MacBook Air can't max out WiFi Speeds

[xaitmi]

Hi, I have a mid 2013 MacBook Air running the latest OSX (Yosemite 10.10.2).

I have an ASUS RT-N66U and a Netgear R7000 NightHawk.

My internet speed is 250 down and 20 up.

When ever I do speed tests on my Macbook Air, I can never max out the speeds, they max out around 110 - 130 down and 20 up.

At first I thought it was because of my RT-N66U, so I bought the Netgear R7000 yesterday.

We tried investigating it in this thread, a lot of details and diagnostics are posted on page 2 and 3.
Here is a link to page 2, scroll a bit down till you see me. http://www.dslreports.com/forum/r29909628-Anyone-using-Rogers-250-with-an-Asus-RT-N66U-~start=30
My username is imtiax there.

Here is a copy of some of the details I posted there.

This is while I'm sitting right next to the Netgear R7000..

http://i.imgur.com/ih3IXlI.png
http://i.imgur.com/lbruX8y.png

Does anyone have any idea why my MacBook is having issues maxing out the connection?

 

[stevech]

are you expecting the WiFi connection rate to match the net yield at the IP layer?
Typically it's 60% of the WiFi rate due to WiFi overhead.

 

[xaitmi]

are you expecting the WiFi connection rate to match the net yield at the IP layer?
Typically it's 60% of the WiFi rate due to WiFi overhead.

I was hoping speeds of at least 200 down over WiFi, whereas I only get ~ 130 down.

My friend brought over his Windows laptop to my house last night, and we ran a speed test and he was able to get around 230 down and 20 up.

 

[L&LD]

Have you tried disabling the power savings of the air or the ac wifi?

That model is made for maximum battery run time, not best performance (WiFi or otherwise).

 

[xaitmi]

Have you tried disabling the power savings of the air or the ac wifi?

That model is made for maximum battery run time, not best performance (WiFi or otherwise).

Tried it, but no difference.

 

[azazel1024]

Which band are you connecting on? I'd suspect the 2.4GHz band, which is why. Last I checked, the Air uses a 2:2 adapter in it and Apple is 20MHz channel widths in 2.4GHz ONLY.

That means you are looking at a max link speed of 135/144Mbps and a max usable (with great router and great client, close to it, no interference) of maybe around 110-115. Are you sure you are getting stable, average speeds of 130Mbps?

Anyway, that would be my suspicion.

 

[xaitmi]

Which band are you connecting on? I'd suspect the 2.4GHz band, which is why. Last I checked, the Air uses a 2:2 adapter in it and Apple is 20MHz channel widths in 2.4GHz ONLY.

That means you are looking at a max link speed of 135/144Mbps and a max usable (with great router and great client, close to it, no interference) of maybe around 110-115. Are you sure you are getting stable, average speeds of 130Mbps?

Anyway, that would be my suspicion.

I am connecting to the 5Ghz Wireless N (Currently connected to my ASUS RT N66U)

http://i.imgur.com/nt4SNC4.png
http://i.imgur.com/n6ybfjl.png

When I do speed tests I usually get between 110-120 down, I've seen it go upto 130 a few times.

What I find strange is that even though the screen shots above say I am connected to a 40Mhz network, which I am, and my MCS index is 15.

40MHZ with MCS of 15 should mean Data Rate of around 270-300Mbit/s whereas my speedtests always look like it's using the 20Mhz which 130-144.4Mbit/s

Here is an MCS chart http://communityforums.rogers.com/t...user-id/829158/image-id/1128i1B0D5B8474F8D75B

 

[thiggins]

What link rate did your friend's Windows notebook report?

The mid-2013 Air has a 2x2 AC radio. It has a max 40 MHz N link rate of 300 Mbps and max
80 MHz AC link rate of 867 Mbps.

Using this chart http://www.smallnetbuilder.com/tools/charts/router/bar/113-5-ghz-dn-c with benchmark type selector set to Maximum.

The best I ever measured for a 3x3 N router 5 GHz downlink, 40 MHz mode was 192 Mbps (ASUS RT-N66) with a 3x3 adapter. 2x2 adapter would have lower throughput.

Looking at AC1200 results (2x2), we see results more like you desire.

Bottom line is that you should be able to get what you want from the R7000, but not the N66U.

 

[azazel1024]

In your testing. Real world is likely to produce a lot higher throughput supposing you aren't running in to interference issues.

Heck, my Intel 7260ac with my TP-Link WDR3600 manages to produce 204Mbps on 5GHz 40MHz same room with the router and 200Mbps on 2.4GHz 40MHz and that is just 2:2. I've seen Intel 3:3 11n adapters produce >280Mbps with good N750/900 routers.

 

[thiggins]

In your testing. Real world is likely to produce a lot higher throughput supposing you aren't running in to interference issues.

Heck, my Intel 7260ac with my TP-Link WDR3600 manages to produce 204Mbps on 5GHz 40MHz same room with the router and 200Mbps on 2.4GHz 40MHz and that is just 2:2. I've seen Intel 3:3 11n adapters produce >280Mbps with good N750/900 routers.

How do you measure throughput in your testing?
Expecting 40 MHz performance in 2.4 GHz is optimistic. The band is just too crowded.

 

[azazel1024]

How do you measure throughput in your testing?
Expecting 40 MHz performance in 2.4 GHz is optimistic. The band is just too crowded.

It isn't too crowded for a lot of people. Yes pretty much any urban or apartment deweller is going to run in to issues. Most people in suburban neighbor hoods aren't going to have significant issues and anyone rural is completely in the clear.

