PSU Fan Direction

[LASERandDVDfan]

As for

"parallel" port modems, I've heard they exist, but I've never seen
one. Serial port modems are always "hard".

I got the terminology mixed.

Choke it up to the always infamous brain-fart. - Reinhart

 

[LASERandDVDfan]

I have always found ACF/ACF2 chipped Rockwell/Conexant "hard" modems

to be reliable perfomers.

Cool. I've not had that much luck with Conexant myself. But, my experience
with them is limited.

Texas Instruments, on the other hand... - Reinhart

 

[Caroline]

Responses to Frank and Asimov follow, along with a few more general
observations.

What did this involve?

From an Earthlink technician, September, 2003:
--------
1. Go to Control Panel
2. Double click Dial-Up Networking icon.
3. Click the Connection icon. Right click to display "Properties." Choose
"Properties."
4. Click through General, Configure, Connection, Advanced.
5. Uncheck the box next to "Use Error Control." Click OK
6. Click "Options."
7. Click "Bring up terminal window after dialing. Click OK. Click OK.
8. Double-click the Connection icon.
9. Click the Connect button. Type user names and passwords as needed.

Under normal circumstances, after the login information has been transmitted,
you will see a message that says, "... starting PPP session." If you see a large
amount of "garbage" characters underneath this line, then most likely there is
some sort of interference on the phone line. Try connecting a different phone
cord from the wall to the computer. If you are using a splitter, try the
connection without the splitter. Contact your local phone company to conduct a
test of the lines.
-----

Six months ago I did indeed get a large amount of "garbage" characters.
Ultimately an Earthlink tech. and I blamed it on changing weather conditions
over a week-long period in my area. This past winter I didn't have the
disconnect problems I have today. I did another phone line check as described
above recently and again got a lot of garbage characters.

snip
Tell them you are having trouble sending/receiving faxes. The telco is
not obliged to provide "computer" grade phone lines, only voice and
fax.

Noted. Thanks.

Query the modem's last call diagnostic report, as described elsewhere
in this thread.

Okay.

Per your other post that talked about obtaining the last call diagnostic report:
I am using the directions at http://www.modemsite.com/56k/x2-hyperterm.asp and
http://www.modemsite.com/56k/usehyper.asp
to try to obtain this but so far no luck. I'm having problems at about step 5 of
the first site. But I'm unfamiliar with a lot of what exactly this is doing, so
I'll keep studying it, experimenting and trying.


Asimov:
I saw your comment about dirty RJ phone connections and will look into this.
Also, I understand about not telling the phone company I'm having modem
problems. Just say I hear noise and echoes on the phone line...

Going back to another conflict in this thread, and for what it's worth:
While my professional technical experience is not specialized to computer
hardware design, I do have power plant experience. I absolutely agree with those
saying that a computer power supply improperly electrically sized (e.g. watts
rating too low) may very well
cause computer freeze-ups. When a power source is overloaded, its output voltage
will
fall and potentially to a point that equipment it feeds will not operate
properly. Transients will be different and potentially detrimental when the
power supply is undersized. And so forth. I can't believe anyone with a real
technical background seriously disputes this.

With a PSU now twice the size of the original one, I am not getting the freeze
ups I used to get before. Before, I had to tread carefully before bringing up a
few applications simultaneously and quickly, due to seeing lock-ups before. Or I
avoided doing this at all. Now it seems I can start up as many applications as I
would ever want, as quickly as I want, and I do not get lock-ups.

Re the operating temperature of the CPU: Of course Ricky and others are right
that an overheated CPU (or other accessories) may detrimentally affect overall
computer
performance. Supercooled conductors are not being installed in certain, larger
electrical power applications just to waste money, and so forth. I'm not even
countenancing the one (I think) post that claims otherwise.

 

[w_tom]

Your (exampled) system was not running at 100% fine. It
crashed - one machine instruction did not process properly -
when peripheral load increased. System must work just fine
with all peripherals being accessed. And for most desktops,
that is 250 watts. You call it frozen or crashed. I also
call it 0% performance. The machine is 100% defective because
insufficient power caused a computer crash. Machine worked at
100% until power was insufficient; then failed. Did not slow
down. It crashed. That is a 100% defective machine.

