Re: Belkin Adapters

On Wed, 02 Sep 2009 09:19:14 -0700, Jeff Liebermann <jeffl@xxxxxxxxxx>
wrote:

On Wed, 02 Sep 2009 08:13:59 -0400, Christopher A. Lee
<calee@xxxxxxxxxxxxx> wrote:

USB dongles aren't so good at higher speeds because of the limitations
of USB itself.

USB 1.1 is limited to 12Mbits/sec, which will certainly be a problem
for 802.11g speeds. However, USB 2.0 is limited to 480MBits/sec,
which is considerably faster than the 54Mbits/sec needed for the
highest 802.11g speed. 802.11n may be a problem for USB, but I've
found that speeds over about 120Mbits/sec very difficult to achieve
and then only at fairly close ranges. Therefore, USB 2.0 should not
be the limiting factor.

USB is simplex (ie one direction at a time) and bit serial. So it
takes longer to send data that is parallel in the computer out to the
adapter and make it parallel again in the adapter, as well as having
to keep transitioning between send and receive down the USB interface.

The 480 mbps also has to handle any handshaking protocol between the
computer and the adapter.

The express card splats the data out to the adapter in parallel. It's
like the difference between counting up to 32 and just saying 32.

I found speed slowing drastically with N as the load
increased.

I assume you mean "throughput" and "system load". It shouldn't slow

Yes.

This surprised me until I discovered how USB uses up CPU.

down. If it does, your computah is hurting for horsepower. Please
note that if it can do 100Mbit/sec on the ethernet port, without a
slow down, then it should also be able to do the same on the USB 2.0
ports. If not, something else is amis, such as a badly written
wireless driver.

I use gigabit Ethernet on the desktop, and the only time I have used
wired Ethernet on the laptop was to connect directly to the cable
modem.

Wouldn't surprise me though. Most of this stuff is badly written crap
because it costs too much to write decent software.

Another poster has pointed out how much CPU is used with
USB, which could be part of the problem but N also pushes the transfer
rate of the USB interface to the limit.

The limit of USB is about 3 times the best you can do with 802.11n.
However, that assume the machine is capable of digesting the data. If
the destination for the data is /dev/null, then you have the best of
all cases, and it should be no problem. However, if you're viewing an
HD MP4 video, the CPU has to divide its time between dealing with the
USB port data, and dealing with displaying the movie in real time. If
both require the full attention of the CPU, there's going to be
visible dropouts (pixelation). It won't be so much the lack of
performance on the USB port, as it will be the inability of the CPU to
keep up with the video stream. It may look like a 802.11n slowdown,
but it's really a video processing problem.

Yet I get far better performance from my D-Link Expresscard, reported
by the task manager and the performance monitor. File transfer times
also reflect this.

I have no problem dealing keeping up with the video stream unless I've
got conflicting security packages. These figures are from the
performance monitor.

If anything the current problem area is the Netgear wireless router.
If I lose my broadband connection it seems to forget work on the local
network while trying to recover. And it's easy to overload it
downloading from torrents (which I don't do all that often)I don't
often download torrents but when I do

In either case applications lose their connection to files on the
networked disk box and don't reconnect when it recovers.

If you wanna do some useful benchmarks, look into iperf and jperf.
<http://openmaniak.com/iperf.php> (tutorial)
<http://sourceforge.net/projects/iperf>
<http://sourceforge.net/projects/jperf>

I'll try these.

These should give you performance data (i.e. throughput) without
simultaneously benchmarking a video application, or hard disk speed.
For confirmation of my guess, also try USB wireless benchmark, while
watching an HD MP4 movie from the local hard disk. I predict you'll
see a major slowdown.

I run HD MP4 off a shared network disk. No problems at all with the
300 mbps express card but stuttering if I use USB.

I've used Netgear and D-Link N USB, and while they've stayed up I
never operated at anything close to the rated 270 mbps. The only
device I have that works as specified in a D-Link Expresscard that
gives me a consistent 300 mbps anywhere in the hose.

Careful there. DLink does something that I consider to be rather
devious. They only display the initial connection speed on the status
page. You may probably get 300Mbit/sec initial connection, when
there's no traffic moving, but as soon as you start downloading, and
the errors accumulate, the speed will drop. Most wireless client
managers do NOT display the inevitable drop in speed, or at best, when
the system slow down and then only after a long delay insuring that
the data cannot be used for tuneing. DLink does it even worse, but
displaying the initial connection speed, and leaving it there. The
bottom line is that you have to do the benchmarking yourself, and
usually with a 3rd party application, as the Microsoft diags usually
report what the NDIS driver is reporting, which is the same that Dlink
uses for its connection manager. Incidentally, sniffing the 802.11
management frames with Netstumbler to get the connection speeds is
useless because it would need to sniff 4 different streams
simultaneously. Actually, I've never bothered to try it, but am
fairly sure sniffing won't work.

This is as reported on the task manager and performance monitor. The
latter "adds things up" to get what should be a fairly accurate
throughput. Although it does have its own overheads.

This is the express card in the laptop.

Oddly enough their USB dongle and the Netgear equivalent report
varying speeds in the places you would look for it in.

