ATT> memory experts> re>p4p800e-deluxe

[ñíñjà¤têç]

For one "occasional" reason or another(I won't get into details, but
we'll say that performance was not up to par...), I was prompted to
perform some memory testing on my " fa$t ram in the above mentioned
"p4p800e-deluxe"

In june I purchased some crucial *balistic pc3200 (2-2-2-6 @2.8v)
4sticks @256mb for a total of 1 gb, just to give me some extra
overhead when and if I decide to heat up my northwood 2.8 @800fsb.
Benckmark gains were not that *steller and the occasional error or
other problem would arise.

Test results from "memtest" on a clean boot would vary from none to
many errors reported. The results were inconsistant, but the errors
were there. Tests were performed at a minimum of 2-3 hrs at 2-2-2-6 /
2.8v and at spd / auto voltage overnight.

Tests results from "memtest86" were similar, ranging from one error
overnight to 1500+ in the span of less than one hour. Very
inconsistant and sporadic.

I broke down the pairs and ran one pair without incident. The second
pair produced a single error in a 3-4hr run. Testing individual sticks
from this pair produced no errors. Tests were run 2hrs+ at 2-2-2-6 /
2.8v and then at spd / auto voltage overnight.

I have a retail enermax 350w psu that I've been using for less than a
year and I'm running stock crucial pc3200 (3-3-3-8) 2x512 mb as I type
this. I'm gonna run this pair overnight, but so far a 1hour run with
memtest86 produced no errors.

Other points of concerning the balistic modules and this board:
-with all ram slots full, spd / auto sets as 2.5-2-2-8, where as cpu-z
reads spd as 2-2-2-8 and its running at 2.5-2-2-8!
-I can set the voltage at 2.65v, 2.75v, or 2.85v and run the memory at
any latency (spd or manual 2-2-2-6) and the computer keeps running.
Most errors are transparent, noticeable only by the responsivness of
any given task. Occastional windows will generate an error report and
this is probably appilcation specific.

So I'm assuming the memory hub on this board is very flaky or it
cannot handle the balistic timmings very well. If it is the latter, I
can live with it, but if it is the former, then how far can I take
this?

I'm gonna throw the balistic pairs in my son's intel board to see what
will happen. Otherwise I appreciate any additional feedback!
Regards,

ñíñjà¤têç

---END---PGP---SIGNATURE---

 

[Gordon Scott]

For one "occasional" reason or another(I won't get into details, but
we'll say that performance was not up to par...), I was prompted to
perform some memory testing on my " fa$t ram in the above mentioned
"p4p800e-deluxe"

In june I purchased some crucial *balistic pc3200 (2-2-2-6 @2.8v)
4sticks @256mb for a total of 1 gb, just to give me some extra
overhead when and if I decide to heat up my northwood 2.8 @800fsb.
Benckmark gains were not that *steller and the occasional error or
other problem would arise.

Test results from "memtest" on a clean boot would vary from none to
many errors reported. The results were inconsistant, but the errors
were there. Tests were performed at a minimum of 2-3 hrs at 2-2-2-6 /
2.8v and at spd / auto voltage overnight.

Tests results from "memtest86" were similar, ranging from one error
overnight to 1500+ in the span of less than one hour. Very
inconsistant and sporadic.

I broke down the pairs and ran one pair without incident. The second
pair produced a single error in a 3-4hr run. Testing individual sticks
from this pair produced no errors. Tests were run 2hrs+ at 2-2-2-6 /
2.8v and then at spd / auto voltage overnight.

I have a retail enermax 350w psu that I've been using for less than a
year and I'm running stock crucial pc3200 (3-3-3-8) 2x512 mb as I type
this. I'm gonna run this pair overnight, but so far a 1hour run with
memtest86 produced no errors.

Other points of concerning the balistic modules and this board:
-with all ram slots full, spd / auto sets as 2.5-2-2-8, where as cpu-z
reads spd as 2-2-2-8 and its running at 2.5-2-2-8!
-I can set the voltage at 2.65v, 2.75v, or 2.85v and run the memory at
any latency (spd or manual 2-2-2-6) and the computer keeps running.
Most errors are transparent, noticeable only by the responsivness of
any given task. Occastional windows will generate an error report and
this is probably appilcation specific.

