Hi,

I have problem with the Pre-N router and PC-Card. When using 2
connected computers, I don't get speeds above 15-20 MBit/s even though
the connection shows up as 108 Mbps in the network connections dialog
in XP. This is when the computers are placed side to side (optimal
distance).

Another question: Is there USB equipment for the Pre-N networks? Is
Belkin the only manufacturer?

ttestp***@hotmail.com (Robin2) wrote in
Are both computers using the Belkin pre-n card? If not then your
speed will be dropped to the rate of the non- pre-n.

Show quoteHide quote

On 9 Mar 2005 03:36:46 -0800, ttestp***@hotmail.com (Robin2) wrote:

Make and model?  If the "pc-card" is not Pre-N, you'll get about
25Mbits/sec thruput which is "conventional" 802.11g performance at
close range.  Not that the 108Mbits/sec is not the thruput but the
connection speed.  You'll get about half that in thruput.

Dunno.

Jeff Liebermann wrote:
Show quoteHide quoteYou can also play with router settings. Try the burst mode ack. Limit
the router to just handling the g protocol. Maybe different security
protocols will make a difference. These might help a bit.

Conventional G is 12 - 13 Mbps.

It's never half, more like 1/4 or 1/3.

On Thu, 10 Mar 2005 06:52:33 -0600, "JB"
<jbrandonbbrem***@hotmail.com> wrote:

Note that the original posting was for a Belkin Pre-N access point and
an unspecified client radio which I assumed to be an "ordinary"
802.11g client.  Also note that the test was performed "side by side"
which I presume would be under ideal conditions.  If you're only
getting 1/4 to 1/3 of the connection speed, then methinks something is
broken in your setup. 

From one of my previous postings...

I keep getting asked "how fast can it go" type questions.  Perhaps
some numbers might help.   This is stolen from an Atheros PDF at:
  http://www.atheros.com/pt/atheros_range_whitepaper.pdf
with some additions and corrections by me.

Non-overlapping    Modulation     Max      Max     Max
Channels -------        |         Link     TCP     UDP
802.11b        3       CCK         11      5.9     7.1
802.11g (with
  802.11b)     3   OFDM/CCK        54     14.4    19.5
802.11g only   3      OFDM         54     24.4    30.5
802.11g turbo  1      OFDM        108     42.9    54.8
802.11a       13      OFDM         54     24.4    30.5
802.11a turbo  6      OFDM        108     42.9    54.8

The paper claims that encryption is enabled for these calculations,
but my numbers seem to indicate that these number are for encryption
disabled.  Dunno for sure.  The Max TCP and Max UDP are the
theoretical maximum thruput rates.

I wasn;'t saying that's what I'm getting, it's the reality of wireless. If
the box says 54 or 108, you won't get half. You'll get 1/3 or 1/4.

- JB

On Thu, 10 Mar 2005 12:38:56 -0600, "JB"
<jbrandonbbrem***@hotmail.com> wrote:

Well, I can arrange to get far LESS than 1/3 or 1/4.  With 802.11g and
a 54Mbit/sec connection, I can usually get about half that up to about
10-15 feet away using the stock antennas.  If I move away furthur, the
signal gets weaker and the access point slows down.  Eventually, if I
go far enough away, the thruput will drop to about 1Mbit/sec before it
dies.

If you quote 12-16Mbit/sec "typical" performance, you should also
check and disclose the connection speed and test configuration.  I
just tried a customers laptop in my office.  27Mbits/sec from a Dell
5150 to a WRT54G at a distance of about 6 feet.  54Mbit/sec
connection.  That's about 50% of the connection speed which I consider
to be typical.

However, when I took the laptop outside, and going through two walls,
the connection speed went down to 18MBits/sec OFDM and the thruput
dropped to about 8.5Mbits/sec.  Again, this is about 50% of connection
speed.

Incidentally, I was using Netstat Live for performance testing:
  http://www.analogx.com/contents/download/network/nsl.htm

Jeff Liebermann <je***@comix.santa-cruz.ca.us> wrote:
My 54g delivers 26-27Mbps in near proximity to the WAP.
Fifty feet and a couple of walls away, the "current bandwidth" in
perfmon.msc bounces around on my 802.11g connection, almost like a signal
strength indicator, between 240000 and 48000.

