Windows 2000 boot up freezes after RAM upgrade

[Guest]

I just upgraded my computer's RAM from 512 MB (2 sticks of 256 MB) to
1.5 GB (3 sticks of 512 MB). The motherboard has 3 memory slots, and
supports 1.5 GB of RAM. All 3 sticks are the same brand (Crucial) and
speed.

Now every time I try to boot up my computer, it goes past the Windows
2000 logo, and right before it shows the blue background for the logon
screen, it hangs. Every time.

Any ideas on what's going on here?

 

[Paul]

I just upgraded my computer's RAM from 512 MB (2 sticks of 256 MB) to
1.5 GB (3 sticks of 512 MB). The motherboard has 3 memory slots, and
supports 1.5 GB of RAM. All 3 sticks are the same brand (Crucial) and
speed.

Now every time I try to boot up my computer, it goes past the Windows
2000 logo, and right before it shows the blue background for the logon
screen, it hangs. Every time.

Any ideas on what's going on here?

Did you test with memtest86+ from memtest.org first ? You're
supposed to do that first, before you boot your Windows OS
again. If you boot a system, with bad RAM present, it can
corrupt your install.

Memtest86+ is available for floppy, as well as CD. The program
prepares a self-booting piece of media - there is no OS present
when it runs.

The program will run forever if you let it. You want to let
it run long enough, to complete about two passes. Usually
a couple hours is enough.

During those couple hours, if there are any memory errors,
you have a problem.

You did not indicate your computer make and model or
motherboard brand and model number. Without that
information I cannot give advice on the best
approach.

If you want to start easy, you should use just one of
the new sticks of memory. You can test the sticks,
one at a time. And 512MB may be enough to boot an OS
as well, for a quick test. A single t12MB stick, will
test faster in memtest86+, than would three sticks.

Memtest86+ will test most of the memory. There is
about 1MB or less of memory, which is reserved by
the BIOS, and that bit cannot be tested. But
otherwise, the memtest86+ program even lifts itself
out of the way, and tests underneath where the code is
stored. So it is relatively thorough, in terms of visiting
as many locations as possible.

If you want more help, give more details about the
computer hardware involved. And also, the part number
of the Crucial modules you decided to buy.

Paul

 

[Jose]

I just upgraded my computer's RAM from 512 MB (2 sticks of 256 MB) to
1.5 GB (3 sticks of 512 MB).  The motherboard has 3 memory slots, and
supports 1.5 GB of RAM.  All 3 sticks are the same brand (Crucial) and
speed.

Now every time I try to boot up my computer, it goes past the Windows
2000 logo, and right before it shows the blue background for the logon
screen, it hangs.  Every time.

Any ideas on what's going on here?

No idea what's going on, but you can narrow it down maybe.

If you put your old sticks back does it work okay?

Just install 1 new stick at a time and see how it goes - does it look
right in the POST screen?

If stick 1 does not work, replace stick 1 until you can get a good
boot with just 1 stick.

Install a second stick. Try again.

Install a third stick. Try again.

Any unusual beeping after the power is on? One short beep is
generally good, any other beep pattern usually indicates a problem
that can be figured out by looking at the documentation for your
motherboard and BIOS.

If you need help with that, supply some details so we don't have to
guess about anything and ask a bunch of questions back and forth.

After looking at your documentation, is there anything you need to do
in the BIOS for this new RAM?

 

[Guest]

Did you test with memtest86+ from memtest.org first ? You're
supposed to do that first, before you boot your Windows OS
again. If you boot a system, with bad RAM present, it can
corrupt your install.

Oops, I guess I should have done that. I was planning on running
memtest86+ later. I guess I didn't feel any urgency to do that, since
I've bought used RAM (and I got these memory modules off ebay) several
times before and never had a problem.

I have my old memory back in my computer now, and I have successfully
loaded the recovery console (which was installed to the hard drive).
I'm running chkdsk on the hard drive partitions now.

Memtest86+ is available for floppy, as well as CD. The program
prepares a self-booting piece of media - there is no OS present
when it runs.

I have a floppy with version 1.65 of memtest86+. Probably at least a
few years old. But should still be OK.

The program will run forever if you let it. You want to let
it run long enough, to complete about two passes. Usually
a couple hours is enough.

During those couple hours, if there are any memory errors,
you have a problem.

Yeah, I just ran it on the new memory that I got and the first pass
failed. I think it was test #5. I checked up on my computer
periodically while it was running, and came back to see tons of errors
on test 5, and it was around the 300MB mark. So that would indicate
the memory module in slot 1 was bad, right? (each memory module is
512MB) I kept watching, the errors kept piling up, even when it went
past 512MB, so I guess that means the memory module in slot 2 was
bad? And the errors never stopped when it went past 1024MB, so I
guess that means the memory module in slot 3 was also bad? The errors
seemed to stop when test 6 started.