My biggest gripe is that the suppliers just say "up to 270mbps"
without giving you any help. You're on your own finding all this out
the hard way.

Everyone lies, but that's acceptable because nobody listens. It's
fully capable of going 270Mbits/sec and can probably be demonstrated
in a Greenfield (i.e. perfect) environment or a laboratory. That's
about the same as a 150mph automobile, that can easily do 150mph on a
straight flat unoccupied road, but might have some difficulties on a
city street. Caveat Emptor.

People do listen though. Especially when they are planning a network.

And the guys in the stores aren't very helpful either.Prime example
was the expansion card I put in the desktop to add a PCMCIA socket
because of all the old PCMCIA cards from various laptops. "Yes, we
sell lots of different cards for laptops, what kind do you want?"
"It's an expansion card for a desktop PC so I can use all the old
laptop cards we all have". "No such thing".

See the data rate chart at:
<http://en.wikipedia.org/wiki/IEEE_802.11n>

It would be nice if there were a standard test environment suitable
for realistic benchmarking, but not this week. Important parameters,
such a room reflectivity, interference, and wall construction are just
too difficult to duplicate. So, the manufacturers supply an upper
limit. In my limited experience, you'll be luck to *maintain* speeds
greater than 54mbits/sec in a "typical" indoor environment.

Too much stuff from different manuacturers doesn't work properly with
each other, and even sometimes from the same one when they use
different chip sets.

No kidding. I'll just ignore that the 802.11n specification isn't
quite done yet. Also, that there are a growing number of
implimentations that claim to be 802.11n. For example:
1. Pure Beam Steering
2. Pure Spatial Multiplexing
3. Both Beam Steering and Spatial Mux.
4. Spatial Mux with anything from 1x2:2 (which I don't consider to be
MIMO but is common with USB adapters) to a max of 4x4:4.
5. All the above in a 20MHz channel or a 40Mhz channel.
6. With 802.11g protection on or off.
7. nd a whole mess of minor implimentation options which may make or
break a connection. Getting all this compatible between manufacturers
is not going to be easy.

One of the reasons I chose my router: Netgear claim that when the
standard is finalised this will reflected in a (probably just
firmware?) upgrade,

At least the Wi-Fi Alliance is testing for pre-802.11n standards
compliance, which should be a big help.

Yes.

Microsoft is also a bit flakey in this area and the application
developers haven't caught up either.

If you mean WZC (wireless zero config) is a bit flakey, that's an

I was being diplomatic.

understatement. It's permanently broken and misdesigned in XP. Vista
wireless is a huge improvement, but still lacking for things like
connection progress indication, error recovery, monitoring, dealing
with large numbers of AP's, etc.

I agree with all of these. Error recovery on the network is
particularly bad.

Maybe Windoze 7 will get it right.

Don't hold your breath. I only got Vista because that's what came with
replacement machines after two successive laptop mother board failures
and a desktop that fried itself during a heat wave when the air
conditioning failed.

I use the machines as tools. Having been a mainframe architecture
level hardware professional in my professional life I don't need to be
a PC nerd now.

Not being in the best of health it's not always convenient to use the
desktop PC so I need to do the same things using the same files from
the laptop. Hence the need for a wireless network that operates as the
manufacturers advertise.

In my never humble opinion, Intel Proset 12 is the closest
approximation of a useful and functional connection manager and
driver. Note that it took many years of constant improvments to make
it that way.

I don't think any of my adapters use the Intel chip set. I know that
when I didn't install properly the D-link came up with a default
Atheros driver.

To be honest I'm disappointed how little the manufacturers do to help.

And especially disappointed that they use different chipsets and
drivers in adapters that behave differently even though they're sold
in the same box with the same labeling.

David
....

Incidentally, I would be interested in your iperf or jperf results if
you decide to run them. Email address in signature below.
.

Relevant Pages

  • Re: explorer.exe freezes
    ... I think the thing to do here is replace my card for starters. ... early versions of USB were much slower than USB 2. ... I've had this card for about 3 years, before the USB connection ... slowest way to connect to the internet. ...
    (microsoft.public.windowsxp.general)
  • Re: Firewire on new iPods?
    ... > card slot, but that's probably a bit more expensive. ... USB 2.0 PCI cards for older Mac desktops ... can also charge from a computer USB 2.0 connection via the USB 2.0 ... USB 2.0 IOGear card) it will on manual SLEEP attempt the display blanks ...
    (comp.sys.mac.misc)
  • Re: Where are the bottlenecks?
    ... >> I'm using a USB 802.11b dongle on USB 1.0. ... can actually slow the speeds down noticably. ... call out to someone complaining about connection speeds who was running ...
    (uk.telecom.broadband)
  • Re: explorer.exe freezes
    ... I've had this card for about 3 years, before the USB connection version was ... a laptop with a pc/mcia wireless broadband card for Internet access. ...
    (microsoft.public.windowsxp.general)
  • Re: Capture Device not recognized in Movie Maker
    ... connection so I guess this is what I will need. ... card so I assume I will have to purchase ... you're not outputting digital video if you use USB. ... I have my video cam connected to my computer by USB. ...
    (microsoft.public.windowsxp.video)
Loading