Booting Problem

Paul · Apr 8, 2013

species8350 wrote:

Yes, but I'm embarrassed to report, I haven't always succeeded in finding
the points in question.

Some will be rather obvious. If you're working with VCore (CPU main supply),
then you look at the caps and inductors for a hint of where to connect.
And the regulator chip will also be near by, and if you know where the
feedback input pin is, you might sample the voltage there.

But some of the regulator configurations are a bit less "regular" in design.
On my P4C800-E motherboard for example, they used linear regulators, designed
using transistors and op-amps, and I probed like crazy over the legs of those,
and I couldn't make any sense out of the results. None of the voltage levels,
aligned with what I was expecting.

And yes, you need to leave the tower powered up and running, when attempting
to make such measurements. The preferred method, is to pull the guts of
the tower out, and put them on a table. Then, the motherboard is in the open,
and it's much easier to work.

This is a poor example, but this shows how much easier it is to work
on a motherboard, if it's out in the open. Depicted, is a mad scientist,
in Taiwan :-)

http://www.legitreviews.com/images/reviews/1341/ecs_debugging.jpg

The hardest part, of operating a motherboard in that way, is
keeping the video card upright. You cannot work with the motherboard
out in the open, if there are any children in the area, as they'll
knock your video card over.

Paul

species8350 · Apr 9, 2013

At present I have one RAM module installed and am getting the very slow BIOS scroll.

Evidently Install mode is enabled (256 MB memory available). I tried insertinbg the Vista disk, but nothing was operational.

I managed to activate startup repair, but got the BSOD. System diagnostics not functional.

Boot sequence: system drives not present, and will not activate.

I noticed a small green light on the motherboard, just under the video card. Do you know what this means?

I tried a number of re-starts.

One: failed at checkpoint [WRIT]

In setup none of the drives are recogised and cannot be enabled.

Setup tells me that 512MB is installed.

I tried a system repair.

Failed C000021a Fatal System error. Shutdown.

I noticed the diagnostic lights are no longer green spheres, but are now green numbers (1,2,3 are green symbols).

At the end, I inserted a live ubuntu (9.1) disk into the cd drive and restarted. The hard disk (which passed a test) and CD drives are listed when I pressed F12, but the floppy and usb drives are not and will not activate.

Ubuntu booted up. I manged to use System Monitor and it told me that the memory is 242.9MB. The system then froze.

I mention all these observations in the hope that they may give a hint as to whether I have a RAM or motherboard problem.

Thanks

Paul · Apr 9, 2013

The green LED near the PCI Express x16 slot, is labeled as
"Standby LED" in the manual. I suspect it's the same as my
Asus motherboards here, and it's simply connected to +5VSB.
It's a means of warning you not to install or remove cards
or RAM modules, because +5VSB is still present inside the PC.
When the machine is switched off at the back, or is unplugged,
then you should see that green LED go off.

Your problem still seems to revolve around memory. Either a
Northbridge memory controller problem (i.e. bad motherboard),
or both memory modules have a problem. A bad VDimm regulator
could cause issues. The detection of 256MB however, is weird.
I'd suspect a missing control/address bus bit when that happens
(like, a dead I/O pad on the Northbridge). You would think that
trying the DIMM in all four different slots, that the symptoms
would change.

There's obviously a limit to "arm chair diagnosis", and at best,
we might be able to narrow things down to say "replace motherboard"
or "replace memory modules'. Not all hardware failures have
physically visible root causes.

*******

This is supposed to be a picture of your motherboard.

http://www.esaitech.com/images/detailed/11/UUU/img55897.jpg

I see four DIMM slots. I can think of four test cases I'd like
you to run.

Slot1 Slot2 Slot3 Slot4
DIMM1 DIMM2 Probably dual channel mode

Slot1 Slot2 Slot3 Slot4
DIMM1 DIMM2 Probably single channel mode

We can also reverse the position of the modules, like this.
Now, this case, does very little for the symptoms. All it does,
is move the defective words of memory around locally. So if
you set up the first test case above, and then tried this one,
you would not learn much. In other words, while dual channel
mode is recommended for best performance in regular usage,
it doesn't provide good diagnostic info when things are broken.

