Slashdot Mirror
802.11ac 'Gigabit Wi-Fi' Starts To Show Potential, Limits
alphadogg writes "Vendor tests and very early 802.11ac customers provide a reality check on 'gigabit Wi-Fi' but also confirm much of its promise. Vendors have been testing their 11ac products for months, yielding data that show how 11ac performs and what variables can affect performance. Some of the tests are under ideal laboratory-style conditions; others involve actual or simulated production networks. Among the results: consistent 400M to 800Mbps throughput for 11ac clients in best-case situations, higher throughput as range increases compared to 11n, more clients serviced by each access point, and a boost in performance for existing 11n clients."
The anonymous coward version of wi-fi.
"Among (sic) the results: consistent 400M (sic) to 800Mbps throughput for 11ac clients in best-case situations"
Best case being: the only device on the network; inside a Faraday cage; on the dark side of the Moon; 3 centimetres away from the antenna.
BTW: Google.... fuck your dictionary. It IS centimetres.
With 20+ APs contending for their own slice of half or third of 5 ghz band. 802.11ac took the best feature of the 5 Ghz band, plenty of non overlapping channels, then turned into back into the quagmire of the 2.4 Ghz band by allowing 80 and 160 Mhz spectrum usage. Of course, the router manufacturers are going to enable 160 Mhz by default even when everyone in the neighborhood is on a 25 Mbps cable modem connection.
Irrelevant. Your experience with N was likely a bad router or user error. Whatever the reason, your results are extremely atypical. It would be irresponsible to compare any sort of standard to a weird oddball experience.
Seriously, if you're experiencing ANY "random disconnects", it's time to update the firmware or flat out get a new router. That should be your first clue that something is wrong with your setup.
Actually it isn't. By far! 1. On a gigabit wired network you get 1Gbit of transfer speed. There is a very small percentage lost to coding but you get well over 100MB/s (up to about 120MB/s) trough a Gbit connection. If you get slower speeds and don't know why, than start searching for the bottleneck! 2. The 400Mbit to 800MBit in a WLAN is the "wire speed". I've never seen transfer rates that are more than 70% of this. So, I expect to get maybe 56MB/s (which is already quite good) out of "GBit WLAN" while I get 120MB/s out of an Ethernet connection almost all the time. Still impressive how they even reach such speeds! That's engineering at it's best!
Yes. You're absolutely right. Because the only use for a WLAN is using the internet...
Linux driver support for most of the 802.11ac devices are still iffy which doesn't help.
2.4GHz is far too crowded. Switch to 5GHz and you should see an improvement, particularly if you're in the same room as the AP.
Hardware manufacturers I'm pointing my my finger at you. The most powerful features of 802.11n are largely unimplemented. Laptop/tablet/phone Support for 3 spatial streams is about as rare and rocking horse shit. Support for even 5 ghz is spotty at best and its hard to find out if whatever piece of hardware you want to consider buying even supports it. Heck even 2 spatial streams at 2 ghz is something your lucky to get unless you spend more than $699 on a laptop. The lowest common denominator for 802.11n and what the "wireless n" wifi support really means for half the devices on the market is a single spatial stream 802.11n at 2 ghz, which is 65 Mbps max. I can buy a mid range smartphone with 4g support and the wifi is still single spatial stream at 2 ghz. Hardware manufacturers have no incentive to put better implementations of 802.11n in their because most customers aren't savvy enough to tell the difference and demand better from device manufacturers. 802.11n is on old specification. There's no excuse why 2 spatial streams can't be the minimum. The silicon to do this is cheap and has been refined for many years.
802.11ac will probably suffer the same fate. The minimum implementation to get the "wireless ac" sticker on the box is going to be what half to three quarters of the devices on the market will support, even 10 years from now.
Reaching far back to my Cisco knowledge from 2003-ish, that's because 802.11 requires acknowledging every single packet, whereas wired Ethernet allows a larger window, so several packets get sent before an acknowledgement. I don't know if that's still the case (perhaps a modern network engineer will confirm, please), but that could be the reason for seeing just about double the transfer speed through a wire. On wireless, you're using almost twice as many packets to receive the same data.
You do not have a moral or legal right to do absolutely anything you want.
Have they implemented the full 256QAM 5/6 rate yet with full 80+80MHZ bonding yet? (160 MHZ of channel bandwidth) using 8 transmit antennas and 8 receive antennas on both AP and wireless clients?
I expect early APs and early chipsets will not yet fully implement all the advantageous features 802.1AC has to offer
They'll have made compromises to save money.
I get 114MB/s over SMB with non-jumbo frames. That is 912mb/s of effective bandwidth. I get about 996mb/s of raw bandwidth when doing an iperf test. 800mb/s on a 1gb wired network is HORRIBLE.
To get those higher speeds outside the lab, you'll need some wifi spray
> The noise floor across the whole of the RF spectrum is rising by an average of 1db a year.
You are correct, but not for the reasons you discussed. If the millions of Transmitters were clean and well designed, they would not cause RF interference to other users (except where they were sharing common frequencies).
The problem is that much of the electronics junk generate spurious harmonics. Plasma TV's, PC's, BPL, etc. all put out a horrendous range of broadband rubbish.
This is compounded by many manufacturers and importers ignoring the existing EMC standards, as well as the corrupt regulatory bodies (FCC etc) turning a blind eye to the cheap plastic junk being imported.
