My P4C800-E is destroying memory sticks - Why?

[Ed]

In 18 years of computing (assembling HUNDREDS of systems) I've NEVER
had RAM go bad. Since I built myself a system with the P4C800-E board
I've already killed 2 strips (2 different brands)

No overclocking at all.

P4C800-E Deluxe
1 Gig (well, now 512MB) Muchkin PC3500 level 1 (blue) DDR
P4 3Ghz
Top of the line 500 watt Antec Power supply

Initial mem (Oct 2003)was Corsair PC3200 (2x512). After 4 weeks one
strip went bad. System started doing some weird stuff - reboots
errors, lockups, corrupt files. Went through all the usual diags (I do
this for a living). Hundreds of errors running Memtest-86. Usual
swapping, isolated the bad stick.

Newegg replaced that pair with a pair of Mushkin PC3500 Blue 2X512
"Level 1".

3 months later EXACT SAME THING is happening to the Mushkin. Isolated
the problem to 1 bad stick.

Anyone else having problems with the P4C800-E Deluxe ?

Voltages on PS all check out just fine.

Only thing left (other then real bad luck) is the motherboard.

Anyone hear of such a problem ??

 

[Paul]

In 18 years of computing (assembling HUNDREDS of systems) I've NEVER
had RAM go bad. Since I built myself a system with the P4C800-E board
I've already killed 2 strips (2 different brands)

No overclocking at all.

P4C800-E Deluxe
1 Gig (well, now 512MB) Muchkin PC3500 level 1 (blue) DDR
P4 3Ghz
Top of the line 500 watt Antec Power supply

Initial mem (Oct 2003)was Corsair PC3200 (2x512). After 4 weeks one
strip went bad. System started doing some weird stuff - reboots
errors, lockups, corrupt files. Went through all the usual diags (I do
this for a living). Hundreds of errors running Memtest-86. Usual
swapping, isolated the bad stick.

Newegg replaced that pair with a pair of Mushkin PC3500 Blue 2X512
"Level 1".

3 months later EXACT SAME THING is happening to the Mushkin. Isolated
the problem to 1 bad stick.

Anyone else having problems with the P4C800-E Deluxe ?

Voltages on PS all check out just fine.

Only thing left (other then real bad luck) is the motherboard.

Anyone hear of such a problem ??

With your multimeter, check the voltage being fed to the DIMMs.
Normally, there will be some kind of conversion circuit on
the mobo, to make 2.5V for the DIMMs. On some motherboards, the
voltage is programmable via the BIOS, which means the regulator
circuit may have some signals coming from GPIO pins on the
SuperI/O chip. If this regulator is malfunctioning or there
is some other ohmic fault on the board, or the regulator is
being misprogrammed, you could be running your DIMMs at 3.3V.

Memory chips themselves have a high voltage rating. But, inside
the chip, they have their own regulators for internal circuits.
If a chip is run with excessive voltage, it gets hot, because
the voltage regulator inside the memory chip will be dumping
heat. There can also be issues with the amplitude of reflections
on the memory bus - this is another reason for keeping the
supply on the chips to 2.8V or so. Reflections clipped by ESD
diodes (parasitic or otherwise) cause currents to flow on every
memory transaction, and if this current is more than the
diode strucures can handle, this is another potential problem
area.

So, start by verifying what is being fed to the DIMMs.

HTH,
Paul

 

[Philip Callan]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Ed wrote:
| In 18 years of computing (assembling HUNDREDS of systems) I've NEVER
| had RAM go bad.

Well, welcome to the 21st century, when working with high quality,
low-static tolerant and very voltage sensitive RAM, things tend to go
wrong easily.

(and for the record, the only RAM I ever had 'break' on me, did exactly
that, 30 pin SIPP, which became 29-pin....) heheh

| No overclocking at all.
| P4C800-E Deluxe
| 1 Gig (well, now 512MB) Muchkin PC3500 level 1 (blue) DDR
| P4 3Ghz
| Top of the line 500 watt Antec Power supply

Okay, your going to have to be more specific about system configuration,
What video card, any other PCI devices? how about Hard drives, optical
drives? Operating system? 2K? XP? XP Pro?

| Initial mem (Oct 2003)was Corsair PC3200 (2x512). After 4 weeks one
| strip went bad. System started doing some weird stuff - reboots
| errors, lockups, corrupt files. Went through all the usual diags (I do
| this for a living). Hundreds of errors running Memtest-86. Usual
| swapping, isolated the bad stick.
|
| Newegg replaced that pair with a pair of Mushkin PC3500 Blue 2X512
| "Level 1".
|
| 3 months later EXACT SAME THING is happening to the Mushkin. Isolated
| the problem to 1 bad stick.

