Desktop and neither of them can get my full 120 mb/s down I get on my hardwired desktop on 2.4 or 5ghz wireless. On 2.4 they will get about 45 mb/s down and on 5ghz they will get about 65 mb/s. I have had the laptop within a few feet of the router when testing so I know that distance is not an issue. My network is also the only 5ghz network visible so there is no interference. I also have an older Dlink Dir 655 2.4ghz router I tried plugging in and I got about the same results on 2.4 ghz as with my Asus Router.

I am beginning to wonder if the real maximum speed of wireless N is around 65 mb/s? That seems strange as both my desktop and laptop have cards rated at 450/300 5/2.4ghz cards in them and are definitely connecting at those data rates when connected to the router according to windows. I would love to try AC but my budget doesn't support it right now.

Per the 'net, Comcast extreme 105 is supposed to be 105Mbps (down, what is the upstream?).

I may be judged in the minority, but I think that consumer/residential users are not getting value when they pay for over 50Mbps down. Unless there's a real deal.


I'd think it would be rare to find Internet hosts that will sustain 100+ Mbps to one user. Occasional bursts, yes. But the average would be a lot lower, due to the Internet and the bandwidth that web servers will allow per user/session.

Re WiFi and your 105Mbps Internet. Perhaps an ideal Speedtest.net host will let you get 105Mbps sustained for a brief test, so you can confirm, using a new/fast PC with a wired LAN connection.

Then on to WiFi. The WiFi connection speed needs to be about 150% of your goal net throughput, e.g., 1.5 times 105Mbps if that's what you really get when using wired connections. In addition to the WiFi connection speed at 1.5 times, you need little or no competition for a clear channel air time among neighbors' WiFi that is within 3 channels of yours (in 2.4GHz), AND neighbors that really use a lot of air time. (Number of SSIDs detected isn't important).

Remember that WiFi is "half-duplex". Ethernet is (usually) full duplex. This means that when pulling a big stream from the Internet, WiFi has to send "ACKs" every few thousand bytes before continuing. And each IP packet is broken into dozens of little WiFi frames (802.11) for transmission when the RF channel/spectrum is unoccupied by some other WiFi near your channel. So the overhead on WiFi is why I speak of needing 1.5 times more WiFi bit rate for the net yield at the IP layer.

So the 105Mbps speed on newer WIFi standards, and with an ideal signal strength, and little airtime competition, might keep up.

But, to what end, given the Internet hosts that avoid hogging?

I also read Comcast customer complaining that they hit the 300GB/mo cap quickly. This cap, say some, was not revealed in the contract which is as ever fuzzy on speeds and caps.

 

[remixedcat]

Windows link rate reported is a theoretical maximum and does not really reflect your actual transfer rate. Also speedtest.net, speakeasy, etc are NOT accurate measure of LAN speeds.

For that I recommend LAN speed test instead, as it will show the true transfer rate.

Also using tools like wifi analytics for windows and android and picking a network that's not congested will give you better performance.

 

[killeraxemannic]

Per the 'net, Comcast extreme 105 is supposed to be 105Mbps (down, what is the upstream?).

I may be judged in the minority, but I think that consumer/residential users are not getting value when they pay for over 50Mbps down. Unless there's a real deal.


I'd think it would be rare to find Internet hosts that will sustain 100+ Mbps to one user. Occasional bursts, yes. But the average would be a lot lower, due to the Internet and the bandwidth that web servers will allow per user/session.

Re WiFi and your 105Mbps Internet. Perhaps an ideal Speedtest.net host will let you get 105Mbps sustained for a brief test, so you can confirm, using a new/fast PC with a wired LAN connection.

Then on to WiFi. The WiFi connection speed needs to be about 150% of your goal net throughput, e.g., 1.5 times 105Mbps if that's what you really get when using wired connections. In addition to the WiFi connection speed at 1.5 times, you need little or no competition for a clear channel air time among neighbors' WiFi that is within 3 channels of yours (in 2.4GHz), AND neighbors that really use a lot of air time. (Number of SSIDs detected isn't important).

