Memtest86 errors

  • Thread starter Thread starter Bob Davis
  • Start date Start date
B

Bob Davis

System: Gigabyte 8KNXP, P4 2.8C, 2gb RAM (Kingston DDR400 KVR400X64C3A/512)
, not overclocked.

When running Memtest86 v3 I get random errors at the very end of test #5.
The "good" is always one of the following: ff7fffff, ffefffff, fff7ffff, or
fffeffff--and the "bad" is always ffffffff.

A second GB of RAM was installed two days ago, and the errors started at
that time. Previous tests with 1gb did not manifest errors. The address
location of the errors is random throughout the 2gb, so it apparently isn't
in the new modules. Error count can be as low as two and as high as 20, and
they always occur just as Test #5 is completing and transitioning to Test
#6. I reversed the two new modules without effect, but since I use a Zalman
HSF the Bank 0, Slot 1 module appears not to be removable without moving or
removing the HSF, I haven't tried the new memeory alone.

Anyone have an idea what's going on here? The computer runs fine and I've
had no trouble yet in real-world experience.
 
I have used MemTest86 myself to discover errors in RAM. However, the reasons
for performing the test have always been because of problem troubleshooting.
Indeed, that's how I discovered a bad stick of Crucial DDR 2100, once.
My best guess regarding your issue: I think your RAM is fine (especially if
you are not having problems). I'd be willing to bet that MemTest86 is having
problems dealing with all of the RAM that is currently in your system.


Hope this helps,

D.
 
I managed to shoehorn the stick in bank0/slot1 past the HSF and installed
the new memory alone in B0/S1 and B1/S3 (remember, I have six slots). The
tests showed no errors, so the error on Test 5 occurs only when all four
sticks are installed. I am now satisfied that the physical memory is solid,
but here's an interesting blurb from the Memtest86 site:

"In the vast majority of cases errors reported by the test are valid. There
are some systems that cause Memtest86 to be confused about the size of
memory and it will try to test non-existent memory. This will cause a large
number of consecutive addresses to be reported as bad and generally there
will be many bits in error. If you have a relatively small number of failing
addresses and only one or two bits in error you can be certain that the
errors are valid."

This sounds exactly what's happening here: When it hits 98% on Test 5 it
spits out from two to 20 errors very quickly, all with multiple bits in
error, usually 10k-20k but more rarely 100. I've never seen any bit count
in between. Also, when it hits 98% it zips to 100% much faster than the
previous progression for Test 5.
 
Sometimes the SPD module on the DIMM misreports timings or the BIOS not
reading the SPD correctly or the BIOS is set to some default "aggresive"
memory timing, which can lead to errors. Look at the sticker on the module
and then look at the timings in your BIOS. If they are different, manually
set them to what the sticker says and then run the test again.

....and then again, its just one free program telling you the memory is
bad...might be able to ignore it...
 
This is my experience and is pre DDR . . . hope it helps.

If you are not exceeding the Spec of each module (bios setting) and each
module functions independently, then the Memory Bus is having a hard time
handling both modules. It may be possible to improve the situation by
using two identical modules if that is not already the case. If there are
more than 2 memory slots, try the last 2 or another pattern. This is
likely due to some impedance mismatch or drive problem that brings the
noise margins down to iffy levels, and/or causes reflections on the bus.

Also try Gold memory tester. The only one I have found as an alternate to
Memtest86.
 
Salvatore said:
This is my experience and is pre DDR . . . hope it helps.

If you are not exceeding the Spec of each module (bios setting) and each
module functions independently, then the Memory Bus is having a hard time
handling both modules. It may be possible to improve the situation by
using two identical modules if that is not already the case. If there are
more than 2 memory slots, try the last 2 or another pattern. This is
likely due to some impedance mismatch or drive problem that brings the
noise margins down to iffy levels, and/or causes reflections on the bus.

Also try Gold memory tester. The only one I have found as an alternate to
Memtest86.
 
If you are not exceeding the Spec of each module (bios setting) and each
module functions independently, then the Memory Bus is having a hard time
handling both modules. It may be possible to improve the situation by
using two identical modules if that is not already the case.


The original two (512 x 2) modules manifested no errors on Test 5, nor did
the new pair in the same slot. I also got a clean result running the new
modules alone in Slots 2 and 5 (I have six total slots).
Only when I place all four in Slots 1,2,4, and 5--with matched pairs in 1 &
4 and 2 & 5--do I get the errors on most passes, although several have
completed without errors. The addresses are random, occur only in the last
98% of Test 5, and contain multiple error bits (usually 10k-10m). This
falls in line with Memtest's note that the program may be getting confused
about the amount of memory:

"There are some systems that cause Memtest86 to be confused about the size
of memory and it will try to test non-existent memory. This will cause a
large number of consecutive addresses to be reported as bad and generally
there will be many bits in error. If you have a relatively small number of
failing addresses and only one or two bits in error you can be certain that
the errors are valid."

I also got this email from Kingston today:

"I recommend changing the timing of the Cas to Ras Delay from 8 to 6 which
should be the
default of your motherboard for DDR400 memory. Also please keep in mind
that sometimes these 3rd party memory testers don't always report proper
readouts. If you computer is running fine as you indicated then I wouldn't
worry about it. If the memory was truly failing as mem test shows then your
computer would not be running properly."