That probably makes it half the US who have no significant 2.4GHz interference issues at their house.

My numbers on the WDR3600 are 200Mbps on 2.4GHz 20MHz and 204Mbps on 5GHz 40MHz.

You mentioned that the best downlink ever was on the N66 5GHz 40MHz of 192Mbps. My WDR3600 clearly beats that at a 300Mbps link rate in 2.4GHz and 5GHz and I've seen other routers and Intel 6300 adpaters with 450Mbps link rate on 2.4 and 5GHz push close to 300Mbps downlink in actual performance.

My Archer C8 manages 224Mbps in 2.4GHz downlink performance connecting to my Intel 7260ac for 40MHz channel width and 116Mbps in 20MHz channel width for 2.4GHz.

 

[thiggins]

I think your estimates of no interference in 2.4 GHz are optimistic, at least on a per capita basis.

I never saw anywhere near that performance when I did open air testing.

What is your test method?

 

[azazel1024]

Transfer a file through robocopy from my server to my laptop, then from my laptop to the server, rinse and repeat with 5-6 files ranging in size from 1GB to 6GB.

Average the values. I've got 5 standard testing locations in my house. Same room, line of sight about 8ft from the router, the next room through one oblique wall about 8ft from the router, across the house through 3 walls about 40ft from the router, the floor below about about 8ft horizontal from the router and then across the house, a floor down through 3 walls and about 50ft from the router.

The 200Mbps on 2.4GHz 40MHz is the average transfer performance on downlink and the 204Mbps is the average performance on 5GHz 40MHz. Uplink performance isn't nearly so good. I'd have to look at my testing notes, but I think it is around 160-170Mbps on 2.4GHz and 5GHz 40MHz. The C8 Archer is also better on the uplink side of things, as well as downlink (ignoring 5GHz, because I haven't tested setting the 7260ac in to 5GHz 11n mode) at around 224Mbps downlink and 186Mbps uplink for 2.4GHz 40MHz to the Intel 7260ac.

The robocopy measurements come pretty close to stacking up with the performance I see with what windows reports through regular SMB transfers utilizing file explorer (obviously there you can't see the true average, just the current reported transfer rate averaged over the past second or so) as well as times transfers.

One behavior I have noticed, especially with 11ac 5GHz is that a lot of times the transfers start a little slower and then ramp up. Not sure if that is some TCP windowing going on or what. With the Archer C8 I see a lot of file transfer behavior where it'll start around 360Mbps or so (close to the router) and then ramp up over 4-8 second to around 410-460Mbps. Much less consistent speeds in the 5GHz 11ac realm then I see in 5GHz 11n or 2.4GHz area where most of the time the transfers start off at full speed or roughly full speed and might vary a bit during testing so long as the client isn't being moved around or nothing else is accessing the router.

 

[L&LD]

If you have more than 6GB of ram in your server and your computers, it may be caching some or all of the files? Yes, the transfers over the network are what we are interested in, but maybe some are still cached.

What are the speeds for the first transfer only? After rebooting both the server and the computer(s) used?

 

[thiggins]

Thanks for the testing details azazel. Windows filecopy does all sorts of optimizations including multi-threading and adjusting for network bandwidth. That could be part of the reason why my single-threaded IxChariot transfers never reach similar speeds.

 

[azazel1024]

If you have more than 6GB of ram in your server and your computers, it may be caching some or all of the files? Yes, the transfers over the network are what we are interested in, but maybe some are still cached.

What are the speeds for the first transfer only? After rebooting both the server and the computer(s) used?

Within 5% or less of the follow on tests. With lots and lots of wired and wireless network testing, the only caching behavior I have seen is RAM caching instead of reading/writing to storage (example, with my 2Gbps SMB Multichannel testing, as utilized as my HDD is, it is limited to about 160MB/sec of read/write performance between machines, but with cached files still resident in RAM, I can get 220-235MB/sec transfers until it has to hit the HDD, as it'll cache in RAM on the recieving end some too). It is still transmitting everything over the wire/wireless.

I have also tested with reboots right before testing and no difference (so I don't bother anymore).

The only difference is typically in the size of the files. Average performance on a 1-2GB transfer on 11ac with the couple of 11ac routers I have tested tends to be slightly lower because of that ramp up performance I typically notice (not always, but maybe 3 in 4 file transfers will start off in the mid 40MB/sec range and then jump to around 50-55MB/sec within a handful of seconds).

It is part of why I never test with files smaller than 1GB in size (because that few second ramp in performance can often skew results). I also use different files on each side (5-6 files on 1-6GB on the server side and a different set of 5-6 files on the laptop side). If I notice a particularly odd result I will tend to throw it out if 3 of the exact same transfer do not repeat the results. Doesn't happen that often, but I do occasionally run in to something where, say, one transfer chugs along at 40-42MB/sec for the entire thing. I re-test and the next 3 transfers all are zipping at 50-55MB/sec.

Windows is pretty good about squeezing optimizations out of network transfers I've noticed. Especially with the later kernels and network stacks. "back in the day" I noticed a fairly impressive gain in wireless performance moving from windows 7 to windows 8.1 on the same network adapter (Intel 2230 at the time) using the exact same Intel driver branch and even the exact same driver build (though obviously the 8.1 drivers and not 7 drivers). At the time on my Netgear 3500L I went from about 140Mbps downlink and 130Mbps uplink 2.4GHz 40MHz performance to 165Mbps downlink and 160Mbps uplink.