Did you bother to note if it was the same slot the last stick failed in?

| Anyone else having problems with the P4C800-E Deluxe ?

Zero problems here.

| Voltages on PS all check out just fine.

How did you determine voltages were ok? software monitoring?
MBM/AsusProbe? or with a multitester?

| Only thing left (other then real bad luck) is the motherboard.

Unfortunately, bad luck sometimes runs in spades.

| Anyone hear of such a problem ??

No, of course, I've never heard of anyone buying Mushkin PC3500 if they
dont overclock, why did you move from Corsair 3200 to ram that only has
value to those who *DO* overclock their FSB and need to maintain a 1:1
ratio.

Nobody here is going to judge you any different if you overclock, in
fact it may help isolate your problem.

For the record, you can google my name and *asus in groups.google.com
and get a recent post I put about my P4C800-E, like I said then, I built
this box by hand, and since the 10th of October 2003, it has been
running in essence 24/7 excluding the occassional reboot after
installing something, rock solid, zero problems.

Philip
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (MingW32)

iD8DBQFANf685sKixANmEMgRApk7AKCqnX6vjNIcZk2GkYjqNsPV9hJDegCfXWQD
RABzn492Npb/CXZKabR3t7o=
=vEMT
-----END PGP SIGNATURE-----

 

[MaH]

If things start to go wrong after a while ALLWAYS suspect the PSU.
It doesnt matter how expensive it is, it can still break and cause havoc.

 

[Ed]

I swear, never overclock. In fact the memory is running below spec
according to the bios (seems the Mushkin eprom isn't properly read by
the P4C800). Never play games. 99% of the time the system is used for
editing videos. I bought that memory only because it was supposed to
be the highest quality. Same for the 2nd batch (Mushkin). I also
liked the pretty blue heat spreaders.

Video card is a All-In-Wonder 9700
Only other PCI device is a Canopus A to D card for video conversion
Win XP-Pro
2 X 80Gig Seagate SATA raid 0 running off the ICH5R
Sony DRU-510 DVD burner
Sony DVD reader

As for the memory slot ..... dammit ... no. I don't remember the first
time nor to I even remember this time (senior-itis). That was my first
thought.

Checked PC voltages at the connector on the MB which jived with the
BIOS display.

Something tells me this is a Motherboard issue. Can't see it being the
PS or it would effect more then just memory.

Ed

 

[Philip Callan]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Ed wrote:
| I swear, never overclock. In fact the memory is running below spec
| according to the bios (seems the Mushkin eprom isn't properly read by
| the P4C800).

Actually it is, almost all memory PC3200+ will identify as PC3200, thats
because there is no JEDEC standard above that yet, so no chip will have
its SPD set higher than 3200.

That is, *it is being read by the board properly* but your right, it is
running 'under speed' for what Mushkin tested it too (PC3500 rates)

| Never play games. 99% of the time the system is used for
| editing videos. I bought that memory only because it was supposed to
| be the highest quality. Same for the 2nd batch (Mushkin). I also
| liked the pretty blue heat spreaders.
|

Yeah, I do a lot of capture / editing / burning with mine from MiniDV,
the board and its chipset are fantastic for this (esp w/ onboard Firewire)

But I game a fair bit too, so I think in a lot of ways I stress my board
more than yours had been while killing memory

| Video card is a All-In-Wonder 9700

Woo, same card, but mines a AIW.
Out of curiosity, why didnt you go for a 9600Pro 8xAGP or something? was
the Videocard a legacy from old box, never understood people buying a
kickass 3d card, and then rarely using it. Most of the video editors
I've seen seem to favor Matrox cards, as their 2D is unparalled and
colors are sharp.

| Only other PCI device is a Canopus A to D card for video conversion
| Win XP-Pro
| 2 X 80Gig Seagate SATA raid 0 running off the ICH5R
| Sony DRU-510 DVD burner
| Sony DVD reader

Dang, you know you got less in your box pulling current than me, so that
kinda shot one thing I was thinking out the window.

|
| As for the memory slot ..... dammit ... no. I don't remember the first
| time nor to I even remember this time (senior-itis). That was my first
| thought.

Heheh, yeah, my guess (and this is just a guess) you may have a bad
slot, with pins bent or somthing in the DIMM slot, something shorting
perhaps?

| Checked PC voltages at the connector on the MB which jived with the
| BIOS display.

Yeah, with an Antec 500w (I only run a TruePower 380S) I didnt think
power quality was going to be a problem, but its something to check.

| Something tells me this is a Motherboard issue. Can't see it being the
| PS or it would effect more then just memory.