Slot1 Slot2 Slot3 Slot4
DIMM2 DIMM1 Probably dual channel mode

But, if you test the single channel case, like this:

Slot1 Slot2 Slot3 Slot4
DIMM2 DIMM1 Probably single channel mode

one of the sticks provides "low memory", and the other stick provides
"high memory". The symptoms then should track where the memory problem
has moved. For example, if really low memory, memory below 640K is
bad, the BIOS may not complete POST. The screen may remain black.
A diagnostic code shows on the diagnostic LEDs. Or, the computer
may just decide to crash in the BIOS.

If high memory is bad, this may upset an OS, but perhaps a program
like memtest86+ can manage to start.

I used this technique, single channel mode, to diagnose a module
with a completely dead memory chip on it, on my Nforce2 motherboard.
One config was completely dead, the other config allowed me to
run memtest86+, and it immediately identified the bad module
(memory was bad starting "half way up" in the address space).

The purpose of doing the two single channel mode tests, is to try to
determine which of the two modules is bad. It assumes a single module
has failed, and the intention is to try to figure out which module is
causing the problem.

But this would not be helping at all, if in fact it is VDimm or Vtt
terminator voltage which are out of spec. But your symptoms are
specific and weird enough, it doesn't really seem like that at the
moment. I just can't put my finger on what logic problem equates
to your symptoms (256MB detected on 512MB, used to work, DIMMs).

Try a subset of the four test cases described above. See if you can
figure out (via the information displayed by the BIOS during POST),
as to which DIMM configuration is single channel mode, then test the
two possible swappings of modules. And see if Ubuntu behaves any
differently.

Paul

species8350 · Apr 9, 2013

Thank you for your patience.

Later, I'lll try the DIMM tests, note the LEDs and BIOs messages, and let you know what I find.

The 256 MB memory message in the BIOS seems to be associated with what the BIOS calls: 'Install mode'. It does not give me the impression that only 256MB is present, but that 256MB is available for this specific mode(Install Mode) - what ever that is. I could not install from the Vista disk.

Did you consider the problem I raised regarding set-up and the fact that I could not enable the floppy and USB drives. The BIOS only sees tha the harddisk and CD drives, but others exist.

I agree with you when you say it be desirable to definatively determine that the problem I have is either associated with the RAM or motherboard, or even both.

Best wishes.

A.

Ps. Someone on ebay is advertising a Dell Inspiron E520 motherboard. Is this the same as the Dell Dimension E520 motherboard? I did ask them, but theyappear not to know. The photograph of the board indicates Inspiron. I am not buying at present, just researching.

species8350 · Apr 9, 2013

Regaging the graphic of the motherboard at:

http://www.esaitech.com/images/detailed/11/UUU/img55897.jpg

The side of the tower is in place at the moment, but as I recall I have two black and two white PCI slots. I can't say that this is my motherboard.

As a matter of interest, why are the memory slots colour coded balck and white?

Paul · Apr 9, 2013

Color coding (where present), is supposed to indicate dual channel pairs.
I generally like to consult a manual, to double-check what the colors may be
hinting at. But in general terms, it's to help a user put a matched pair
of DIMMs, in same-colored sockets, for best performance.

For single channel testing then, you'd be doing the opposite, and using
one color of each.

But it's best to use the manual as a check. On retail motherboards,
the first BIOS screen, the text part, will say "Single Channel Mode"
or "Dual Channel Mode", to signify what positions the modules are in.

There's more to it than that, but that's sufficient information to
help you with the two matched DIMMs you have in hand right now.
That limits the possible permutations and combinations.

*******

http://www.avforums.com/forums/moth...new-motherboard-dimension-e520-c2d-e4300.html

...................+++++
...................+FAN+
...................+++++

==========(black)
==========(white)

==========(black)
==========(white)

So that looks to me, like for dual channel mode, you could try
the two white slots.

And for the single channel mode test configurations, try a
black and white pair next to each other.

Paul

species8350 · Apr 11, 2013

Back soon

species8350 · Apr 13, 2013

I have now tried the studies that you suggested., ie,

Mem1 Mem2 --
mem1 - Mem2 -

mem2 mem1 --
mem2 - mem1 -

In all four cases the BIOS moved about one quarter across the screen and stopped.

In all four cases I observed green lights (not the numbers) at positions 3 and 4.