So I'm assuming the memory hub on this board is very flaky or it
cannot handle the balistic timmings very well. If it is the latter, I
can live with it, but if it is the former, then how far can I take
this?

I'm gonna throw the balistic pairs in my son's intel board to see what
will happen. Otherwise I appreciate any additional feedback!
Regards,

ñíñjà¤têç

---END---PGP---SIGNATURE---

stick with OCZ ram

 

[ñíñjà¤têç]

Why ? Are you running this board with low latency OCZ memory ? I've
always supported Crucial and so far there has been no reason to
deflect my loyalty.

As an update, the *Balistic memory is going on 6hrs(memtste86) on the
Intel board (865perl) with pat enabled (by default), with not an
error! All settings are auto and it's running at 2-2-2-7. However, I
must also add that the chip is a 533mhz northwood so that means the
memory is running at 266mhz or pc2100. Maybe the chips are miss
labelled pc3200? Nawwww, can't be...I'll see if I can change the
timmigs.

A 5 hr test with memtest86 on the p4p800e-deluxe produced no errors as
well... with the bottom grade crucial modules out of the intel rig. So
it seems some parts cannot run together and in fact there is truth to
mfg's authorized dual channel ddr vendor's lists...

ñíñjà¤têç

---END---PGP---SIGNATURE---

 

[Paul]

Why ? Are you running this board with low latency OCZ memory ? I've
always supported Crucial and so far there has been no reason to
deflect my loyalty.

As an update, the *Balistic memory is going on 6hrs(memtste86) on the
Intel board (865perl) with pat enabled (by default), with not an
error! All settings are auto and it's running at 2-2-2-7. However, I
must also add that the chip is a 533mhz northwood so that means the
memory is running at 266mhz or pc2100. Maybe the chips are miss
labelled pc3200? Nawwww, can't be...I'll see if I can change the
timmigs.

A 5 hr test with memtest86 on the p4p800e-deluxe produced no errors as
well... with the bottom grade crucial modules out of the intel rig. So
it seems some parts cannot run together and in fact there is truth to
mfg's authorized dual channel ddr vendor's lists...

ñíñjà¤têç

I've run 4x512MB PC3200 2-2-2 Ballistix on my P4C800-E with no
problems (zero errors). It could be there is a problem with your
motherboard, such as a wobbly Vdimm or termination voltage regulator.
I don't know whether Asus RMA repair, would be able to detect and
repair a problem like that. There is always a chance, if you RMA
the board, they'll return it with "no fault found". When a board
is operating flaky, there is no guarantee they'll be able to
detect the problem. (It really needs to go through the factory
test procedure.)

Have you tried Prime95 Torture Test ? (mersenne.org)
I'd be interested in what you find while running Prime95.

Paul

 

[ñíñjà¤têç]

I've run 4x512MB PC3200 2-2-2 Ballistix on my P4C800-E with no
problems (zero errors). It could be there is a problem with your
motherboard, such as a wobbly Vdimm or termination voltage regulator.
I don't know whether Asus RMA repair, would be able to detect and
repair a problem like that. There is always a chance, if you RMA
the board, they'll return it with "no fault found". When a board
is operating flaky, there is no guarantee they'll be able to
detect the problem. (It really needs to go through the factory
test procedure.)

Have you tried Prime95 Torture Test ? (mersenne.org)
I'd be interested in what you find while running Prime95.

Paul

As time allows, I'll try that out (prime95). As for the RMA, this was
actually a replacement for an RMA'd p4p800d...and I certainly agree to
the flaky parts idea. I'll try this memory for a bit and see what
prime95 turns up.

Oh, I was mistaken in my second post, the *Balistic memory is in fact
running at 332mhz (pc2700 with pat @ 2-2-2-7). Intela auto settings
seem to work quite well and at least the intel board knows good memory
when it's there! Also, eveytime I check Asus probe or AI booster, all
my voltages are right there...

I just can't believe how much I'm enjoying using my comp again, it's
that noticeable...
Regards,

ñíñjà¤têç

---END---PGP---SIGNATURE---

 

[ñíñjà¤têç]

Paul:

As I mentioned previously, no problems here since swithcing to the
generic crucial memory modules (3-3-3-8). But I did notice some extra
settings in the bios which I would ask your opinion on:

-dram frequency now shows more options, ie: 266mhz ---> 533mhz,thats
up by at least 1...(533mhz)
-DDR reference voltage now shows 2.55v --->2.85v, thats up by
1...(2.55v)
-AGP VDDQ now shows 1.5v ---> 1.8v, that is also up by 1...(1.8v)

So simply changing the memory to the Balistic module has this effect
in my bios. Why ? Simply through the spd? Or can the bios/board
determine the max amperage?