But back to the orginal post which cites a completely
different concept. Let's say that power supply is sufficient
for motherboard but not sufficient for motherboard and
peripherals. An earlier poster said the motherboard will
process instructions faster if the power supply is sized
larger. No, CPU either processes instructions at 100% speed
or it crashes; for either supply. A larger supply does not
make the CPU execute faster - as another would have us
believe.

Same for temperature. Another suggested that if CPU got
warmer, then it might process slower. Again, nonsense.
Either that CPU works at 100% or heat creates timing changes
inside CPU: machine instruction no longer executes properly.
Machine works at either 100% or it crashes, freezes, works at
0%.

Many might say the problem is opening Explorer only because
they made a conclusion based only upon observation. That is
the formula for disaster. They said the same thing about
Challenger (it launched last time at below 32 degrees;
therefore will always launch fine at that temperature). It is
a basic concept of science that does not change even for
computers. The concept demands knowledge both from underlying
theory AND from experimental evidence. Anything less is
characteristic of a junk science conclusion.

To say Explorer crashed the system and to not first collect
basic facts is classic junk science. It was not Explorer that
caused the crash. It was an insufficient power supply whose
failure was not noted until after Explorer was started.

IOW a computer expert first learns basic electrical
principles - understand both the underlying theory and has
associated real world evidence. Having both learned those
principles and having experience by personally building
computers (at the IC level - not assembling a computer), then
I have long understood how heat and insufficient power can
crash computers. Also learned how to avoid those problems in
advance by using engineering principles and math, and by
taking data with appropriate tools.

Nothing in a power supply or in temperature will make a
computer execute faster (ignoring, for the moment, protection
circuits found in Intel processors). Either it runs at 100%
or it crashes.

Suggested here is that the earlier poster said something
different from what your example demonstrates. A probable
reason for confusion or misunderstanding?

 

[w_tom]

The fastest way to identify an undersized power supply is to
load up many tasks simultaneously - to access many peripherals
simultaneously. Then monitor DC voltages with a 3.5 digit
multimeter. No appreciable voltage drop should be measured;
voltage should not drop into the lower 1/4 of limits if supply
is sufficiently sized.

A major difference exists between super cooled verse a
semiconductor running at cooler room temperatures. Currently,
supercooled conductors are only for superconductivity - to put
more power through smaller wires (ie Detroit Edison). Totally
irrelevant to computers and semiconductors. In power
transmission systems, a cooler wire can carry more current.
Again completely irrelevant to how fast a CPU works.
Furthermore, a CPU cooled too much can fail. Digital
electronics have both an upper limit and a lower limit for
proper operation. Heat does not make digital electronics
faster or slower.

Concerning step 5 of the Hyperterminal procedure at
http://www.modemsite.com/56k/x2-hyperterm.asp :
Instead of entering just AT, enter AT&F . IOW modem may
respond to commands but not echo them back to screen, or be
setup for other responses. Enter AT&FL3DT. Modem should at
least sound dial tone on speaker or (very faintly) click an
off hook relay.

Again, you want to test modem with a direct connection into
NID; modem connection not via existing household phone wires.
You must confirm problem is on their (telco) side; not on side
of NID that you are responsible for.

 

[Franc Zabkar]

Another thing, too. Ever since I can remember, computer rooms have been ran
around 68-72 degrees, to keep the computers cool. So, if cooling a chip is
not important, how come companies continue to improve on the cooling systems
of the Processor Chips in computers? How come Muti Million dollar companies
spend more money on the conditions of their computer rooms (not meaning
actual hardware), then they spend on insurance in a year (Just
exaggerating)? There has to be some logic to it. I am sure Microsoft would
not spend money where it is not needed.

Rick

Of course cooling a chip is important, but there comes a point where
factors other than temperature will limit performance gains. For
example, the capacitance of a p-n junction will limit the switching
speed of a logic gate.