Remember that WiFi is "half-duplex". Ethernet is (usually) full duplex. This means that when pulling a big stream from the Internet, WiFi has to send "ACKs" every few thousand bytes before continuing. And each IP packet is broken into dozens of little WiFi frames (802.11) for transmission when the RF channel/spectrum is unoccupied by some other WiFi near your channel. So the overhead on WiFi is why I speak of needing 1.5 times more WiFi bit rate for the net yield at the IP layer.

So the 105Mbps speed on newer WIFi standards, and with an ideal signal strength, and little airtime competition, might keep up.

But, to what end, given the Internet hosts that avoid hogging?

I also read Comcast customer complaining that they hit the 300GB/mo cap quickly. This cap, say some, was not revealed in the contract which is as ever fuzzy on speeds and caps.

The extreme 105 is supposed to be 105 down 20 up. On a wired connection I consistently get 120 down and 25 up. I actually got the service for the upload speed because I do a lot of steaming and uploading to youtube and it's the fastest available in my area for upload speeds.

I definitely get a sustained 100 mb/s+ download speeds though. I have downloaded quite a few games on Steam and they will download at around 14 MB/s throughout the whole download of 20+GB.

Windows link rate reported is a theoretical maximum and does not really reflect your actual transfer rate. Also speedtest.net, speakeasy, etc are NOT accurate measure of LAN speeds.

For that I recommend LAN speed test instead, as it will show the true transfer rate.

Also using tools like wifi analytics for windows and android and picking a network that's not congested will give you better performance.

I am on the correct channels per InSSIDer. Also I am the only network on the 5ghz band here. Also I never said I was using speedtest.net or speakeasy. I have been using http://testmy.net/ which claims they are more accurate than the flash based speed tests.

 

[stevech]

Windows link rate reported is a theoretical maximum and does not really reflect your actual transfer rate. Also speedtest.net, speakeasy, etc are NOT accurate measure of LAN speeds.

For that I recommend LAN speed test instead, as it will show the true transfer rate.

Also using tools like wifi analytics for windows and android and picking a network that's not congested will give you better performance.

But OP was concerned about ISP/WAN speeds, not just LAN speed.

 

[overdrive31]

Look at the Router charts of the AC based routers here at SNB. Mainly their minimum downlink speeds in both bands. Now compare those measurements to your router class. On par with what you're seeing(N) and expecting(AC) heh. Time to upgrade to AC!

Just to be clear, the results are from LAN to LAN testing not WAN to LAN like you're testing with. The tests also reflect 3 stream performance, while almost all AC client devices are still just 2 stream. At the very least, your range should improve considerably as well as some increase in speed on 5GHz(~90Mbps). If you upgrade, use broadcom client devices for best results.

 

[azazel1024]

Not all clients are created equal. A good N300/300Mbps client might be able to push 150-180Mbps in an uncluttered wifi environment, but a crappy one might be lucky to get 80Mbps with a tail wind.

 

[stevech]

Not all clients are created equal. A good N300/300Mbps client might be able to push 150-180Mbps in an uncluttered wifi environment, but a crappy one might be lucky to get 80Mbps with a tail wind.

true

AND... don't confuse net yield of an IP session on the Internet with the raw WiFi bit rate. The raw WiFi bit rate has to be much higher than the goal net yield, because of high overhead in WiFi protocols, and competition for air time.

 

[killeraxemannic]

So what I am gathering is that the speeds I am seeing are normal for wireless N?

 

[azazel1024]

I wouldn't say they are normal for wireless n, but they might not be abnormal for the client and router you are using dependent upon environmental variables.

http://www.anandtech.com/show/7127/the-joys-of-80211ac-wifi/2

Just look at that link and the speeds. Ignore the 802.11ac results if you want and just look at the 11n clients in there. All but the MSI are at least 300Mbps clients. Notice that there are poop ton running at pretty slow speeds?

I wouldn't really consider getting ~60Mbps out of 300Mbps before overhead a good situation, unless of course you have competing networks or have a lot of interferance (or low signal strength), but thems the breaks with different clients.

Some are just crap and others are pretty good.

An Atheros based USB 2.4GHz wifi adapter I have, supposedly 150Mbps capable, hits about 30Mbps with a tail wind stuck in my laptop in the same room as my AP running in 40MHz mode. My Asus tablet with a 150Mbps Broadcom wifi adapter in it hits 70-80Mbps.