Then I put my 2 old sticks of 256MB ram back into the computer, ran
memtest86+ again, and the first pass was successful.

You did not indicate your computer make and model or
motherboard brand and model number. Without that
information I cannot give advice on the best
approach

It's an old computer, an Abit KT7-RAID motherboard, which has 3 memory
slots, and the manual says it supports up to 1.5 GB of RAM. I'm
running Windows 2000, and see no reason why I can't upgrade to 1.5GB.

If you want to start easy, you should use just one of
the new sticks of memory. You can test the sticks,
one at a time. And 512MB may be enough to boot an OS
as well, for a quick test. A single t12MB stick, will
test faster in memtest86+, than would three sticks.

Do you think the fact that one of the memtest86+ tests seemed to fail
across the whole 1.5GB range (well, at least starting from 300MB when
I came back to check on the computer) indicates that all 3 modules are
bad? Seems unlikely, but I guess that's possible.

The person I got these from on ebay said I can return them within 7
days, so I will probably do that, and try to get 3 modules from
someone else.

Memtest86+ will test most of the memory. There is
about 1MB or less of memory, which is reserved by
the BIOS, and that bit cannot be tested. But
otherwise, the memtest86+ program even lifts itself
out of the way, and tests underneath where the code is
stored. So it is relatively thorough, in terms of visiting
as many locations as possible.

If you want more help, give more details about the
computer hardware involved. And also, the part number
of the Crucial modules you decided to buy.

The new modules are Crucial CT64M64S4D7E (512MB SDRAM PC133 CL2).

My old modules are Crucial CT32M64S4D7E (256MB SDRAM PC133 CL2).

 

[Paul]

Oops, I guess I should have done that. I was planning on running
memtest86+ later. I guess I didn't feel any urgency to do that, since
I've bought used RAM (and I got these memory modules off ebay) several
times before and never had a problem.

I have my old memory back in my computer now, and I have successfully
loaded the recovery console (which was installed to the hard drive).
I'm running chkdsk on the hard drive partitions now.


I have a floppy with version 1.65 of memtest86+. Probably at least a
few years old. But should still be OK.


Yeah, I just ran it on the new memory that I got and the first pass
failed. I think it was test #5. I checked up on my computer
periodically while it was running, and came back to see tons of errors
on test 5, and it was around the 300MB mark. So that would indicate
the memory module in slot 1 was bad, right? (each memory module is
512MB) I kept watching, the errors kept piling up, even when it went
past 512MB, so I guess that means the memory module in slot 2 was
bad? And the errors never stopped when it went past 1024MB, so I
guess that means the memory module in slot 3 was also bad? The errors
seemed to stop when test 6 started.

Then I put my 2 old sticks of 256MB ram back into the computer, ran
memtest86+ again, and the first pass was successful.


It's an old computer, an Abit KT7-RAID motherboard, which has 3 memory
slots, and the manual says it supports up to 1.5 GB of RAM. I'm
running Windows 2000, and see no reason why I can't upgrade to 1.5GB.


Do you think the fact that one of the memtest86+ tests seemed to fail
across the whole 1.5GB range (well, at least starting from 300MB when
I came back to check on the computer) indicates that all 3 modules are
bad? Seems unlikely, but I guess that's possible.

The person I got these from on ebay said I can return them within 7
days, so I will probably do that, and try to get 3 modules from
someone else.


The new modules are Crucial CT64M64S4D7E (512MB SDRAM PC133 CL2).

My old modules are Crucial CT32M64S4D7E (256MB SDRAM PC133 CL2).

I would test the sticks individually. Less bus loading, so that
the usage of one stick should remove loading as a consideration.

By testing the sticks one at a time, you may also see a pattern
emerging. Such as two new sticks passing, and one failing.

Say two sticks passed individually. Then I'd install them
together, in slot 1 and slot 3, and do another couple
passes.

If that passes, then I'd add one old stick. This test case,
with the three sticks, is to see if bus loading is an issue
for your motherboard or not.

If you notice (as I have), that Test 5 "does all the testing",
you can use the controls in memtest86+ to advance to Test 5,
and skip the other tests. By doing that, you may be able to
run more test cases in a shorter time.

Eventually, you'll find a "maximum" configuration that
is working for you. And a number of other interesting cases
that don't work. You might see a pattern to it.

Once your characterization is complete, then you can make up
your mind as to whether it is worth returning all of them.