I have tested with iperf to a wired desktop, and timed file transfers.

I would agree with Jeff that it is around half of the link speed, which is
consistent with the throughput I was seeing in 802.11b as well.

Good that you're enjoying that kind of access. My point, though, is that
most routers rated at 54 will run at a real-world 12. Also, routers rated at
108 or 125 will run at about 35 - 40. That's in normal conditions without a
lot of interference from maybe 10 feet (some routers will run slower from
shorter distances). I've tested 60-70 routers under all conditions, so I'm
just saying the average is 12. For example, the Apple Extreme G runs at 12.
And most of the networking companies -- as much as they talk about fast
access and theoreticl speeds -- will quote realistic numbers like 35 for
MIMO and 45 for Super-G.

- JB

On Thu, 10 Mar 2005 22:16:42 -0600, "JB"
<jbrandonbbrem***@hotmail.com> wrote:

Show quoteHide quoteMethinks you missed my point.  The TCP thruput is a roughly a fixed
percentage of the connection speed.  40% to 50% is typical.  The
connection speed varies with the S/N ratio and BER (bit error rate),
where the system intentionally slows down to keep the BER within
usable limits.  You can't just say "it runs at 12Mbits/sec" without
specifying the connection speed or at least the conditions of the
test.  There are no "typical", "average", or "all" conditions.  More
bluntly, any test conditions without numbers is worthless. 

For example, I have a typical 2 year old laptop running an unspecified
802.11g wireless card to a commodity access point across the room.
Based upon your experience, what thruput should I be getting?  Notice
that it would be almost impossible to predict without knowing the make
and models, antenna arrangements, range, reflective surfaces,
interference potential (glass wall facing the city), type of
benchmarking software, or protocols used (TCP vs UDP, VPN, WPA, WEP,
PPTP, etc).  All I'm asking is that if when proclaim that "most
routers rated a 54 will run at a real-world 12", that you specify
approximately under what conditions and distance the number was
obtained.  At the very least, specify if the stated speed is with
indoor (no walls), office (some walls), or outdoor, use.

For what it's worth, the office next to mine has an Apple Airport
Extreme 802.11g.  When I test their laptops (mix of iBooks and Dell
51xx laptops), they all work between 20-30Mbits/sec.  The range is
between 6 to 15ft.  However, there is one old Micron PIII W2K laptop
that only does about 10-12Mbits/sec that I've been unable to obtain
faster performance.  Since it's always on the road, it's been
difficult to confiscate and fix.  The user doesn't seem concerned as
it's fast enough for his use.

Incidentally, in another thread you seem to believe that the thruput
and performance are heavily dependent upon the processor.  That may be
true for really ancient laptops and computahs, but just about anything
faster than a Celeron 300 will do 25Mbits/sec thruput with any 802.11g
access point up to about 10ft.  More specifically, my ancient Micron
Transport 2 300Mhz PII laptop running WinME and a DWL-G630 rev B1
talking to a DI-624 does it nicely.

Show quote Hide quote This doesn't make any sense. So, it's 50% at best and lower with
interference. Huh?

You're talking about test conditions and results, I'm not. I'm talking about
general 54, 108, and 125 labels on boxes and what the real world results
will most likely become. Your approach is good for a specific router, PC
Card, environment, and such. Mine is just a general "it's about 1/3 or 1/4
of what the box label says."

I think it's kind of funny how you are asking me questions. It's like this.
If I said, when you go into that forrest, you might bump into a tree because
they are close together. So, you are dismissing that general advice by
saying, what is the size of the oak trees versus the maples, and what is the
specific diameter of each tree. Then, you throw out this blind question for
me: I bumped into a tree, can you tell me what kind it is based on all your
(ahem) "research." Huh, can you? Well, no. I never said I was talking about
trees, I'm talking about the forrest.

In wireless, there's either "this is what you might expect, maybe, something
to think about" or there is benchmarks on specific products under certain
conditions.