Hell yeah Bad PSU's start trashing parts as fast as Darl McBride lies....

| Ed

Well, honestly I think your right, in that the board is faulty, I would
try to RMA it if still possible, but please dont take it as an
indication the this model board is bad, I think you just got the bad one
so that I could get the good one! (We both bought in Oct 2003) hehe.

Good luck with ASUS (or your Vendor)
Let us know what happens

Philip
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (MingW32)

iD8DBQFANmO35sKixANmEMgRApgbAKDTI+A98i82bFrLtM86lcM6+tnw7wCfYtt+
y0nu7N0nKg7v+O0UJzZIODQ=
=a7yM
-----END PGP SIGNATURE-----

 

[Ed]

Out of curiosity, why didnt you go for a 9600Pro 8xAGP or something? was

the Videocard a legacy from old box, never understood people buying a
kickass 3d card, and then rarely using it. Most of the video editors
I've seen seem to favor Matrox cards, as their 2D is unparalled and
colors are sharp.

Err, the video card is an ALL-IN-WONDER 9700 PRO ("100" better then
the 9600) AGP 8X - I wanted to use the video capture off it but for
some reason sound was out of sync with the video when I did
analog->digital - so I bought the Canopus board and still use the
Tuner to watch some TV. It replaced an old Radeon AIW that served me
well for about 3 years.

RMA'ing the board to ASUS would be a joke. They expect you to be w/o
a motherboard for 4 weeks (stupid system - no x-ship) and I don't
believe Newegg will do an RMA after 4 months. I may just buy a new
one then sell this to some jamoak on eBay (just kidding - on the eBay
part).

 

[Kyle Brant]

|
| With your multimeter, check the voltage being fed to the DIMMs.
| Normally, there will be some kind of conversion circuit on
| the mobo, to make 2.5V for the DIMMs. On some motherboards, the
| voltage is programmable via the BIOS, which means the regulator
| circuit may have some signals coming from GPIO pins on the
| SuperI/O chip. If this regulator is malfunctioning or there
| is some other ohmic fault on the board, or the regulator is
| being misprogrammed, you could be running your DIMMs at 3.3V.
|
| Memory chips themselves have a high voltage rating. But, inside
| the chip, they have their own regulators for internal circuits.
| If a chip is run with excessive voltage, it gets hot, because
| the voltage regulator inside the memory chip will be dumping
| heat. There can also be issues with the amplitude of reflections
| on the memory bus - this is another reason for keeping the
| supply on the chips to 2.8V or so. Reflections clipped by ESD
| diodes (parasitic or otherwise) cause currents to flow on every
| memory transaction, and if this current is more than the
| diode strucures can handle, this is another potential problem
| area.
|
| So, start by verifying what is being fed to the DIMMs.
|
| HTH,
| Paul


I find the comment that DDR ICs have onboard regulators rather odd, so
I had a look at a Micron 512Mb chip datasheet.
http://download.micron.com/pdf/datasheets/dram/ddr/512MBDDRx4x8x16.pdf

There is no mention of an onboard regulator. Further, it strikes me
that it would be overkill for the mobo to provide 2.5v regulated to
the DIMM slots with each sdram chip further regulating the supply
signal, resulting in over-redundancy. I suspect the sdram chips have
an internal zener to help reduce the effect of Vdd spikes. However, I
do not believe sdram ICs include an internal voltage regulator. If
I'm incorrect, don't hesitate to direct me to further reading.

 

[Paul]

|
| With your multimeter, check the voltage being fed to the DIMMs.
| Normally, there will be some kind of conversion circuit on
| the mobo, to make 2.5V for the DIMMs. On some motherboards, the
| voltage is programmable via the BIOS, which means the regulator
| circuit may have some signals coming from GPIO pins on the
| SuperI/O chip. If this regulator is malfunctioning or there
| is some other ohmic fault on the board, or the regulator is
| being misprogrammed, you could be running your DIMMs at 3.3V.
|
| Memory chips themselves have a high voltage rating. But, inside
| the chip, they have their own regulators for internal circuits.
| If a chip is run with excessive voltage, it gets hot, because
| the voltage regulator inside the memory chip will be dumping
| heat. There can also be issues with the amplitude of reflections
| on the memory bus - this is another reason for keeping the
| supply on the chips to 2.8V or so. Reflections clipped by ESD
| diodes (parasitic or otherwise) cause currents to flow on every
| memory transaction, and if this current is more than the
| diode strucures can handle, this is another potential problem
| area.
|
| So, start by verifying what is being fed to the DIMMs.
|
| HTH,
| Paul


I find the comment that DDR ICs have onboard regulators rather odd, so
I had a look at a Micron 512Mb chip datasheet.
http://download.micron.com/pdf/datasheets/dram/ddr/512MBDDRx4x8x16.pdf

There is no mention of an onboard regulator. Further, it strikes me
that it would be overkill for the mobo to provide 2.5v regulated to
the DIMM slots with each sdram chip further regulating the supply
signal, resulting in over-redundancy. I suspect the sdram chips have
an internal zener to help reduce the effect of Vdd spikes. However, I
do not believe sdram ICs include an internal voltage regulator. If
I'm incorrect, don't hesitate to direct me to further reading.