- Franc Zabkar

 

[Franc Zabkar]

Responses to Frank and Asimov follow, along with a few more general
observations.



From an Earthlink technician, September, 2003:
--------
1. Go to Control Panel
2. Double click Dial-Up Networking icon.
3. Click the Connection icon. Right click to display "Properties." Choose
"Properties."
4. Click through General, Configure, Connection, Advanced.
5. Uncheck the box next to "Use Error Control." Click OK
6. Click "Options."
7. Click "Bring up terminal window after dialing. Click OK. Click OK.
8. Double-click the Connection icon.
9. Click the Connect button. Type user names and passwords as needed.

Under normal circumstances, after the login information has been transmitted,
you will see a message that says, "... starting PPP session." If you see a large
amount of "garbage" characters underneath this line, then most likely there is
some sort of interference on the phone line.

Actually, those "garbage" characters are probably legitimate binary
data which make up the PPP packets. In fact you should see these same
characters with EC (error correction) enabled.

As I understand it, after an initial text based login sequence, the
dial-up session switches to binary PPP mode. This means that any
strange characters received during the username and password phase are
genuine garbage, while those received after the session switches to
PPP mode may or may not be so.

In any case, a better test for line quality is to enable EC and then
query the modem's last call diagnostic report. The modem will tell you
how many "I-frames" were transmitted and received, and how many of
these resulted in errors. On a good connection I get about 1 error in
1000 packets, while on a bad connection I can get as many as 1 in 20.
You can minimise the error rate, and therefore maximise your
throughput, by limiting the modem's max connect speed.

See http://www.modemsite.com/56k/x2-linklimit.asp

BTW, if you choose to disable EC, you can get an idea just how noisy
your last connection was by viewing the total CRC errors in your
ppplog. Note that in the case of external serial modems some of these
CRC errors may be caused by serial buffer overruns.

See http://www.modemsite.com/56k/ppp.asp

Okay.

Per your other post that talked about obtaining the last call diagnostic report:
I am using the directions at http://www.modemsite.com/56k/x2-hyperterm.asp and
http://www.modemsite.com/56k/usehyper.asp
to try to obtain this but so far no luck. I'm having problems at about step 5 of
the first site. But I'm unfamiliar with a lot of what exactly this is doing, so
I'll keep studying it, experimenting and trying.

Have you enabled command echo with ATE1? Until you do this, you may be
typing blind.

BTW, if you wish to retrieve the modem's last call diagnostic report,
avoid using the AT&F command just prior to doing so. On some modems
this will reset the diagnostic data. On others, like my own
Rockwelloid, AT&F resets the "reason for disconnect", but leaves the
data intact.

With a PSU now twice the size of the original one, I am not getting the freeze
ups I used to get before. Before, I had to tread carefully before bringing up a
few applications simultaneously and quickly, due to seeing lock-ups before. Or I
avoided doing this at all. Now it seems I can start up as many applications as I
would ever want, as quickly as I want, and I do not get lock-ups.

I have to wonder about how PSUs are rated. Your initial post stated
that your PSU was a 90W device which, at face value, seems seriously
low. But then I doubt that many, if any, PC PSUs actually deliver
anything near their claimed output. In fact I've only ever heard of
one PC PSU that even regulates properly. By that I mean that the vast
majority of PSUs "regulate" by taking a weighted average of the +5V
and +12V rails, and then assume that the other supply rails will fall
within spec solely on the basis of their turns ratio. A poor design,
IMHO ...


- Franc Zabkar

 

[Caroline]

C wrote
snip but comments noted

Have you enabled command echo with ATE1? Until you do this, you may be
typing blind.

Evidently. No idea how to enable anything. I need to dig into this a lot more.
(It's probably way more efficient for me to do this than to have you
troubleshoot via an internet forum, though your trouble is appreciated.)

BTW, if you wish to retrieve the modem's last call diagnostic report,
avoid using the AT&F command just prior to doing so. On some modems
this will reset the diagnostic data. On others, like my own
Rockwelloid, AT&F resets the "reason for disconnect", but leaves the
data intact.