Two obstentially equal clients used in the same location and radically different speeds.

Of course about the only thing you can do is either replace the access point/router if you know/suspect it to be bad, play with driver settings to see if something is misconfigured, or if it is possible, swap out the wifi adapter/card for a better one. Because...it might not just be the client. In the link above, look at the 7260 in 2.4GHz...now the author might well have a lot of competing networks or some kind of 2.4GHz interferance going on...but same room, with a Netgear 3500l as the access point, on the EXACT same wifi card, I get 170-180Mbps...nearly double the speeds that the author is seeing.

To get more complicated, drivers and OS play a big role to. Previous to the 7260 in my laptop, I had an Intel 2230. Under Windows 7sp1, I was getting around 105-110Mbps, sometimes 120Mbps on a really good day. I loaded Windows 8.1 on to the laptop and it jumped to 160-170Mbps!

No other changes, I had been running the latest Windows 7 drivers, no configuration differences in the driver properties, nothing. A change in OS lead to a speed increase of around 50%.

At least in my experience, Realtek, on the Wifi side of things, tends to be crap. As does Qualcomm/Atheros. Broadcom often produces pretty good networking gear, but I've run in to some really spectacular driver issues with them (more than ANYONE else). Intel pretty much makes solid stuff, but even they aren't bug free 100% of the time. The hardwired networking world is a little difference, I tend to see pretty good Realtek gigabit adapters, okay Broadcom network adapters, crappy Atheros and excellent Intel network adapters.

I am hoping Qualcomm having bought up Atheros will change that ship around, but at least so far they don't seem to have.

 

[stevech]

In the post above, and others, we should state if a speed number (megabits/sec = Mbps, little b)... is the WiFi connection speed or the measured IP layer rate, and if the latter, is it LAN/LAN or LAN/WAN and how measured.

40MHz mode in 2.4GHz is not recommended by many users, unless in a rural setting.

 

[sfx2000]

I have done a few speed tests lately after upgrading my internet connection to the Comcast extreme 105 on my wireless and I have found that I am not getting my full download speed over wireless on my RT-N65R.

What is the brand and model of your cable modem?

Getting to those speeds requires a DOCSIS 3.0 modem, and Support may have to push an updated configuration file to the modem.

 

[TonyH]

Hi,
My speed deal is 50/3 by Shaw cable. When I speed test against known server, I always
get ~50/3 but when downloading or streaming, depending where it is coming from, speed
varies any where like from 7 to full 50 any time of the day. Play with QOS set up for an improvement.

 

[overdrive31]

Not all clients are created equal. A good N300/300Mbps client might be able to push 150-180Mbps in an uncluttered wifi environment, but a crappy one might be lucky to get 80Mbps with a tail wind.

It's important to match up router and client chipsets esp. since Broadcom AC also fully supports beamforming(Realtek and MediaTek clients support it as well). Reason enough for my recommendation.

I have to disable beamforming on the router to get decent performance with the 7260(Intel clients do not support beamforming). Another one of those, put the G clients on a separate AP to avoid impacting the N clients scenario. The OP wants wireless speed to perform as fast as their ISP provides. The only 3 stream AC client(desktop client) available uses what chipset? Broadcom.

That article you posted, did it use any AC Broadcom cards? Only the highest performing of the bunch, which was the WD bridge. Mostly the advantage being 3 stream vs. 2 stream, but I'm sure beamforming kept the speeds high as the distance became greater. It's rather puzzling that the author claims they never connected with more than 2 streams using any AC clients including the bridge. Uh, how? Router and bridge are both 3 stream, there shouldn't be any limiting factor. I guess they assumed it was 2 stream because they expected higher speeds despite having something to compare to.

According to Tim's testing with AC, the results for maximum performance are on par with the author's for 3 stream AC. So they assumed only 2 stream with the WD bridge because...they expected WAY(800-900Mbps?) more speed. Maybe with 802.11ac Wave 2 if we're lucky. May need to wait for 802.11ad for those speeds more than likely.

I know beamforming isn't listed in the specs for the WD networking equipment, but the hardware does support it. How come the speed of the WD was greater than 1/3 the Intel, shouldn't 3 stream vs 2 stream be close to 1/3 faster? In some cases it was > 2x faster, could that be because Intel and beamforming don't mix? Looks that way now doesn't it.