Just one specific example. Once upon a time, manufactures used linear-mode power supplies with large transformers. In an effort to reduce costs, they have universally changed to using switch-mode supplies. These supplies are certainly cheaper, but they almost always generate much higher levels of radio interference.
There's a trade-off: Being able to buy cheap electronics means that there's a good chance you will be unable to enjoy it due to the resulting interference levels.
Switch to 5GHz and you should see an improvement
Combined with further reduction in range. With an ASUS N56U, in the middle of nowhere with no interference, 2.4GHz becomes unreliable at around 700ft. 5GHz drops out somewhere around 450ft.
AFAIK, the WiFi standard allows you to do accumulated bulk transfers similar TCP Nagle's algorithm before being ack'd. I believe this dates back to 802.11g because they (the WiFi alliance) realized that the overhead for WiFi is more expensive in certain cases.
I can echo this, almost exactly.
That's not the Nagle algorithm. Nagle is about delaying before sending packaets, to prevent lots of small packets instead of one big one.
You're probably being limited by the R/W speed of your hard drive and O/S.
I consistently get 100MB/s over my network between two machines that are capable of reading and writing at least that fast to their storage systems, and this is with cheap Realtek gigabit equipment.
I would not call those results atypical. Signal strength will drop, though in many cases 5GHz will be cleaner in the first place so it makes up for it in quality.
But the client behavior when presented with multiple APs on both 2.5GHz and 5GHz, and when presented with multiple APs some of which are N-capable and some just a/b/g is generally abysmal. We have lots of clients that students bring from their simple single-AP 2.5GHz home networks and just cannot cut it in a WPA-enterprise environment with lots of infrastrcuture APs around. They jump around between APs constantly, often choose APs based on mysterious metrics which are probably the worst choice of available APs, and very often the worst of them manage to trigger themselves to re-ask for credentials despite being told to remember them -- I don't know how that got into their codebase, but we've got several users that get constant cred popups. To top it off the UI on the devices has been dumbed down to the point where there is no user-level control to the degree of selecting preferred BSSIDs or tweaking any parameters whatsoever. Most cannot even tell the user what BSSID they are currently connected to.
I'm glad 11ac is going to force device manufacturers to start putting 5G antennas in again, but anyone running an enterprise WAN would be well advised to increase their AP density to 5G full coverage, drastically reduce the tx power on their 2.5GHz radios so they look quieter than the 5G radios, and not rely on the devices falling back to b/g/n-2.5 reliably. WiFi driver software is apparently written by conpanies that have invested zero into recreating real-world "BYOD" scenarios for QA purposes.
Someone had to do it.
I bet you have a NetGear WNDR3800 with Stock firmware.
Nothing but disconnects and reboots. Especially with N.
Netgear's answer: Discontinue the product.
Obama's legacy: (N)othing (S)ecure (A)nywhere and (T)error (S)imulation (A)dministration
If you are trying to reach an access point 150m away, you are not in a densely populated area. 2.4GHz will work fine for you without interference. The lower range of 5GHz is an advantage, it helps ensure that the band has low interference.
(Although I am sure it will get crowded soon. 60GHz, here we come...)
Finally! A year of moderation! Ready for 2019?
I'm getting 7-9 MB/sec with my 802.11ac adapter, whether on the 2.4GHz band or the 5GHz band. So at least the theoretical speed of wireless G is finally real.
It is much faster than the 2.5MB/sec I was getting on the so-called "a/b/g/n" adapter.
For comparison, on actual gigabit ethernet, I get about 88-100MB/sec.
1.) Wired ethernet will give you more than that; as another responder said you're likely to be limited by the write speed of your storage devices. If you really want to test it, create a RAM disk on both machines, and transfer a file from one RAM disk to another - you'll see yourself saturate the line pretty quickly.
2.) Wireless is inherently half-duplex, because you can't transmit and receive at the same time on the same frequency. Dual-band technology was supposed to help out with that, but it only works if everything on the LAN supports it. Else, one band will be more crowded than the other, and messes will still be made. Additionally, the more devices you've got running on your wireless network, the greater a collision domain you've got.
3.) Pursuant to 1 and 2, when you've got large numbers of devices, it's where wired really shines. no one gets a good signal with a hundred other access points and endpoints on a LAN, but if you've got everyone wired to a half decent switch, everyone can communicate much more efficiently. Additionally, you may be seeing poor speeds if your gigabit switch is low quality and has a poor amount of backplane bandwidth.
In a warehouse, actually. The only "interference" is from trying to get a signal through concrete and steel spaceframe. We brought an access point to connect to the equipment we were installing wirelessly, until the customer could get around to installing their own wireless infrastructure. It was a dual-band access point, and the 2.4GHz signal was significantly higher performance at range, for obvious reasons. Strangely enough, when the customer did install their infrastructure, it was 802.11a.
No number for worst case latency - Something needed so VoIP actually works.
I suppose it is not very good or they would have mentioned it.
yes but its only half duplex.. o wait
My summery of the technology would be.
Got a bleeding edge MB last month, included ac wifi. Looked into buying an ac router. Not that many actually available, and the few that were cost 150-250$.
I will stick with my g router which probably is worth about 10$.
I will wait until they become a bit more affordable.