I've seen hints of the presence of voltage regulation on various
web pages, but this is the first mention I found when looking a
few days ago:

http://www.cisl.columbia.edu/courses/spring-2002/ee6930/papers/00799866.pdf

"The chip is powered from a 2.5-V external supply, which is
regulated internally to 2.1 V for support circuitry, 1.5 V for
the array, and 1.0 V for RWD."

I'm still looking for a paper with a little more detail on internal
powering, because I think there is a lot more to it than this
terse description. Implementation details are not something to be
found in a chip datasheet, as the details can contain intellectual
property.

Paul

 

[Kyle Brant]

| In article <[email protected]>, "Kyle
Brant"
|
| > | >
| > |
| > | With your multimeter, check the voltage being fed to the DIMMs.
| > | Normally, there will be some kind of conversion circuit on
| > | the mobo, to make 2.5V for the DIMMs. On some motherboards, the
| > | voltage is programmable via the BIOS, which means the regulator
| > | circuit may have some signals coming from GPIO pins on the
| > | SuperI/O chip. If this regulator is malfunctioning or there
| > | is some other ohmic fault on the board, or the regulator is
| > | being misprogrammed, you could be running your DIMMs at 3.3V.
| > |
| > | Memory chips themselves have a high voltage rating. But, inside
| > | the chip, they have their own regulators for internal circuits.
| > | If a chip is run with excessive voltage, it gets hot, because
| > | the voltage regulator inside the memory chip will be dumping
| > | heat. There can also be issues with the amplitude of reflections
| > | on the memory bus - this is another reason for keeping the
| > | supply on the chips to 2.8V or so. Reflections clipped by ESD
| > | diodes (parasitic or otherwise) cause currents to flow on every
| > | memory transaction, and if this current is more than the
| > | diode strucures can handle, this is another potential problem
| > | area.
| > |
| > | So, start by verifying what is being fed to the DIMMs.
| > |
| > | HTH,
| > | Paul
| >
| >
| > I find the comment that DDR ICs have onboard regulators rather
odd, so
| > I had a look at a Micron 512Mb chip datasheet.
| >
http://download.micron.com/pdf/datasheets/dram/ddr/512MBDDRx4x8x16.pdf
| >
| > There is no mention of an onboard regulator. Further, it strikes
me
| > that it would be overkill for the mobo to provide 2.5v regulated
to
| > the DIMM slots with each sdram chip further regulating the supply
| > signal, resulting in over-redundancy. I suspect the sdram chips
have
| > an internal zener to help reduce the effect of Vdd spikes.
However, I
| > do not believe sdram ICs include an internal voltage regulator.
If
| > I'm incorrect, don't hesitate to direct me to further reading.
| >
| > --
| > Best regards,
| > Kyle
|
| I've seen hints of the presence of voltage regulation on various
| web pages, but this is the first mention I found when looking a
| few days ago:
|
|
http://www.cisl.columbia.edu/courses/spring-2002/ee6930/papers/0079986
6.pdf
|
| "The chip is powered from a 2.5-V external supply, which is
| regulated internally to 2.1 V for support circuitry, 1.5 V for
| the array, and 1.0 V for RWD."
|
| I'm still looking for a paper with a little more detail on internal
| powering, because I think there is a lot more to it than this
| terse description. Implementation details are not something to be
| found in a chip datasheet, as the details can contain intellectual
| property.
|
| Paul


Good and interesting info. Onboard regulators with lower voltage
outputs are perhaps best explained by the fact that the higher
frequencies of these devices are better implemented with a power rail
as low as possible. I note the article is 5 years old, so in terms of
computer technology, this article is actually quite old. Thanks for
the info. As to the revelation of intellectual property, it strikes
me that such technology is best protected through the patent process
in view of the known methodologies for reverse engineering an IC and
the relatively limited lifetime of such devices. Given patent
protection in most countries requires disclosure of the invention
sufficient for "one skilled in the art to practice the invention"
(this is the US legal requirement), I suspect further info can be
found via patent searches.