 

[KGB7]

You need to do a file transfer from PC that is hardwired in to a router, to Laptop over wifi, using a single large file. This is called LAN to LAN speed test, it requires no software of any kind.

Then you start tweaking your router settings and wifi NIC settings. NIC has many settings with in its properties.



I have FIOS 50/25, and after running multiple speed tests on testmy.net, i get 8/3, thats in Mbps. So this clearly tells me that this site is flawed. Yet when i torrented LinuxMint, i maxed out my FIOS. I was getting 56Mbps down. LAN to LAN, i get 60Mbps on 2.4Ghz band @20Mhz.


So if you still want to use a flawed testing method; WAN to LAN. Then at the least use Comcast speed test website.

 

[killeraxemannic]

What is the brand and model of your cable modem?

Getting to those speeds requires a DOCSIS 3.0 modem, and Support may have to push an updated configuration file to the modem.

In the post I stated that I am getting the correct speeds over wired ethernet so this would mean there is not an issue with my cable modem. It is the Motorola SB6141. Hard wired on my desktop I get 120/25 fairly consistently when running speedtests from http://testmy.net/

 

[killeraxemannic]

You need to do a file transfer from PC that is hardwired in to a router, to Laptop over wifi, using a single large file. This is called LAN to LAN speed test, it requires no software of any kind.

Then you start tweaking your router settings and wifi NIC settings. NIC has many settings with in its properties.



I have FIOS 50/25, and after running multiple speed tests on testmy.net, i get 8/3, thats in Mbps. So this clearly tells me that this site is flawed. Yet when i torrented LinuxMint, i maxed out my FIOS. I was getting 56Mbps down. LAN to LAN, i get 60Mbps on 2.4Ghz band @20Mhz.


So if you still want to use a flawed testing method; WAN to LAN. Then at the least use Comcast speed test website.

I don't want to use that program because you have to pay for it. Test my net and Comcast speed test website give me almost identical results most of the time. I will give the large file transfer speeds a test tomorrow and see what kind of results I get.

 

[azazel1024]

It's important to match up router and client chipsets esp. since Broadcom AC also fully supports beamforming(Realtek and MediaTek clients support it as well). Reason enough for my recommendation.

I have to disable beamforming on the router to get decent performance with the 7260(Intel clients do not support beamforming). Another one of those, put the G clients on a separate AP to avoid impacting the N clients scenario. The OP wants wireless speed to perform as fast as their ISP provides. The only 3 stream AC client(desktop client) available uses what chipset? Broadcom.

That article you posted, did it use any AC Broadcom cards? Only the highest performing of the bunch, which was the WD bridge. Mostly the advantage being 3 stream vs. 2 stream, but I'm sure beamforming kept the speeds high as the distance became greater. It's rather puzzling that the author claims they never connected with more than 2 streams using any AC clients including the bridge. Uh, how? Router and bridge are both 3 stream, there shouldn't be any limiting factor. I guess they assumed it was 2 stream because they expected higher speeds despite having something to compare to.

According to Tim's testing with AC, the results for maximum performance are on par with the author's for 3 stream AC. So they assumed only 2 stream with the WD bridge because...they expected WAY(800-900Mbps?) more speed. Maybe with 802.11ac Wave 2 if we're lucky. May need to wait for 802.11ad for those speeds more than likely.

I know beamforming isn't listed in the specs for the WD networking equipment, but the hardware does support it. How come the speed of the WD was greater than 1/3 the Intel, shouldn't 3 stream vs 2 stream be close to 1/3 faster? In some cases it was > 2x faster, could that be because Intel and beamforming don't mix? Looks that way now doesn't it.

I posted the article more to show examples of what a crappy 802.11n client look like (check the Realtek chipset NIC there at abysmal speeds, half or less of most of the other 802.11n clients).

I can't attest to 11ac performance on the 7260 yet, but its bested all 11n clients I've tried with my network setup by a good 10-15Mbps (300Mbps 2.4GHz 40MHz, Netgear 3500L). I've tried 4 300Mbps NICs and the Intel 7260 is the fastest of them.

At any rate, I've seen enough people with the 7260 getting real world >40MB/sec speeds on the 7260 with the AC66u and the C7 Archer and several who pushed over 50MB/sec that I am pretty confident it is a good performing client.