I haven't given this another shot yet but likely will sometime in the next few
months.

Seems like most of my disconnects occur when I'm typing a long email or Usenet
post. I've checked my Outlook Express settings and some other modem-related
setting involving being "idle," but no luck.

I have to wonder about how PSUs are rated. Your initial post stated
that your PSU was a 90W device which, at face value, seems seriously
low.

No, I think it's right. This was and still is a very barebones computer. I don't
do video games. I have only a single hard drive, CD-Rom drive, and floppy drive.
Nothing else extraordinary at all.

The old power supply has a label on it that says the max. output is only 70
Watts. This contrasts with what Gateway and other sources say about it being a
90-watt power supply. Probably a little semantical distinction about "power
rating."

But then I doubt that many, if any, PC PSUs actually deliver
anything near their claimed output.

This would be consistent with what I say above.

In fact I've only ever heard of
one PC PSU that even regulates properly. By that I mean that the vast
majority of PSUs "regulate" by taking a weighted average of the +5V
and +12V rails, and then assume that the other supply rails will fall
within spec solely on the basis of their turns ratio. A poor design,
IMHO ...

Haven't looked at this.

 

[Ricky Eck]

Seems like most of my disconnects occur when I'm typing a long email or Usenet
post. I've checked my Outlook Express settings and some other modem-related
setting involving being "idle," but no luck.

Ok, this MAY be a possability. SOME IP's reserve the right to "Disconnect"
you if your connection is idle for a certain, prederterminad amount of time.
I.E. NetZero would disconnect me after two hours if there was "No User
Signal" sent. What that ment, is if I was not actualy typing in a web page,
or something like that, it concidered me idle. Even if my mail was checking
every 10 min. I use a program called "TrackPass", through RealOne. It is
for my racing, that gives me real time stats during a Cup Race, and Time
Trials. Anyways, NetZero did not count this as a "User Signal", mainly
because the computer was doing all the transmiting. It was computer
generated. How in the world it knows this, I don't know. But it does. I
don't know how many times, during the short time I used dial-up between my
moves, that my TrackPass got shut down, because NetZero dissconected on me
(unlimited internet, my Butt, and don't get me started on that stupid
HighSpeed Junk they offer... Now, is this the same problem for you? I
don't know. Could be. I don't know how long a "Long Email" is. However, I
would look at your TOS provided by your ISP, and see if this could apply.
Just an Idea.

The old power supply has a label on it that says the max. output is only 70
Watts. This contrasts with what Gateway and other sources say about it being a
90-watt power supply. Probably a little semantical distinction about "power
rating."

This may be an idea. The 70 Watt may be RMS, and 90 Watt PTP? (Many Amps
claim this, 100 watts, then you look in fine print "PTP", RMS is actualy
like 75 Watts or some other number. Just made a number up) Or, Gateway may
have screwed ya.. LOL Na, a major company screwing a Cust. It could
never happen. I know I have a Gateway, and I am happy with it so far. No
Probs, and have had it for about 1.5 years. Just hope it stays that way.
My Mom has a Dell, and had nothing but Probs with hers.

Rick

 

[Caroline]

Ok, this MAY be a possability. SOME IP's reserve the right to "Disconnect"
you if your connection is idle for a certain, prederterminad amount of time.
I.E. NetZero would disconnect me after two hours if there was "No User
Signal" sent. What that ment, is if I was not actualy typing in a web page,
or something like that, it concidered me idle. Even if my mail was checking
every 10 min. I use a program called "TrackPass", through RealOne. It is
for my racing, that gives me real time stats during a Cup Race, and Time
Trials. Anyways, NetZero did not count this as a "User Signal", mainly
because the computer was doing all the transmiting. It was computer
generated. How in the world it knows this, I don't know. But it does. I
don't know how many times, during the short time I used dial-up between my
moves, that my TrackPass got shut down, because NetZero dissconected on me
(unlimited internet, my Butt, and don't get me started on that stupid
HighSpeed Junk they offer... Now, is this the same problem for you? I
don't know. Could be. I don't know how long a "Long Email" is. However, I
would look at your TOS provided by your ISP, and see if this could apply.
Just an Idea.