Paul

 

[Guest]

I would test the sticks individually. Less bus loading, so that
the usage of one stick should remove loading as a consideration.

By testing the sticks one at a time, you may also see a pattern
emerging. Such as two new sticks passing, and one failing.

Say two sticks passed individually. Then I'd install them
together, in slot 1 and slot 3, and do another couple
passes.

If that passes, then I'd add one old stick. This test case,
with the three sticks, is to see if bus loading is an issue
for your motherboard or not.

If you notice (as I have), that Test 5 "does all the testing",
you can use the controls in memtest86+ to advance to Test 5,
and skip the other tests. By doing that, you may be able to
run more test cases in a shorter time.

Eventually, you'll find a "maximum" configuration that
is working for you. And a number of other interesting cases
that don't work. You might see a pattern to it.

Once your characterization is complete, then you can make up
your mind as to whether it is worth returning all of them.

Thanks for the help.

I tested each stick individually, and they all passed!

Then I tested a stick in slot 1 and a stick in slot 3, and that
passed. I'm testing a stick in slot 1 and a stick in slot 2 right now
(I'm guessing that will pass).

So the system can handle 1GB of RAM, but not 1.5GB. The manual for my
motherboard specifically says that it supports a 512MB memory module
in all 3 slots, so I wonder why it doesn't work?

 

[Sid Elbow]

So the system can handle 1GB of RAM, but not 1.5GB. The manual for my
motherboard specifically says that it supports a 512MB memory module
in all 3 slots, so I wonder why it doesn't work?

I've seen exactly the same Windows booting problem you describe when
trying to upgrade the memory on a laptop. Laptops are notoriously fussy
about the exact ram that they will accept .... nominally the same
specification of ram from different manufacturers may or may not work.

Desktop machines are usually reckoned to be less sensitive in that
respect - perhaps because the electrical design of the motherboard
doesn't have to skimp on power in all the circuits the way a laptop
does. However, it's probably still possible for something similar to
happen in some instances on a desktop and it might depend on just how
many modules you use. If you chose another manufacturer of the same ram
modules it might perhaps take all three (or just one ).

 

[Paul]

Thanks for the help.

I tested each stick individually, and they all passed!

Then I tested a stick in slot 1 and a stick in slot 3, and that
passed. I'm testing a stick in slot 1 and a stick in slot 2 right now
(I'm guessing that will pass).

So the system can handle 1GB of RAM, but not 1.5GB. The manual for my
motherboard specifically says that it supports a 512MB memory module
in all 3 slots, so I wonder why it doesn't work?

I don't know how much control you have, in the BIOS. Since it is a
KT7, perhaps the FSB is limited to FSB200 ? The memory may run at
PC100 or PC133 (depending on whether "host clock + PCI" is selected
for the RAM).

In the memory timing parameters, there are Tcas, Trcd and so on.
You can try setting the memory timings manually, such as bumping
up the numerical value of Trcd, one above its normal value. That
would "relax" the timing.

To review how the board is currently set up, drop back to one
stick. Install the single 512MB stick, in the slot furthest from
the processor. Boot the OS, and see if you have stability.

Download CPUZ from cpuid.com

http://www.cpuid.com/cpuz.php

You can use that, to record how the BIOS is setting things up.
Record the values, so if necessary, you can enter them manually
in the BIOS.

Tcas has a relatively large effect. Trcd a smaller effect (it
sometimes affects the ability to overclock a RAM). The thing is,
depending on the RAM, it might be relatively sluggish CAS3 stuff.
In which case, there might not be a lot of room to adjust it higher.
I.e. The BIOS may not accept CAS4 as a setting. If it did, then
perhaps more bus loading could be handled.

I don't know where I'd find a manual for that board now, as
Abit/UAbit has stopped making motherboards, and their documentation
is rather limited.

Paul

 

[Guest]

I don't know how much control you have, in the BIOS. Since it is a
KT7, perhaps the FSB is limited to FSB200 ?

The CPU FSB/PCI Clock is set to 100/33MHz.

The memory may run at
PC100 or PC133 (depending on whether "host clock + PCI" is selected
for the RAM).

DRAM Clock is set to HCLK+PCKCLK

There's also:
Enhance Chip Performance: enabled
Force 4-Way Interleave: enabled
Enable Dram 4K-Page Mode: enabled

In the memory timing parameters, there are Tcas, Trcd and so on.
You can try setting the memory timings manually, such as bumping
up the numerical value of Trcd, one above its normal value. That
would "relax" the timing.

