re: BSOD and Memory on Athlons
Friday, March 11, 2005 at 2:03 pm
Windows XP Annoyances Discussion Forum
Posted by Jack Gulley
(5917 messages posted)
Just to add a little more fuel to the fire on Athlon and memory related problems.
I have an Athlon PC3200 system, standard setting (no overclocking) MSI board, that
after several months of running Windows XP without problems, started throwing many
of the above BSOD's and could find nothing wrong. Latest versions of MemTest86_v3.2 ran OK over night.
Gave up and started running Windows ME on the system so I could use it. But then
things started going wrong there also and after switching back to XP, it was having
more frequent errors than before.
Was talking to a neighbor with several machines and he was having the same problem
with them. So I took three of his, mine and another of mine that had not had any
problems and set them up and started testing and swapping parts around. Found that
some combinations of different memory modules would not work in one machine, but
run OK in others. Some would only work when paired with specific other modules. Others
would work only when ran alone. Some only in specific combinations on a specific
machine (not just a specific model of board). (All PC3200) Then discovered that putting
in PC2700 (old good quality and well tested and used), worked on all the machines?
Just a little slower at a lower bus speed.
What got me, was that one 512MB PC3200 module would work just fine by itself,
yet when combined with any other module on two of the systems, would only show up
as a 256MB module??? Yet it worked just fine on two other systems. In one case, this
was the same exact system board and BIOS level, yet it worked OK in one and failed
in the other.
Also noticed that these memory parts all worked just fine in any combination on
a Giga-Byte P4 system board (unless there was also a memory error detected by MemTest86).
Did some creative swapping around of parts, weeded out one memory module that
was suspect, and one that was going bad, took notes and got all the systems working
error free. And had one module replaced by the vendor who sold it to me, but a different
replacement "brand".
First point is, that there is some memory bus compatibility problems with the
Athlon processors and/or the system boards they are used on. It could be a simple
thing such as the quality of the manufacturing process of each individual system
board and the thickness of the layers between signal paths that allow cross talk
or
signal quality problems between the processor and the memory sockets. This is something
system board makers should check closely and frequently, but seldom do in volume
production because of the expense of destroying boards to do the testing.
Then a week later, my machine with a single 512MB replacement module in it started
throwing errors again. Ran MemTest86 and it would pickup a single address and bit
error about once every eight to ten hours. Same bit every time. It also did the same
thing in other machines. This was a "cheap replacement" Vdata Markvision Value 512MB
DDR PC3200 CL=2.5 module, with a "Life Time Warranty". Over the next few days, the
error rate became more frequent to the point that it is now almost useless. Having
to run with a PC2700 module right now.
The second point is, that a single bit error like this can cause a very hard to
diagnose problem. Changing drivers or other programs just change the chances of the
failing bit causing a problem and the nature of the problem. It is a well known fact
that when such an error starts to occur, it it often temperature sensitive at first
and only gets worse over time until the part has a complete failure of the bit.
Started taking a close look at the memory modules and the others that were known
to cause problems in specific systems. They all had one noticeable thing in common.
The tops of the chips all had a rough or course texture, like they had been ground
off, instead of the normal smooth gloss texture from normal molding of the chips!
This is a sign that these parts may have been "re-branded or remarked", as they all
have the same low quality of stamping of the chip information on the refurbished
surface.
Just try to find out any information about Vdata or MarkVision parts and their
"Life Time Warranty". No address or contact point. Much the same for the other "suspect"
parts. And they are usually sold by vendors who have a no return policy after x number
of days, and who do NOT handle Warranty returns of these parts for you.
Taking it a step farther, I downloaded the latest copy of CPU-Z
(and found a bug in it which the author quickly fixed for us over night) and ran
it on all of these modules to look at the SPD information on their supported speed
specifications, and Manufacturer's name information. Again, all of the different
modules with problems and re-branded chips had a Manufacturer's ID in the SPD chip
of 0000000000000000 or blanks! Clearly a violation of the SPD memory
identification standards and an obvious attempt to evade Warranty responsibility,
if not an attempt to defraud.
Third point is, there appears to be a lot of low quality memory in the market.
While I don't think this is the cause of most of these problems, it sure points
out that there is a lot of "suspect" memory floating around in the market place and
that it can cause problems. And that some vendors may have quality problems with
their system board manufacturing process.
|
All messages in this thread [show all]
 |  | | | | |  | Confused (Adri1456: Fri, May 14, 2004, 3:29 pm) |
| | | | | |  | Crash (greine12: Tue, Apr 6, 2004, 4:57 pm) |
| | | | | | | | | re: Crash (MadMan: Sat, Apr 10, 2004, 1:19 pm) |
| | | | | | re: BSOD and Memory on Athlons (Jack Gulley: Fri, Mar 11, 2005, 2:03 pm) |
| |
| |
| |
Return to the Windows XP Discussion Forum
|
|