So, how do you know how fast the throughput is on that iBook? Because I know
you are not using Ixia products. And the Apple Extreme G is not *capable of
running at Super-G or Afterburner speeds, so I think you are confused about
the testing results.

There's a context to that thread that you didn't include here. In real-world
tests, the Belkin Pre-N runs at a suprrisingly different speed based on the
laptop you use. That doesn't mean the laptop is always such a huge factor
for every router.

On Fri, 11 Mar 2005 11:50:16 -0600, "JB"
<jbrandonbbrem***@hotmail.com> wrote:

Under ideal conditions, I get thruput equal to about 50% of the
connection speed.  If I add interference, the percentage will drop.

I ignore the labels and marketing hype.  50% of the connection speed
at best.  Anything you do beyond that makes it worse, such as adding
802.11b clients, inter fence, reflections, obstructions, other
co-channel users etc.

I only have one point to make.  Don't throw numbers into an answer
unless you are prepared to substantiate where you obtained those
numbers and under what conditions.  I don't think that's too much to
ask.

How fast does my car go?  Mine goes about 30 miles per hour on the
average.  How much does that tell you about how fast *YOUR* car will
go.  However, if I specify the conditions under which I came up with
30 miles per hour, you might be able to consider the 30 miles per hour
as useful information.

In engineering, I work with benchmarks and reproduce able numbers.
When I put on my marketing hat, I work with expectations, perception,
and a large collection of intangibles.  I literally have to do a
personality change to make the transition.

Timed file copy and various benchmarking programs running on a test
PC.  Find a big file.  Time how long it takes to download.  Do the
math.  It's not terribly accurate, but good enough.  However, I goofed
(memory fault).  I think it's 15-20Mbits/sec, not 20-30Mbits/sec.
I'll check again when I get to my office tonight.

See the graph at:

http://www.jiwire.com/apple-airport-extreme-base-station-modem-antenna-port-performance-lab-tests.htm
20 mbits/sec for Airport Extreme at 3ft.  You don't need Super-G or
Afterburner to get 20Mbits/sec.  I get that up to about 10ft.
However, I cheat.  The airport base station has a hang on antenna to
improve the signal.

I haven't done any Belkin Pre-N testing, so I'll take your word for
it. 

Incidentally, I avoid the term "real world tests" as it reeks of
anecdotal evidence and irreproducible results.  By avoid, I do use the
term where there is no way to predict or calculate performance, but
much prefer reproducible benchmarks and detailed calculations.  I
don't expect everyone (or anyone) to agree or adopt my personal
preferences, but it would be nice.  May I humbly suggest that if you
insist on offering "real world test" results, offer them as a range of
expected values, that cover perhaps 1 or 2 sigma bell distribution of
probably conditions.  Methinks that would be a bit more believable
than a single number.

In (ahem) real world tests, you'll get 1/3 to 1/4. If you get half, it's
because you are a) adding antennas, b) not testing correctly or c) fudging
the numbers.

That's why you're an engineer and I'm a writer. People want to understand
things, not wade through meaningless statistics. That's why, when I say most
standard G routers run 12 Mbps, it's useful information. And that's why when
I say Super G goes 45, MIMO goes 35, and Afterburner goes 40, it's something
that can guide people. So, in your car analogy (I prefer the trees example
myself), I would say most street cars can go 55 with no problem. Then, the
consumer can say, oh - so it's not a suped up sports car that can go 150, or
a junker from 1972 that can only go 45 in third gear -- I'm interested,
thanks for the adivce, gotta go see my Lexus dealer.
You're running a program on a PC to test an iBook? Interesting. Also, not
accurate. The only way to test is with Kismac, AirStumbler, or some other
tool that at least gives you a percentage of signal strength and a quality
indicator. But there's no QCheck for Mac (yet).

I'm just going to kind of leave this one alone.

On Fri, 11 Mar 2005 15:00:27 -0600, "JB"
<jbrandonbbrem***@hotmail.com> wrote:

Would you entertain 3rd party results?  See:

WRT54G does about 20-22Mbits/sec.
  http://www.tomsnetworking.com/Reviews-39-ProdID-WRT54G-9.php

Apple Airport Express at 24Mbits/sec.
  http://www.tomsnetworking.com/Reviews-181-ProdID-AIREXPRESS-8.php

I can dig out more if you want them.  54Mbit/sec connection yields
about half that in thruput.

