| Author |
Message |
J. Clarke
Guest
|
 Posted:
Sun Nov 21, 2004 9:41 pm Post subject:
Re: Confirm: Which is Faster - External Versus Internal Mod |
|
|
TE Cheah wrote:
| Quote: | | Have you disabled data compression on your modem?
No, as shown in mdm log.
| Even a crappy old v34+
| modem (33.6k) can overstuff a 115.2KB serial port.
Unless you have a 16650 / 750 / 950 UART & can download ( from a v34
mdm ) more bytes per second than with a 16550 UART, I'll believe ZDnet's
6-95 article.
|
Believe what you want to, but (a) ZDnet is hardly authoritative, and (b)
find out what standards have been released since.
In the computer industry 10 years is an eternity.
--
--John
Reply to jclarke at ae tee tee global dot net
(was jclarke at eye bee em dot net) |
|
| Back to top |
|
 |
Franc Zabkar
Guest
|
| Posted:
Mon Nov 22, 2004 12:59 am Post subject:
Re: Confirm: Which is Faster - External Versus Internal Mod |
|
|
On Sun, 21 Nov 2004 16:14:43 +0800, "TE Cheah" <no@spam.biz> put
finger to keyboard and composed:
| Quote: | | Are you claiming that an external modem can somehow push data into the
| computer faster than the rate of the serial port used for the
| connection?
No, ZDnet's PC Magazine ( USA version ) June'95 issue's article had said
that 115.2 kbps DTE speed is not a bottleneck to any DCE speed <128 kbps
e.g. ISDN 1 B channel's 64 kbps ( however compressible data is ) , let alone
a v92 mdm's maximum 56 kbps.
|
My modem's DTE rate is set to 230400bps. When downloading this page
....
http://modemfaq.home.att.net/manfs.htm#Contents
.... Win95 System Monitor reports a steady throughput of 16KBps for the
most part, with a maximum of 18KBps. The DCE speed during the entire
V.90 dialup session is 46667bps, compression is V.42bis, and images
are turned off. My ISP does not support software compression.
Setting the port rate to 115200bps would limit the transfer rate to
11.5KBps.
| Quote: | | that's 2249618 bytes in 693.09 seconds, about 3246 bytes/sec.
What about upload time, servers ( of ISP & websites )' congestion ?
| theoretical limit for a 115200,N,8,1 serial port is 11,520
| bytes/sec.
I think a N,8,1 connection uses only 9 bit per byte.
|
No, on the DTE side the word format is 1 start bit, 8 data bits, 1
stop bit.
See my post entitled "Test results - compression and error
correction":
http://groups.google.com/groups?selm=3b5ff0c7.6013011%40news.dingoblue.net.au&output=gplain
- Franc Zabkar
--
Please remove one 's' from my address when replying by email. |
|
| Back to top |
|
|
Geoffrey Welsh
Guest
|
| Posted:
Mon Dec 13, 2004 9:16 pm Post subject:
Re: Confirm: Which is Faster - External Versus Internal Mod |
|
|
J. Clarke wrote:
| Quote: | v.42bis provides up to 4:1 compression for some types of data. More
recently, v.44 was released, which provides even greater compression.
|
That 4:1 figure is utterly baseless: the "up to" is an outright lie, as
V.42bis could, IIRC, provide over 100:1 compression on sufficiently
repetetive data*, and the 4:1 number was probably chosen to one-up MNP7
(which claimed "up to 3:1"), and that number was probably chosen to one-up
MNP5 (which claimed "up to 2:1"), and that number was probably just chosen
because doubling your throughput sounded particularly good. Even then,
those numbers appeared in marketing drivel; they are not quoted in the
Microcom technical documents that I have.
Now let's get to the nitty-gritty: how much compression do you actually get
on REAL data? On GIFs, JPGs, MP3s, MPEGs: NONE. Precompressed data doesn't
compress well and, in many circumstances, can actually GROW when data
compression techniques are applied. Fortunately, V.42bis detects when it's
losing ground and actually turns compression off when it hurts more than it
helps.
| Quote: | Since the compression and decompression take place in the modem for a
controller-based modem, the port speed does become the bottleneck.
|
Not unless you're transmitting ridiculously repetetive data (such as
database dumps in uncompressed files) and, even then, if you're doing it
over TCP/IP the layers of overhead and framing hurt the compression rate.
* Welsh's Law of Data Transmission: the _information_ rate of a compressed
channel is a constant, regardless of the data rate, because compression by
definition optimizes redundant data. Corollary: the only way to get
extremely high throughput via extremely high compression is to send data so
repetitive that it contains practically no useful information and is
therefore not worth sending at any speed.
--
Geoffrey Welsh <Geoffrey [dot] Welsh [at] bigfoot [dot] com> |
|
| Back to top |
|
|
J. Clarke
Guest
|
| Posted:
Tue Dec 14, 2004 12:23 am Post subject:
Re: Confirm: Which is Faster - External Versus Internal Mod |
|
|
Geoffrey Welsh wrote:
| Quote: | J. Clarke wrote:
v.42bis provides up to 4:1 compression for some types of data. More
recently, v.44 was released, which provides even greater compression.