Also, my Antec SurePower-SL350 (not enermax) has these specs:
+5v - 35a max/1.5 min
+12v - 16a max/.8 min
+3.3v - 28a max/.5 min
-5v - .5a max/0
-12v - .8a max/0
+5sb - 2 max/0
(+5v,+3.3v,+12v = 330w max)
(+5v,+3.3v = 230w max)

The P4P800e-deluxe has the lan active and the promise raid controler,
in addition I have: (all retail products...)

-ATI X850XT agp8x (no issue at crisis event)
-Audigy 2ZS
-1 ata hdd 30gb maxtor fdb (primary/single)
-2 ata hdd 30gb each maxtor fdb (promise-raid0)
-plextor dvdrw and liteon cdrw (secondary/master,slave)
-2 - 3pin zalman 80m fans (mobo plugs)
-2 - 4pin generic 80m w/color led
-usb is active and working, no peripherals plugged in at crisis event.

I'm thinking that after installing the 4 dimms of Balistic, some of my
power rail(s) may not be adaquet. Otherwise why else would the generic
dimms run without fault? I have 2 more generic dimms to put in for a
total of 4, which I will test tonight. I checked out that prime95
link, but I couldn't find any reference to torture test..?
Regards,

ñíñjà¤têç

---END---PGP---SIGNATURE---

 

[Paul]

Paul:

As I mentioned previously, no problems here since swithcing to the
generic crucial memory modules (3-3-3-8). But I did notice some extra
settings in the bios which I would ask your opinion on:

-dram frequency now shows more options, ie: 266mhz ---> 533mhz,thats
up by at least 1...(533mhz)
-DDR reference voltage now shows 2.55v --->2.85v, thats up by
1...(2.55v)
-AGP VDDQ now shows 1.5v ---> 1.8v, that is also up by 1...(1.8v)

So simply changing the memory to the Balistic module has this effect
in my bios. Why ? Simply through the spd? Or can the bios/board
determine the max amperage?

It is possible you never cleared the CMOS or used the BIOS
"Load Setup Defaults" function, after your last BIOS flash ?
When you plugged in new DIMMs, at the very least, the BIOS
recomputes DMI/ESCD information, and perhaps the CMOS got
fixed at the same time. Just a guess. If plugging in the
Ballistix again, makes the options revert to their old
values, that would shoot the theory down. The voltage
options offered by the BIOS should be a stable part of
the BIOS interface, and the range offered should not
jump around.

The motherboard cannot measure the current used by the DIMMs.
(The Super I/O chip only measures voltages, and there are
a limited number of channels available.) There have been
motherboards, where the MOSFETs burned up, so it appears
there isn't even a current limiting function there.

The JEDEC spec only declares the interface type (byte 8), but
doesn't state the voltage to be used. The BIOS uses its
best judgement, based on the fact that chip specs are 2.6V
for DDR400 and 2.5V for slower DDR RAM. Overclocker RAM can
have other voltages mentioned by the manufacturer, which is
how the RAM is able to overclock. Only the user can apply
those voltages.

(JEDEC SPD spec)
http://web.archive.org/web/20030417070529/http://www.jedec.org/download/search/4
_01_02_04R11A.PDF

Also, my Antec SurePower-SL350 (not enermax) has these specs:
+5v - 35a max/1.5 min
+12v - 16a max/.8 min
+3.3v - 28a max/.5 min
-5v - .5a max/0
-12v - .8a max/0
+5sb - 2 max/0
(+5v,+3.3v,+12v = 330w max)
(+5v,+3.3v = 230w max)

The P4P800e-deluxe has the lan active and the promise raid controler,
in addition I have: (all retail products...)