Reminder... your first reply in this thread was:

  "Conventional G is 12 - 13 Mbps.
  (...)
  "It's never half, more like 1/4 or 1/3."

Wanna amend that in view of numerous 802.11g benchmarks demonstrating
approximately 40-50% thruput?


I think you had better ask your people what they want.  The one's I
know, that I work for, and most of those in this group want
explanations.  Judging by the email I receive, many want even more
details.  Personally, I find one-line unsubstantiated answers to be
less than useful.  For example, in another thread, you basically
suggested to the OP that he toss his Sony Vaio and purchase Wi-Fi
certified hardware.  That style of answer is no better than RTFM or
"search Google".

However, you are correct that there is a class of client and reader
that "just wants it to work".  No understanding is needed or desired.
Tell me what to change and I'll get on with life.  You're absolutely
correct that my style of writing is not for this type of person.  I
try to avoid them, but they're everywhere.  I have clients also like
that, who don't want to listen to explanations, or are too busy to be
concerned.  I generally have problems with them in that things have to
be done over and over and over again.  In most cases, they don't want
to learn how to do anything on their computah because it's so much
easier to just ring my cell phone in the middle of the night.  I
charge for the effort and they're happy.  If this is your target
audience, I wish you great success as it seems to be a growth market.
But, I don't really want to participate in delivering unsubstantiated
conclusions, irreproducible results, and magic numbers from nowhere.

Incidentally, I judge my acquaintances and friends by their
willingness and ability to learn.  I don't care what they already know
and I don't care about their IQ.  If they're learning, I try to help.
If they stagnate in some manner, I leave them to their own devices.

Touche.  I guess is you specify the hardware or vehicle, it might be
useful information.  I live in the mountains and driving fast is not
an option with my ancient pickup truck.  Actually, I just realized
that I can dive into my GPS logs and give you almost exact numbers.
For the path 3 days, my top speed was 55 mph.  The truck was moving
for 5 hours in the last 72 hrs over a distance of 55 miles.  Therefore
my average moving speed was 11 miles per hour and my average overall
speed was 0.069 miles/hr.  Isn't that more useful?

There is a relationship between signal quality (S/N ratio) and
thruput.  However, I'm only interested in the final output.  I don't
need to know the S/N ratio to measure the signal strength.  It's nice
to know S/N, connection speed, protocols used, encryption, etc, but
not necessary.

Frankly, I don't know much about OS/X and Mac's.  However, I do know
and love Unix.  So, I dive into the shell (terminal window) and use
ftp, rcp, and possibly rcmd for remote benchmarks from the command
line.  For UDP testing, I just cat a big file to an IP socket that
ends in either /dev/null or a named pipe to a counter program.  No wiz
bang gui-fied program needed.  If I'm really in a rush, I'll fire up
Windoze (NETBIOS over TCP/IP) file sharing and just time and copy a
known file size.  No big deal.  In this case, the test was to an
Airport Extreme with a Dell 5150(?) laptop plugged into the LAN port
(through a switch) via ftp.  PC's do talk to Mac's and versa-vica.

When I get busy, anything without a dollar sign in front of it gets
ignored including this discussion.  At this moment, I have the time.

Show quote Hide quote In what part of the country is 20 half of 54?

I noticed you conveniently switched your test results update here to the
Express, which is a completely different product from the Extreme G. The
Extreme G gets 12.

Keep posting those standard G test results -- most will be closer to 1/3,
some will be 1/4. You do realize that 20 Mbps is closer to 1/3 than 1/2
right?

I stand by that advice -- especially considering the cost of a PC Card these
days.