From the lights diagnostics this indicates memory failure, however, I noticed that when using only one RAM module that on occasions the BIOS crawled across the screen, although I had very little control - no mouse, only F12 on the keyboard, no detection of drives, etc.

Memory or motherboard?

Thanks

Paul · Apr 13, 2013

If it was the memory, the symptoms would not be the same for each case.

Your memories are doing the same thing, whether installed in one pattern,
or reversed. Single channel mode, if there was a difference between
one configuration and another, it would show one module as being
more defective than the other. When I was debugging a memory failure
here, an entire dead chip on one of two memory modules, one single channel
configuration booted, the other one crashed in the BIOS. So night and day
difference in symptoms. You're not seeing that.

I think it's motherboard.

The thing is, much of your motherboard is working.

All I can guess at the moment, is the problem is near the DIMM sockets,
or is a problem with the memory controller. The problem seems to relate
to the number of memory accesses. The more memory present in the system,
the worse it works. It implies the memory is working, like, 99.5% normally,
and a few accesses are getting corrupted. Like VDimm is too low, or Vtt
is not centered in voltage, half way between VDimm and GND.

So to fix it, in a "swap it out world", you'd change motherboards.

That's all I can suggest, looking at your problem from here.

You've done enough test cases, I can't think of any more to shed more
light on it. Borrowing a single stick of RAM, known good, would
be another test case. If it behaves the same as all the other RAM
tests, then you swap the motherboard.

If I was doing it, about the only thing I'd do extra, is get out the
multimeter and attempt to find the voltages I thought, might be
causing the problem. Before looking for another motherboard.

I have done more extensive analysis of hardware, in the lab I used
to work in. There, I could connect a couple logic analysers to
a defective piece of electronics, and trace firmware execution.
I don't have anything like that here, at home.

Paul

species8350 · Apr 14, 2013

I don't have any testing equipment, or access to another RAM module.

We both think its a motherboard problem. So I'll start looking.

Can you suggest a cheap motherboard that will be compatible (BTX case, LGA 775, I believe chipset G965, processor Pentium D 915, 2.8 GHz). Refurbishesd o secondhand would be fine, its quite an old tower.

Thank you for all the help that you have provided and for staying with this thread. Much appreciated.

Best wishes.

A

Paul · Apr 14, 2013

The BTX is going to limit your options.

Just plug your motherboard details into Ebay and
see what's there. That way, you get something
mechanically that's going to fit. See if there
is a number on the surface of the motherboard,
like WG864 or the like. Some of the numbers on the
motherboard, appear to refer to standards of some sort,
and are not actually model info.

When I needed an LGA775 board for my current
build, LGA775 was already old as a standard.
And I paid more for the board than I should have.
My choices were rather limited. And the boards
were ATX form factor.

In your case, there won't be a lot of BTX boards,
that might align with your cooling solution and
so on. If you get the exact same board, that
removes some of the headaches (header for front panel
wiring). If you substitute something else, there
are going to be more details to worry about.

I have a couple computer recyclers in town, and
I suppose I could walk my motherboard into a shop
like that, and see what they've got that is similar.

The only thing you have to be careful of, is certain
Dell models had bad capacitor problems. And for those,
if you saw a motherboard for sale, it would also have the
bad caps. So if you were thinking of substituting another
Dell model of motherboard, check the history of that
computer, to see if there were problems. If the motherboard
advertising said "re-capped", it might be OK to buy. But
some of those products, the fallout rate might be 90%.

"Suit Over Faulty Computers Highlights Dell’s Decline"
http://www.nytimes.com/2010/06/29/technology/29dell.html?_r=0

So if you were thinking of taking a board from a GX620,
and stuffing it into your E520, that might be risky from
a capacitor point of view. You need to research, whether
a particular model of motherboard is one of the risky ones.

I have seen whole, used Dell computers (perhaps off lease), with
an advertising banner of "limited time only", and that
puts a smile on my face. As it implies it'll work for
a limited time, if I buy it and take it home.

In terms of the bad capacitor problem here, I've lost a couple
power supplies to the bad caps. The capacitors had the
orange-brown colored rust stains on top, and a "sizzling"
sound came from the power supply. So far, no motherboards
have blown on me, and that's a surprise.

Paul