The advanced section has the following:

Bank 0/1 DRAM Timing: Turbo (that's the fastest, other options are
SDRAM 8/10ns, Normal, Medium, and Fast)
Bank 2/3 DRAM Timing: Turbo
Bank 4/5 DRAM Timing: Turbo
DRAM Bank Interleave: 4-Way (other options are 2-Way, Disabled)
Delay DRAM Read Latch: AUTO (other options are NO Delay, 0.5ns, 1.0ns,
1.5ns)
MD Driving Strength: Hi (other option is Lo)
SDRAM Cycle Length: 2 (other option is 3)
Memory Hole: Disabled (other option is 15M-16M)

To review how the board is currently set up, drop back to one
stick. Install the single 512MB stick, in the slot furthest from
the processor. Boot the OS, and see if you have stability.

Download CPUZ from cpuid.com

   http://www.cpuid.com/cpuz.php

You can use that, to record how the BIOS is setting things up.
Record the values, so if necessary, you can enter them manually
in the BIOS.

Tcas has a relatively large effect. Trcd a smaller effect (it
sometimes affects the ability to overclock a RAM). The thing is,
depending on the RAM, it might be relatively sluggish CAS3 stuff.
In which case, there might not be a lot of room to adjust it higher.
I.e. The BIOS may not accept CAS4 as a setting. If it did, then
perhaps more bus loading could be handled.

Well my RAM is CAS 2, so I can try changing that SDRAM Cycle Length to
3.

I don't know where I'd find a manual for that board now, as
Abit/UAbit has stopped making motherboards, and their documentation
is rather limited.

It can be found here: http://safemanuals.com/user-guide-instructions-owner-manual/ABIT/KT7-RAID-_E

 

[Paul]

The CPU FSB/PCI Clock is set to 100/33MHz.


DRAM Clock is set to HCLK+PCKCLK

There's also:
Enhance Chip Performance: enabled
Force 4-Way Interleave: enabled
Enable Dram 4K-Page Mode: enabled


The advanced section has the following:

Bank 0/1 DRAM Timing: Turbo (that's the fastest, other options are
SDRAM 8/10ns, Normal, Medium, and Fast)
Bank 2/3 DRAM Timing: Turbo
Bank 4/5 DRAM Timing: Turbo
DRAM Bank Interleave: 4-Way (other options are 2-Way, Disabled)
Delay DRAM Read Latch: AUTO (other options are NO Delay, 0.5ns, 1.0ns,
1.5ns)
MD Driving Strength: Hi (other option is Lo)
SDRAM Cycle Length: 2 (other option is 3)
Memory Hole: Disabled (other option is 15M-16M)



Well my RAM is CAS 2, so I can try changing that SDRAM Cycle Length to
3.


It can be found here: http://safemanuals.com/user-guide-instructions-owner-manual/ABIT/KT7-RAID-_E

I don't know what a number of those settings do, so I'm
at a disadvantage right away.

You could try setting the DRAM clock to HCLK, basically
slowing it down to PC100. Then run the RAM test with
three sticks. That would at least prove it can work with
the three sticks present.

Or, leave it at HCLK+PCKCLK and try CAS3 for a test run,
and see how far it gets.

I don't understand the implications of "Turbo" in this case.
Many memory settings, apply to all slots at the same time.
While in principle, a chip could have specific behaviors
per slot, (like maybe Tsetup/Thold compensation), I don't
really see anything "speed related" that could be associated
that way. For example, if you apply CAS3, it applies to all
slots at the same time.

I know there are some more modern boards, where a "Turbo" setting,
basically just forces the CAS to CAS2. But the information you
present above, doesn't look like that.

You'd need to find a review article for the KT7-RAID, where
they go over the settings.

At least your testing is getting you somewhere. But tracking
down what all those settings do, is a much more challenging
project.

Good luck,
Paul

 

[Paul]

The CPU FSB/PCI Clock is set to 100/33MHz.


DRAM Clock is set to HCLK+PCKCLK

There's also:
Enhance Chip Performance: enabled
Force 4-Way Interleave: enabled
Enable Dram 4K-Page Mode: enabled


The advanced section has the following:

Bank 0/1 DRAM Timing: Turbo (that's the fastest, other options are
SDRAM 8/10ns, Normal, Medium, and Fast)
Bank 2/3 DRAM Timing: Turbo
Bank 4/5 DRAM Timing: Turbo
DRAM Bank Interleave: 4-Way (other options are 2-Way, Disabled)
Delay DRAM Read Latch: AUTO (other options are NO Delay, 0.5ns, 1.0ns,
1.5ns)
MD Driving Strength: Hi (other option is Lo)
SDRAM Cycle Length: 2 (other option is 3)
Memory Hole: Disabled (other option is 15M-16M)



Well my RAM is CAS 2, so I can try changing that SDRAM Cycle Length to
3.