My RAS-to-CAS delay setting will go no higher than 4, and it is set to 3. I
will be calling him tomorrow.
If there are more than 2 memory slots, try the last 2 or another pattern. This is
likely due to some impedance mismatch or drive problem that brings the
noise margins down to iffy levels, and/or causes reflections on the bus.

I'll try moving the second pair to slots 3 and 6, and will post back later.
Also try Gold memory tester. The only one I have found as an alternate to

Do you mean Memonster Gold? I can find no references to "Gold Memory
Tester" with a Google search.
 
The original two (512 x 2) modules manifested no errors on Test 5, nor did
the new pair in the same slot. I also got a clean result running the new
modules alone in Slots 2 and 5 (I have six total slots).
Only when I place all four in Slots 1,2,4, and 5--with matched pairs in 1 &
4 and 2 & 5--do I get the errors on most passes, although several have
completed without errors. The addresses are random, occur only in the last
98% of Test 5, and contain multiple error bits (usually 10k-10m). This
falls in line with Memtest's note that the program may be getting confused
about the amount of memory:

"There are some systems that cause Memtest86 to be confused about the size
of memory and it will try to test non-existent memory. This will cause a
large number of consecutive addresses to be reported as bad and generally
there will be many bits in error. If you have a relatively small number of
failing addresses and only one or two bits in error you can be certain that
the errors are valid."
That is why I suggested the other memory test.
I also got this email from Kingston today:

"I recommend changing the timing of the Cas to Ras Delay from 8 to 6 which
should be the
default of your motherboard for DDR400 memory. Also please keep in mind
that sometimes these 3rd party memory testers don't always report proper
readouts. If you computer is running fine as you indicated then I wouldn't
worry about it. If the memory was truly failing as mem test shows then your
computer would not be running properly."

My RAS-to-CAS delay setting will go no higher than 4, and it is set to 3. I
will be calling him tomorrow.


I'll try moving the second pair to slots 3 and 6, and will post back later.


Do you mean Memonster Gold? I can find no references to "Gold Memory
Tester" with a Google search.
GoldMemory v5.07
Copyright 1993,2003 Michal Tulacek
 
I have NO experience with your motherboard or MoBos with 6 Memory slots,
so I cannot accurately comment on their architecture. Again, I have no
hands on with DDR. Waiting until all the intel CPU/system bus speed
escalation is finished before I build another system.


That said, I would look in the bios to see if there is a setting for drive
current for the Memory bus just to see if this feature is present. I have
seen one or 2 with it but cannot recall if it was of recent vintage.

There is perhaps some validity to my feeling that impedance mismatch
exists. This *"might"* not occur with 4 identical modules if that is so.
If memories had cables, this is the kind of problem that might change
symptoms with different cable lengths.

If it is drive power related, you might be able to see the voltage (DVM)
at the memory slot decrease with all 4 modules present. If you can borrow
2 more, 6 might make things much worse. Not an easy one to sort out.

Try this to see if it is all MemTest86 generated

http://www.goldmemory.cz/download.htm
 
This is what investigating these "spooky" problems is like.

No one thing is assumed good. As you note, you may have discovered the
MoBo having marginal (or deteriorating) support for multiple memory banks.
OR found a memory module going bad, explaining why it gave no errors alone
at first.

I remember one of the blackouts in NYC in my teens. It proceeded slowly
and the first sign was the TV set freaking. I had it well apart before I
saw the small change creeping in to the lighting.

With such a mature industry as the PC, one would think that these things
would have been worked out despite the climb in device frequency.
Experience says differently though. I remember being amazed that memory
would work well on plug in modules :-]

Hope all turns out sensibly for you.

Are you referring to DIMM voltage? If so, I can boost it upward a few 10ths
from the standard 2.5v, but I tried 2.6v to no avail. That would be it.

As a recent development, I began getting crashes from a long-used DOS app
(Foxpro) that escalated yesterday to four incidents. At first I blamed a
software conflict, but after the computer rebooted itself once yesterday I
began looking more closely at a RAM failure. Thinking the culprit was the
new memory, I placed the old pair alone in slots 1 & 4 and to my surprise I
got 411 errors on pass 1 of Test 5 and >600 on the second pass. I then
placed the new (matched) set in the same slots and received no errors in
seven passes. Kingston asked that I send the old modules in for
replacement.

What's confusing is that I took these "bad" modules to a friend's 875p Asus
system and could not generate any errors with the same pair that was causing
the trouble before on my system. I shipped them to Kingston today, but I'm
not sure what's going on here. The new modules do not generate errors, but
the old do, either alone or combined with the new. The results of these
tests make no sense to me, unless you could assume the Gigabyte mobo is more
temperamental with memory than the Asus. But even then, what about the old
set (not a matched set) causes it to err on my setup and not on my friend's?

Electronics is that way when it comes to high frequencies and design
inconsiderations.
 
Hope all turns out sensibly for you.

I'll await receipt of Kingston's replacements and keep my fingers crossed.
One good thing is that while the old modules were not a matched pair, the
new ones will be.
 
Back
Top