Somewhere around 90% of all customers want general guidelines. In a Media
Play today, I overheard a husband and wife asking a sales clerk how they can
get multiplayer on their Xbox when it is down in the living room. I was
going to jump in, but then the clerk gave a pretty good answer (wireless
router, wireless game adapter from Microsoft). Did this couple or their
bored teen care that much about detailed environmental specs? WPA or VPN?
Not a hoot. They cared about getting Mechwarrior (the game they were
holding) to run. Now, here's a question for you. If you were there, and they
said what kind of access speed should they expect for the 54, 108, 125, or
MIMO product on the shelf, as the clerk, what would you have said? And
before you answer, I want you to imagine the look on the teenager's face,
and think seriously about the other ten customers lingering nearby with
puzzled looks, and the manager scowling off in the corner thinking terrible
thoughts about the TPS reports he would be generating about you and the
stale coffee he would be administering to himself.

Uh-huh.

Show quoteHide quoteAnd you are using this example -- you know, things like GPS logs -- as a
sales pitch at a Lexus dealership? I think it's good you are not a writer or
a salesman!

Show quoteHide quoteYeah, and you'll get 12. And, also, Tom's didn't use a Mac for the test it
seems.

Okay.

On Sun, 13 Mar 2005 21:31:13 -0600, "JB"
<jbrandonbbrem***@hotmail.com> wrote:

Where does it say to always use the lowest number?   22Mbits/sec is
41% of 54mbits/sec.  OK, I'll concede that it's not exactly half.
I've gotten about 25Mbits/sec using IPerf at about 6 ft.

Yep.  TomsNetworking didn't have an Extreme G benchmark.  I figured it
was close enough.  I guess not.  I'll confess that I don't know much
about Apple products.  The Airport at the neighboring office is the
one that looks like a large while plastic mushroom with an aftermarket
hang-on antenna.  I guess that's an Extreme G. 

Here's some benchmarks by Mac.com

http://homepage.mac.com/techedgeezine/2004-0402_g-turbo_shootout1.htm
I'm not sure why they find it necessary to tabulate RSSI and "number
of bars" as some sort of benchmark.  They really buried the
performance test on page 9.
  Airport Extreme G Base Station -
  Apple Extreme G PC Card: 38 sec. = 1,129 KBps
  BAsante AeroLAN-XG 35 sec. = 1,258 KBps (10.3% faster)
  Apple Airport (.11B) PC Card : 82 sec. = 537 KBps
  (Note that this is KBytes not bits)
So, the 1129KBytes/sec = 9Mbits/sec, which sucks.  No clue what they
were trying to do.

CNET review offers 16.8Mbits/sec in their review at:

http://reviews.cnet.com/Apple_AirPort_Extreme_base_station_wireless_access_point/4505-3245_7-20852329-5.html?tag=top
I was getting 20-30Mbits/sec when I tested it using several laptops.
However, I may have screwed up and will retest when I get to the
office on Tues.

This is kinda weird.

http://www.tigerdirect.com/applications/SearchTools/item-details.asp?EdpNo=602207
The performance comparisons on the bottom of the page are from CNET,
but I couldn't find the original text on the CNET web pile.
18.4Mbits/sec for the Airport Extreme-G only.

Well, with the Apple Airport Extreme G Base Station, you're apparently
correct.  Most of the other benchmarks appear to be running in the
22-28Mbits/sec range.

Yep.  20Mbits/sec is 37% which is fairly close to 33%.  22Mbits/sec
(the higher number) is 41% which is about half way between 33% and
50%.  Would it help if I amend my estimate to "up to 50% of the
connection speed"?

Show quoteHide quoteI don't have the slightest idea what to tell them.  I'm the worlds
lousiest salesperson and would probably bury them in technobabble.  If
forced to supply an answer, I would probably cover my ass and give
them a range of performance values that conveniently are proportional
to price tag on the hardware.  After all, the customers do have some
vague idea that price and performance are proportional.

Now, plop yourself in my shoes.  You just burned a good chunk of the
marketing budget paying an overpriced lab to run performance and
reliability tests on your unreleased product.  The numbers come back
looking like someone had thrown darts at a target, with no semblance
of a bell curve.  Of course, the lab claims they did everything
correctly.  What numbers do you give them for the advertising
literature?  Marketing wants to use the best numbers and ignore the
rest.  Engineering wants to cover their ass and use minimum values so
that nobody comes back and demands that the field units perform to
published specifications.  Legal wants to use the tweaked FCC cert
numbers.  The ad agency is waiting.  What do you tell them?