It can be found here: http://safemanuals.com/user-guide-instructions-owner-manual/ABIT/KT7-RAID-_E

[Re-sent via AIOE. Articles from my other news server, don't get through
to people using the MS server.]

I don't know what a number of those settings do, so I'm
at a disadvantage right away.

You could try setting the DRAM clock to HCLK, basically
slowing it down to PC100. Then run the RAM test with
three sticks. That would at least prove it can work with
the three sticks present.

Or, leave it at HCLK+PCKCLK and try CAS3 for a test run,
and see how far it gets.

I don't understand the implications of "Turbo" in this case.
Many memory settings, apply to all slots at the same time.
While in principle, a chip could have specific behaviors
per slot, (like maybe Tsetup/Thold compensation), I don't
really see anything "speed related" that could be associated
that way. For example, if you apply CAS3, it applies to all
slots at the same time.

I know there are some more modern boards, where a "Turbo" setting,
basically just forces the CAS to CAS2. But the information you
present above, doesn't look like that.

You'd need to find a review article for the KT7-RAID, where
they go over the settings.

At least your testing is getting you somewhere. But tracking
down what all those settings do, is a much more challenging
project.

Good luck,
Paul

 

[Guest]

I don't know what a number of those settings do, so I'm
at a disadvantage right away.

You could try setting the DRAM clock to HCLK, basically
slowing it down to PC100. Then run the RAM test with
three sticks. That would at least prove it can work with
the three sticks present.

Or, leave it at HCLK+PCKCLK and try CAS3 for a test run,
and see how far it gets.

I don't understand the implications of "Turbo" in this case.
Many memory settings, apply to all slots at the same time.
While in principle, a chip could have specific behaviors
per slot, (like maybe Tsetup/Thold compensation), I don't
really see anything "speed related" that could be associated
that way. For example, if you apply CAS3, it applies to all
slots at the same time.

I know there are some more modern boards, where a "Turbo" setting,
basically just forces the CAS to CAS2. But the information you
present above, doesn't look like that.

You'd need to find a review article for the KT7-RAID, where
they go over the settings.

At least your testing is getting you somewhere. But tracking
down what all those settings do, is a much more challenging
project.

Thanks for the help. Changing CAS to 3 did not help, that still
failed. So I changed it back to CAS2, and then changed the DRAM Clock
to Host CLK. Now it passes!

Any idea how much of a performance hit it will be to run RAM at PC100
instead of PC133?

 

[Paul]

Thanks for the help. Changing CAS to 3 did not help, that still
failed. So I changed it back to CAS2, and then changed the DRAM Clock
to Host CLK. Now it passes!

Any idea how much of a performance hit it will be to run RAM at PC100
instead of PC133?

You can test that for yourself. I use SuperPI and have it
calculate 1 million digits of PI. The benchmark returns
an execution time. (The number of digits, should be selected
so that the amount of data is larger than the processor cache,
and 1 million digits is 8MB or so. On some of the newer processors,
it helps to benchmark a larger number of digits.) Note - about
half the time, when I try to reach this site, there is a server
problem. So if it fails to appear in your browser, try again
later.

Main page for SuperPI download
http://www.xtremesystems.com/pi/

Download link
http://www.xtremesystems.com/pi/super_pi_mod-1.5.zip

What you'd do, is try two sticks at HCLK+PCKCLK, run the
benchmark. Then try three sticks at HCLK, and compare the
two results. That gives a percentage impact for the memory
change. Since in both cases, there is plenty of room to
run the benchmark, the amount of memory should make no
difference to the results.

SuperPI takes some of the guesswork out, by being an
actual benchmark with a timer. It should give you a
good idea, what the application impact would be.

Some applications are more memory dependent than others.
For example, WinRAR depends more on memory, than your
average program. So if you wanted to see the maximum
amount of impact, you'd try compressing or decompressing
a WinRAR archive.

But I like SuperPI for benching, and have used it a lot
for tuning. I believe the original source code for SuperPI
was lost some time ago. The current version is maintained,
by editing the binary code in the file. So features were
added, without actually having the base code. (A guy in
Japan wrote the original code.) According to this, it
was written in 1995. Some of the timer features (to prevent
cheating and improve accuracy), were added by binary editing
of the executable.

http://en.wikipedia.org/wiki/Superpi

Paul

 

[Buffalo]

The CPU FSB/PCI Clock is set to 100/33MHz.

Try runing both at the same speed. Some boards have problems with the ram
running at a different speed than the FSB.
Buffalo