As a note, not sure if it is OS, cluttered 2.4GHz, poor performance on the WD AC router...but I think I noted in my previous post, my 7260 with Windows 8.1 posts up speeds of 170-180Mbps on 2.4GHz 300Mbps...compared to the around 100Mbps that the author was managing.

Based on their provided 105Mbps downstream ISP speeds...a decent 2:2 2.4GHz 40MHz setup should competently be able to handle that easily, at least at modest distances from the router with a good router and a good client.

Again...mine pushes 170-180Mbps, well beyond 105. Heck, even on Win 7sp1 and my older Intel 2230, I pushed 120-130Mbps.

a single stream 11ac client should exceed 105Mbps at resonable distances from the router. Now if you want to max it at longer ranges, you'll need a significantly more powerful router/client combo, or multiple APs.

My 170-180Mbps turns in to 10-15Mbps on the otherside of my house (50ft, a 4ft thick masonry chimney and 4 interior walls away)...which is why I have a router in my basement on the exact opposite end of my house, only configured as a 20MHz so no channel sharing, but the worst performance I get in my house is 50Mbps on my laptop and typical is more like 70+Mbps anywhere in the house. Closer to 100Mbps in my basement by the router or LoS with it, or 170-180Mbps in my familyroom or the next door dinning room.

Fast enough for me, and with a 75/35 (85/32 real world) FIOS connection, I am happy enough on the internet side of things. Heck, for outside, I got a 10/100 port based 300Mbps 2.4GHz router. There I care more about range than about speed so I don't feel that the 100Mbps port limit will effect me, besides I'll probably run it in 20MHz mode to reduce nearby interferance. I am a good distance from my neighbors, but outside the nearby network signal strength is a lot higher.

In my family room it is -86dB for the strongest network nearby, step right out the back door of the family room on to my back deck and it is -77dB. Go to the area of my backyard where I care about getting signal and a seperate network from my rear neighbors is around -60dB...so getting on a channel seperate from theirs (its the only competing network at medium+ strength there, but I think theirs is set to 40MHz) and also at a higher power is crucial. Shouldn't be an issue with LoS to the antennas and only 80ft or so. 15-20Mbps of a nice solid connection is a lot more important than craming down 150+Mbps for file transfers. Internet/email/basic video streaming is what I care more about out there.

I do want 11ac, and soon, in part because on of my APs is dying and I do often want to transfer multiple gigabytes from my laptop to the server or vice versa. So 300-400Mbps would be real swell...

 

[overdrive31]

The article, in terms of 2.4GHz N performance, was biased towards the Intel, there wasn't any 2.4GHz "AC era" clients to compare to(the usb client was single stream, so not viable). Your idea of it being the best for 2.4GHz N, based on what comparison? Just other N only clients? That is a fair comparison? Don't knock it til you try it. There is definitely a solid performance benefit in 5GHz AC with Broadcom, and beamforming works on both bands(some routers do not support it on 2.4GHz). Now how would it benefit to ignore such based on older tech vs the latest Intel hardware that lacks features that improve the current tech with 2.4GHz N?

Tim, how about a review on AC clients...

Place your bets..which client will be the top performer with beamforming enabled?

You thought the WD router suffered from poor performance? It's beamforming...with my 7260 and beamforming enabled, the bandwidth gets a 50% cut. The dropouts people report with the 7260, likely a result of beamforming being enabled.

To me, 2.4GHz is only for convenience. There are no guarantees from one moment to the next, with so many factors that can disrupt it.

 

[azazel1024]

My point isn't that the 7260 is the bestest of the best on 2.4GHz, though in my testing for the variety of clients i've tried, it is for 2:2.

My point is, that there are a lot really BAD 802.11n clients, probably a lot of really bad 802.11ac clients out there. The link and my few brief examples are simply showing that, yeah, there is a HUGE different in the clients you are running as well.

So, move away from fixating on the 7260 and your experience on 11ac versus mine on 11n and 2.4GHz. My SOLE point is, there are good clients and bad clients and there is a huge performance gulf between them.

Just because something is 300Mbps or whatever, doesn't mean it is going to be remotely as fast as another device that is 300Mbps.