That 4:1 figure is utterly baseless: the "up to" is an outright lie, as
V.42bis could, IIRC, provide over 100:1 compression on sufficiently
repetetive data*, and the 4:1 number was probably chosen to one-up MNP7
(which claimed "up to 3:1"), and that number was probably chosen to one-up
MNP5 (which claimed "up to 2:1"), and that number was probably just chosen
because doubling your throughput sounded particularly good. Even then,
those numbers appeared in marketing drivel; they are not quoted in the
Microcom technical documents that I have.
|
It was all based on "average text" and in an application that I worked with
at the time that involved the transmission of a great deal of text the 4:1
was pretty close. And it was better than MNP5-we didn't have any hardware
that supported MNP7.
| Quote: | Now let's get to the nitty-gritty: how much compression do you actually
get
on REAL data? On GIFs, JPGs, MP3s, MPEGs: NONE. Precompressed data
doesn't compress well and, in many circumstances, can actually GROW when
data
compression techniques are applied. Fortunately, V.42bis detects when
it's losing ground and actually turns compression off when it hurts more
than it helps.
|
Well all of that is just fine but it's irrelevant--if you're transmitting
that kind of data then V.42bis doesn't help you and you're limited by the
line speed. But everybody doesn't transmit that kind of data.
| Quote: | Since the compression and decompression take place in the modem for a
controller-based modem, the port speed does become the bottleneck.
Not unless you're transmitting ridiculously repetetive data (such as
database dumps in uncompressed files) and, even then, if you're doing it
over TCP/IP the layers of overhead and framing hurt the compression rate.
|
Somewhat. But that does not alter the fact that the port speed can be the
bottleneck if compression and decompression occur in the modem. And
everybody doesn't use TCP/IP for all data transmissions.
| Quote: | * Welsh's Law of Data Transmission: the _information_ rate of a compressed
channel is a constant, regardless of the data rate, because compression by
definition optimizes redundant data. Corollary: the only way to get
extremely high throughput via extremely high compression is to send data
so repetitive that it contains practically no useful information and is
therefore not worth sending at any speed.
|
Whoever "Welsh" is. If you know something about the data then you can make
use of that knowledge. In text-compression algorithms for example short
codes are often used to transmit the most common characters. In video
compression only the changes from the previous frame get transmitted, and
so on. You can claim that compression never works until you're blue in the
face, in the real world it works fine on many classes of data.
--
--John
Reply to jclarke at ae tee tee global dot net
(was jclarke at eye bee em dot net) |
|
| Back to top |
|
|
J. Clarke
Guest
|
| Posted:
Thu Dec 16, 2004 5:18 am Post subject:
Re: Confirm: Which is Faster - External Versus Internal Mod |
|
|
Geoffrey Welsh wrote:
| Quote: | J. Clarke wrote:
Whoever "Welsh" is.
Uh, did you miss the "from" field?
If you know something about the data then you can make use of
that knowledge. In text-compression algorithms for
example short codes are often used to transmit the most common
characters. In video compression only the changes from the previous
frame get transmitted, and so on.
Thanks for the primer, but in addition to using modems in a wide variety
of applications over the past two decades, I've also implemented data
compression in a couple of applications.
|
Well aren't you special.
| Quote: | You can claim that compression never works until you're blue in the face
Did you read my message any better than you read the from field, or is
this an intentional attempt to discredit me by lying about what I said?
|
Perhaps if you would state your thesis succinctly in one sentence before
defending it it might be easier to figure out your point. If you are
claiming that the port speed is never ever a bottleneck you're full of
crap.
--
--John
Reply to jclarke at ae tee tee global dot net
(was jclarke at eye bee em dot net) |
|
| Back to top |
|
|
Geoffrey Welsh
Guest
|
| Posted:
Thu Dec 16, 2004 5:18 am Post subject:
Re: Confirm: Which is Faster - External Versus Internal Mod |
|
|
J. Clarke wrote:
| Quote: | Whoever "Welsh" is.
|
Uh, did you miss the "from" field?
| Quote: | If you know something about the data then you can make use of
that knowledge. In text-compression algorithms for
example short codes are often used to transmit the most common
characters. In video compression only the changes from the previous
frame get transmitted, and so on.
|
Thanks for the primer, but in addition to using modems in a wide variety of
applications over the past two decades, I've also implemented data
compression in a couple of applications.
| Quote: | You can claim that compression never works until you're blue in the face
|
Did you read my message any better than you read the from field, or is this
an intentional attempt to discredit me by lying about what I said?
--
Geoffrey Welsh <Geoffrey [dot] Welsh [at] bigfoot [dot] com> |
|
| Back to top |
|
|
Franc Zabkar
Guest
|
| Posted:
Thu Dec 16, 2004 11:36 am Post subject:
Re: Confirm: Which is Faster - External Versus Internal Mod |
|
|
On Mon, 13 Dec 2004 11:16:16 -0500, "Geoffrey Welsh"
<reply@newsgroup.please> put finger to keyboard and composed:
| Quote: | J. Clarke wrote:
v.42bis provides up to 4:1 compression for some types of data. More
recently, v.44 was released, which provides even greater compression.