I don't think this is what is happening. I'll get thrown off the connection
after maybe 20 minutes of typing an email or Usenet post.

I am trying to get people off this problem because, with all due respect and
affection for the great analyzers, engineers, and technicians in the world, you
can't be aware of the nuances to my problem. Like I bought the house (where my
computer now is) in August of last year. The disconnect problems started in
September. They went away as the weather seemed to change. Then they returned
about a month ago (after we had mega-rain, which has stopped, but... ). Prior to
August, in my former residence, I had not had disconnect problems. I have not
been monkeying with the software. So yes it may very well be my new
(10-year-old) house's phone line problems... Some of you are no doubt slapping
your head as you speak, and hopefully killfiling this thread at this point ;-)
Sorry. I find it hard to convey every detail in these posts. It's why I tried to
keep the threads narrow. I really just wanted to know whether reversing the fan
direction had ever been a problem in anyone's experience!

BTW The fan flow reversal does not seem to be a problem. I'm still gonna buy a
little heat sink fan, as Ricky and others I think suggested, sometime soon, as I
figure at a minimum it might help the overall life of the CPU.

Aside: Someone asked about how often I clean the inside of the case and its
hardware. Answer: About once a year. Any more often and there wouldn't be much
to clean.

This may be an idea. The 70 Watt may be RMS, and 90 Watt PTP?

snip

Oops. I made a mistake. The nameplate on the old power supply says: "max.
combined power on +5V & +3.3V output is 70 watts." But as I'm sure you know,
there's a +12 Volt output, too. On my computer this output is rated at 1.5 A.
This of course is an additional 18 Watts. Thus the nameplate also says "90 W.
Max." (Close enough.)

 

[Ricky Eck]

I don't think this is what is happening. I'll get thrown off the connection
after maybe 20 minutes of typing an email or Usenet post.

I am trying to get people off this problem because, with all due respect and
affection for the great analyzers, engineers, and technicians in the world, you
can't be aware of the nuances to my problem. Like I bought the house (where my
computer now is) in August of last year. The disconnect problems started in
September. They went away as the weather seemed to change. Then they returned
about a month ago (after we had mega-rain, which has stopped, but... ). Prior to
August, in my former residence, I had not had disconnect problems. I have not
been monkeying with the software. So yes it may very well be my new
(10-year-old) house's phone line problems...

I wasn't tring to drag the post, I just seen you post that you got booted
off after writing a long letter. That popped into my head that NetZero used
to do that.

Rick

 

[Caroline]

[After installing a power supply twice the watt rating of the old power
supply--]

Performance seems better. For example, it seems I can dump more tasks on the
computer at the same time. (This sometimes froze things up.) I am still getting
some spontaneous Internet disconnects but am now suspecting the phone lines and
weather variations in my area.

To be complete, I have an update:

I switched to the integrated modem ("56K AC-Link Voice Modem") two days ago. I
also switched the telephone access number I use with Earthlink. My rate of
disconnects has gone down considerably--only one in the last 2.5 days, vs.
half-a-dozen or more daily in the recent past. (My time on the computer and
typing messages did not change notably in the last 2.5 days.)

I'll switch back to the external modem soon and see if the problem was a
"crowded" (or otherwise deficient-in-some-way) access number.

 

[w_tom]

Did you make the direct connection to NID - to determine if
noise is from internal or external source? Demonstrated is
one modem that tends to be more noise resistant (has a better
DSP algorithm) than the other. Test still does not identify
source of noise and resulting disconnect. You still have
disconnects. Therefore you are not yet identifying reason for
failure. Long before we discuss fixing the problem, first
concentrate on identifying the problem. Hyperterminal has
always been a powerful tool for identifying a modem verse line
problem in part because it is so simple to use and because it
provides fundamental facts without the summarizations (basic
data removed) from Windows.