So if you haven't heard of the new super-duper fast wifi standard, there's 802.11ac. Supposedly faster than 802.11g and 802.11n. Promises up to 450mbps on the 2.4Ghz band, or 1300mbps on the 5Ghz band.
I finally got my hands on a USB wireless device and a router.
The verdict:
7 MB/sec on 2.4Ghz
7-9 MB/sec on 5Ghz (although I was able to get 13.5 MB/sec on a different computer)
So much for "450mbps" on 2.4Ghz or "1300mbps" on 5Ghz.
Still much faster than the 2.5 MB/sec I was getting before on the so-called "a/b/g/n" card. At least the "54mbps" promised by old standards has finally been delivered.
If you want throughput/speed, use Ethernet. If you can't run CAT5e/6 around your dwelling, then consider using PowerLine adapters. Don't complain about having to plug a cable into a device -- I don't want to hear it. :-)
If you're testing throughput using CIFS/SMB: don't. The protocol is slow as molasses at numerous levels. Use FTP.
But in general all 802.11 protocols are shit. Anything wireless is generally shit, for the record -- that applies universally to speed, reliability, all of the above. The entire design is "spew crap and hope it gets there". No thanks, rather not.
FTP has to repeat
slow-start for each copied file unless you're putting them in a tarball, zipfile, disk image, or other archive first. HTTP at least reuses connections. Keep this in mind when you run speed tests. Also keep it in mind when you try to upload a big web application like MediaWiki to a shared web host; you want one with shell access because uploading the application as a tarball and extracting it on the server is a lot faster than opening a connection for each file.
Does passive mode FTP still open additional connections? I thought NAT was so prevalent that nobody uses active mode FTP anymore. Edit: yes, passive mode does open additional connections...
Tepples is worrying about something that isn't relevant to the discussion. Do not let his words sway what I am telling you; please re-focus.
For FTP throughput testing, you use a single large file for tests -- say, a 600MByte ISO image. Socket creation time has virtually no impact on this kind of test. You need to test across your LAN -- do not ever use an Internet-bound host for wifi throughput tests. A proper test: wifi client (e.g. laptop) connecting via FTP to an Ethernet-connected server (e.g. desktop box).
Unrelated: both passive and active FTP mode are still used today. Failure to support both results in problems -- and I cannot tell you how many times I have had to explain how the FTP protocol actually works (so many times on DSLR and linksysinfo.org!) to people setting up/configuring firewalls, because most people "think" they understand it but actually don't.
I don't use wifi because I don't want to run cables - my main desktop machine is tethered, as is my PS2. I do like Wifi for handheld devices and portability - my GF brings her laptop over here sometimes and it's nice to be able to let her use the internet without having to string Ethernet across the floor. Plus, I know some people here may not like Apple devices, but I'm quite fond of my jailbroken iPod touch and heavily use it with my wifi.
I got WiFi because of my GF's android tablet thing, smartphone and WiiU and sometimes my laptop and her laptop. I cannot run any cables in this apartment (through walls anyway) and it is hard to get any connection to the bedroom besides WiFi.
Since we're already off topic, I use wifi for a distance of ~12ft on my main desktop because there are doors on either wall between my modem and PC. I love wireless so much even my internet access is wireless from our provider. To be fair/honest I suppose it's more due to the fact I don't have much for internet provider options living in the sticks. And I guess I'm lazy and easy to please, why route a cable if I don't have to. Especially if the wifi isn't the limiting factor on my internet.
Like I said -- look into PowerLine adapters. No "wall tunnelling" or "routing of cables" needed. It works, and the throughput is (on most products -- not all, but most) as advertised. It's also AES encrypted (and given how it works the encryption is important) -- but it's all 100% transparent to the system/OS/machine (as it should be).
For handheld devices that absolutely warrant wifi, yep, you have no real choice in the matter. The protocol and technology (wireless) still sucks and is extremely error-prone, however. My approach is always: if you have the choice of using Ethernet, take it, else use PowerLine adapters; for handhelds you gotta use wifi, just nature of the beast.
I've used adapters through the wall. There are conditions where they work well, and conditions where they don't, depending on what the power situation in your house is. They're worth a try, at least. I've had situations where their bandwidth is severely limited, or drops out for sustained periods due to other electrical usage in the house (e.g. A/C, vacuum cleaner, etc.), and I've had other situations where it worked reliably for a long time.
I almost exclusively use wifi, because I very seldom find it lacking (there are 6 computers in this house using wifi daily, none wired). There are occasional signal drops, of course, but not frequently enough to be a problem. Maybe once a week one happens in the middle of a download and I need to resume/restart. Not a big deal, but it depends on your usage. Obviously there are lots of situations where wifi reliability is very poor, but it can also be very good. If you need a constant uninterrupted signal, though, there's no substitute for being plugged in.
Large transfers (e.g. >10GB) I generally just stick on a portable hard drive, but with the speeds dwedit's talking about for 802.11ac maybe I wouldn't even bother.