Stephen said:
I addressed this question to the "general" group and someone suggested that I
turn the question to the hardware group.
I recently installed new memory in an old (2000) HP. It had Win ME on it
when I bought it and I changed to XP Professional, but did not have efficient
RAM. I put two new 256mb memories, specified for this HP model, but I am
getting an error message saying that "We apologize for the inconvenience, but
Windows did not start successfully. A recent hardware or software change
might have caused this." Remedies suggested on the error page include
starting in safe mode, safe mode with networking, etc., and last known good
configuration and finally start Windows normally. None of those work. I
have also gone into the BOIS setup via F1, but do not know my way around
those screens. The memory does show up successfully in the BIOS setup
screens. Any ideas of how to get XP to boot?
Thanks much.
If the computer has a floppy drive or a CDROM drive you can boot from,
you can use this program.
http://www.memtest.org/
For the floppy version, you'll need to go to a working computer, and use
the downloaded program to format a blank floppy. The floppy will not
contain a file system, when the program is finished preparing the floppy.
The memtest86+ floppy cannot be viewed in Explorer. Now, plug the
prepared floppy into the computer with the new RAM, and allow the
memtest program on the floppy to boot the computer.
An alternative is to prepare a CDROM version of memtest. Download
the ISO9660 type file from memtest.org . Use a CDROM burning program
such as Nero, or any other program which recognizes how to transfer
an ISO9660 to blank media. ISO9660 is unique, in that it contains
information in its format, for making bootable media. The program
used to burn a CD, has to know how to handle an ISO9660 file, to
get proper results. If you burn the CDROM correctly, and look
at the CDROM in Explorer, chances are you'll see a single file
called "README.txt" on the CD. As with the floppy disk case,
once you've prepared the new CD, you can use it to boot the
computer that has the new RAM.
This is what the screen looks like when memtest86+ starts.
http://www.memtest.org/pics/i875-big.gif
The two fields "Cached" and "RsvdMem", when added together, should
more or less total up to the amount of RAM you have installed in the
computer. For example, in that screenshot, the total is close to
256MB of RAM. On computers with built-in graphics in the chipset,
part of the Reserved memory could be used for the display frame
buffer.
The memory test will run forever, if you leave it. You want to run
one or two complete passes, to have some confidence the memory is
working. Usually this takes an hour - but if you see errors printed
on the lower half of the screen, then you're done with memtest
and know the memory has a problem.
To quit the program, press the <esc> key, and the computer will reboot.
Now, the reason I'm familiar with this process, is I was using it
today. I'll relate the story, so you can see what the whole procedure
might look like, and maybe it will be relevant to your symptoms.
A couple weeks ago, I turned on the computer, and it blue screened
during boot, claiming a hardware problem. At first, I thought perhaps
my hard drive was bad. I have two boot drives, and tried the alternate
boot drive, but it gave an error message as well. Next, I tried
removing a stick of RAM, and the computer could boot. I was relieved
that the hard drive was OK, and now I could test the memory.
Depending on the order that the two memory sticks were placed
in the computer, I couldn't even get memtest86+ to boot or
get the computer to pass the BIOS startup tests. My
motherboard has four memory slots, and the configuration I
ended up using, was good_stick, bad_stick, good_stick, so
that the bad one was sandwiched in the middle of the good
ones (I had another spare stick available for use). For some
reason, the BIOS could not detect that the bad stick was bad,
when the bad stick wasn't right next to the CPU. By trying
the sticks in a different order, you may be able to get the
BIOS to work, and get memtest86+ booted.
So after figuring out how to position the RAM so I could
test it, I ran memtest86+. Within the first minute, I got
errors. The same byte position in the stick was affected, on
each errored word. (The DIMM is 8 bytes wide, and the same byte was bad in
each case. Memtest86+ works on 32 bits at a time, and the memory
is 64 bits wide, so you have to figure out from the memory address,
as to whether the upper or lower 32 bits are affected.) One of my
memory chips can no longer be written, and uninitialized garbage
comes back from the bad chip.
So that is an example of figuring out what is wrong. On the
one hand, by using just the bad stick of RAM in the computer,
I couldn't do anything. The BIOS wouldn't POST with just
the bad stick present. By putting a good stick nearest the
CPU, and the bad stick in a slot further away, I was able
to start the computer, and boot memtest86+.
I purchased new memory a couple days ago, and it arrived today.
The first thing to do, was to boot with memtest86+, with the
new RAM installed. This is important, because it reduces the
risk that Windows will get corrupted on the first boot. (Sometimes
new RAM is bad when you get it.)
In the worst case, with new RAM, some new RAM has the power rails
shorted together, and the motherboard gets burned and damaged. But
more likely, your memory could have some errors in it, either
transient errors or errors at fixed locations. As long as the
memory is good enough to allow the BIOS to start, then you can
use memtest86+ for testing.
Once memtest86+ is done, there are a couple choices. In the past, I've
booted a Linux LiveCD, like Knoppix (knopper.net) or Ubuntu
(ubuntu.com). Those are Linux CDs that don't have to be installed
to a hard drive to work. I can run the Prime95 program, from
mersenne.org, and use its "torture test", to further test the
memory. If Prime95 will run for four hours, with not a single
error detected, then I consider it safe to boot Windows from
my hard drive.
In this case, I didn't bother with the Linux test, and after
my memtest86+ testing was finished, I booted into Windows. I'm
running the Windows version of Prime95 in the background, while
typing this, and have Prime95 set to test most but not all of
my new memory. I reduced the allocation in the Options of the
Prime95 program, so there would be room for my news reader. So
far, Prime95 has run for two hours and no errors yet.
As one other poster suggested, you should compare the old and
known working memory, to the new memory. Compare the test
results with the old memory, to the test results with the
new memory.
When adding RAM to a computer, you should be absolutely
sure the power is off. My motherboard has a green LED that
lights if +5VSB is present. If the green LED is lit, it isn't
safe for me to add or remove sticks of RAM. Switching off
the switch at the back of the computer, will disable +5VSB.
But another alternative, is to unplug the computer before
changing the RAM. You can cause pretty expensive damage
if you don't follow this rule.
When installing the RAM, you should take antistatic precautions.
Bring the new RAM and packaging to the same electrical potential
as the chassis of the computer. While holding some bright
shiny metal on the chassis, use your other hand to install the
RAM. You can use that procedure if you don't own a wrist strap,
as normally you'd use a wrist strap to bring your body to the
same potential as the computer chassis.
In terms of memory limits, Win98 was limited to about 512MB
(without some fiddling around). WinXP should not be limited in
that way, so you could use more RAM without a problem.
So give memtest86+ a try, and see how much memory it reports,
what timing values it shows and so on.
Hope that helps,
Paul