Actually, this is a rhetorical question because such debates are never
this clear or obvious.  At one company, engineering makes all such
decisions.  In another, it's marketing, legal, or maybe the autocratic
president.  It's the same as in your store example.  The customers
(and managers) all have their expectations and agendas and it's
difficult to make them all happy simultaneously.

Actually, I do quite a bit of writing.  None of it is for a general
audience.  Most of it is research and technical summaries.  Some
science fiction business plans and lots of market analysis reports.  I
try to keep my explanations simple and often succeed.  It's difficult
in usenet news to judge the abilities of someone posting a question.
If I guess wrong, I may overwhelm a beginner or insult an expert.  I
get quite a bit of email asking for details so I think I can assume
that the general usenet audience are not all looking for a fast fix.

You didn't really answer my question.  So, I'll simplify it for you.
Please select which one *YOU* would prefer to be told (not what you
thing everyone else would prefer) for 802.11g performance.
1.  12Mbits/sec
2.  Up to 25Mbits/sec
3.  40-50% of connection rate.
4.  33% of connection rate.
5.  20-25Mbits/sec up to 6ft under ideal conditions, TCP, using IPerf
benchmark, on a 100baseTX-HDX connected computer to a XXXXX wireless
router and a XXXXX mini-PCI wireless card in a P4-2.4GHz laptop.
Obviously, I prefer #5.

Do you mean a Mac is slower than a PC for 802.11g benchmarks?  The
only Mac's I work on are from friends and accomplices, so I don't see
too many of them.  Frankly, I've never bothered to do a side by side
wireless comparison with a comparable PC laptop.

Well, actually I'm stalling.  The plumbing in the bathroom is clogged
and I'm waiting for my chemical attack to finish.

Show quote Hide quote Isn't it all about averages>? I mean, your ad agency would not want
techogibberish, and even if they did, they do not have the magic ring
decoders for it anway. But they want a number. So, I can get an Apple
Extreme G down to 5 Mbps no problem, and I can get a generic Linksys or
Netgear standard G up to 22. But under most conditions, where there might be
interference, chicken wire in the walls, etc. you can say something like 12.
But, of course, they never do use a number like that. They use 54 Mbps with
a disclaimer that your rate will be lower (without saying how much lower).
I use number 4.

A lot of people don't reaize how much the processor and OS can affect
throughput. The iBook runs at 700 MHz. On a PB G4, the Base Station Extreme
G is blazingly fast, running at about 18 Mbps from ten feet. Same
conditions, same router, same card, same everything, the iBook runs at
10-11.

thats normal for Belkin 'Pre-n' ...dont believe the marketing hype !

You should be able to get 35.

They may NOT work at their best when close together, try max distance but
still within the room and without encryption.

Although other manufacturers have, or will shortly have, MIMO AP's and
adapters but as yet there isn't agreed standard,
so it will be trial and error to see which kit works with another one.
Regards,
Martin

I've tried all possible distances between the wifi components..

Also, the Belkin PC-Card is a Pre-N card. To the best of my knowlegde,
WEP was not activated.

I get approx, 16 Mbit/s max and that's pretty constant, it never seems
to get higher even for a very short period of time.. Very odd. It
seems to stay on that 16 Mbit/s level for normal usage as well, ie.
doesn't dip below 16Mbit/s too often.

Ps. When using a 802.11g PC-CARD the speed maxed out at 11Mbit/s.
Quite odd that too, isn't it?

Do you have the latest PC Card driver?
Do you have G only turned on on the router?
Have you tested from 10 feet?
Do you have any other 2.4 GHz equipment such as microwave or portable phone
nearby?
What is the speed of your laptop?
Have you tried a speed test with just one computer cpnnected?
Is anyone else in your area connecting and you don't know it?

- JB

Linksys WRT54G now a brick....but wireless works...
VLAN & SSID
3Com 3crwe454g72
ICS not present in XP?
Restricting wireless networks a laptop can access
File Sharing Problem
will "enhanced" 802.11g devices help interference?
Animated tool for Wi-Fi?
HSB1 -AP Hawking Range Extender-Anyone try one? (versus Lnksys)
Linksys compability w/ HyperWRT