Since the compression and decompression take place in the modem for a
controller-based modem, the port speed does become the bottleneck.
Not unless you're transmitting ridiculously repetetive data (such as
database dumps in uncompressed files) and, even then, if you're doing it
over TCP/IP the layers of overhead and framing hurt the compression rate.
|
What about my example from John Navas' FAQ? Furthermore, when I was on
dialup I often saw transfer rates in excess of the standard port rate
when downloading .pdf datasheets. Admittedly most web pages do not
exceed 11.5KBps, especially when using v.42bis, but I download a *lot*
of datasheets.
- Franc Zabkar
--
Please remove one 's' from my address when replying by email. |
|
| Back to top |
|
|
Geoffrey Welsh
Guest
|
| Posted:
Sun Dec 19, 2004 5:28 am Post subject:
Re: Confirm: Which is Faster - External Versus Internal Mod |
|
|
J. Clarke wrote:
| Quote: | Perhaps if you would state your thesis succinctly in one sentence
before defending it it might be easier to figure out your point. If
you are claiming that the port speed is never ever a bottleneck
you're full of crap.
|
I never say "never" - well, hardly ever. I even pointed out that
ridiculously high compression can be achieved in special circumstances. The
short thesis, basically, is that the average user need not lose sleep over
the throughput they're losing because of the 115200 bps serial port. For
most users, the average throughput will be well below that bottleneck and
momentary bursts of compressible data won't be significant enough to justify
going out of your way to bypass it. Franc's observation about PDF files is
interesting because the average dialup user might well spend a significant
amount of time downloading large PDF files (as opposed to my example of
uncompressed database dumps.)
--
Geoffrey Welsh <Geoffrey [dot] Welsh [at] bigfoot [dot] com> |
|
| Back to top |
|
|
Geoffrey Welsh
Guest
|
| Posted:
Sun Dec 19, 2004 5:36 am Post subject:
Re: Confirm: Which is Faster - External Versus Internal Mod |
|
|
Long time, no speak. Been busy. Not really back, just visiting.
Franc Zabkar wrote:
| Quote: | What about my example from John Navas' FAQ?
|
I did a quick search and came up with a chart that basically said connect
speeds over 50K result in higher than 11K/sec throughput, but no discussion
of how that number was arrived at. Do you have a URL for a more detailed
discussion?
| Quote: | Furthermore, when I was on
dialup I often saw transfer rates in excess of the standard port rate
when downloading .pdf datasheets. Admittedly most web pages do not
exceed 11.5KBps, especially when using v.42bis, but I download a *lot*
of datasheets.
|
Now, that I do find interesting. I should not have been surprised by it,
because I have noted that the content of PDFs varies very, very, very much,
with the worst looking like some kind of bitmap of a scanned page in stead
of text peppered with Postscript-style commands. I guess that repeated
commands compress much better than the text in which they're embedded, which
would also be true of RTF and even [SG|HT|X]ML documents to varying degrees.
And, of course, PDFs are more likely to be in a typical users' surfing
habits than my extreme example of uncompressed database dumps.
Thanks for the correction.
--
Geoffrey Welsh <Geoffrey [dot] Welsh [at] bigfoot [dot] com> |
|
| Back to top |
|
|
none
Guest
|
| Posted:
Fri Dec 24, 2004 8:02 am Post subject:
Re: Confirm: Which is Faster - External Versus Internal Mod |
|
|
On 17 Nov 2004 11:07:12 -0800, brucekg78@yahoo.com (Bruce Gillis)
wrote:
| Quote: |
Now, I am more confused than ever. I want the fastest 56K modem
around but I am uncertain whether to go with an internal or external
modem. Can someone knowledgeable about modems clarify this for me?
Much appreciated,
Bruce
|
I have a high-end 150 buck external (serial) 3Com business modem on my
machine and my wife has an 8 buck win modem based on the Lucent chip
set.
Overall the net practical difference between the two, in our
experience has been zero.
On our noisy rural lines they both consistently connect at identical
speeds and just as fast. About 44000 to 48000.
The only real difference is that the 3Com works better as a fax modem
because it has a more complete command set for use a fax modem.
In fact if I had to do it over again I would simply put in a pci win
modem like my wife's-get rid of all the clutter of power supply and
power cords, serial cable, plus the modem it's self which takes up a
fair amount of room. In addition I would free up a serial port.
Another thought-modems are prone to burnout from lighting storms and 8
bucks is a lot cheaper than 150.
I think the difference in performance between a hardware or software
modem might of meant something back in the days of the 486 but with
modern cpu's there really is no difference. |
|
| Back to top |
|
|
|
|
|
|
FAQ
Memberlist
Register
Profile
Log in to check your private messages
Log in