-ATI X850XT agp8x (no issue at crisis event)
-Audigy 2ZS
-1 ata hdd 30gb maxtor fdb (primary/single)
-2 ata hdd 30gb each maxtor fdb (promise-raid0)
-plextor dvdrw and liteon cdrw (secondary/master,slave)
-2 - 3pin zalman 80m fans (mobo plugs)
-2 - 4pin generic 80m w/color led
-usb is active and working, no peripherals plugged in at crisis event.

I'm thinking that after installing the 4 dimms of Balistic, some of my
power rail(s) may not be adaquet. Otherwise why else would the generic
dimms run without fault? I have 2 more generic dimms to put in for a
total of 4, which I will test tonight. I checked out that prime95
link, but I couldn't find any reference to torture test..?
Regards,

ñíñjà¤têç

The RAM is likely powered from 3.3V, via a linear regulator circuit.
The linear regulator drops the voltage from 3.3V to the 2.55-2.85
range. Asus likes to use a 14 pin quad opamp for their linear
regulating functions, with MOSFETs used as the series pass element.

On my P4C800-E, the load on 3.3V is about 14.4 amps or so. That is
with 4 sticks, DDR400 2-2-2, PAT enabled, running memtest86+. With
one stick of RAM, and sitting in the BIOS screen, the current draw
drops to 8.1 amps from 3.3V. That is the range of current consumption.
And the 5V consumption by the motherboard itself is less than 1 amp,
which means you will not even be close to the combined power limit
for +3.3V and +5V output windings.

I would check the Asus download page, and see if there are any
BIOS releases that mention "improved memory stability" or the
like. For example, on my P4C800-E, using the 1014 BIOS that my
board came with, the BIOS memory timings displayed on the screen,
are not the settings actually being used by the RAM. I had to use
CPUZ (www.cpuid.com) to verify the RAM timings, and they were all
over the place. Also, if you are offered a setting like "Turbo",
don't use it - setting things up manually makes a lot more
sense.

It could be that your problems would clear up, with a particular
BIOS release. I don't really see the hardware being responsible,
unless:

1) Northbridge is fried (and your 3-3-3-8 RAM test proves that is
not the case).
2) Vdimm or Vterm are out of spec. Again, your 3-3-3-8 RAM test
proves that is not the case.
3) Ballistix is out of spec. Testing the Ballistix on another
board will prove or disprove that.

I would try Googling or using the private forums, to see if any
BIOS releases for your board, have had problems with the setup
of memory timing. Try to find a BIOS that other people say works
for them, then repeat your Ballistix testing. And verify via
CPUZ, that the BIOS is doing what it says it is doing - my
BIOS lied to me.

As for Prime95, there is an option called "Torture Test", and
that is what you use for testing. The mixed test does some
big and some small arrays - the big array tests memory, the
small array stays in cache and makes the processor hotter.

HTH,
Paul

 

[ñíñjà¤têç]

It is possible you never cleared the CMOS or used the BIOS
"Load Setup Defaults" function, after your last BIOS flash ?
When you plugged in new DIMMs, at the very least, the BIOS
recomputes DMI/ESCD information, and perhaps the CMOS got
fixed at the same time. Just a guess. If plugging in the
Ballistix again, makes the options revert to their old
values, that would shoot the theory down. The voltage
options offered by the BIOS should be a stable part of
the BIOS interface, and the range offered should not
jump around.

After flashing, it is always my habit to load set up defaults save and
reboot, then make my bios adjustments. I am positive those extra
settings were not present during the balistic/memtest testing fase...
In fact, I remember flashing it with the raid0 install intact, just in
case I missed one of my pre-install adjustments...

I was hoping that would be the case...with the bios values being a
"set" option.

The motherboard cannot measure the current used by the DIMMs.
(The Super I/O chip only measures voltages, and there are
a limited number of channels available.) There have been
motherboards, where the MOSFETs burned up, so it appears
there isn't even a current limiting function there.

The JEDEC spec only declares the interface type (byte 8), but
doesn't state the voltage to be used. The BIOS uses its
best judgement, based on the fact that chip specs are 2.6V
for DDR400 and 2.5V for slower DDR RAM. Overclocker RAM can
have other voltages mentioned by the manufacturer, which is
how the RAM is able to overclock. Only the user can apply
those voltages.

(JEDEC SPD spec)
http://web.archive.org/web/20030417070529/http://www.jedec.org/download/search/4
_01_02_04R11A.PDF


The RAM is likely powered from 3.3V, via a linear regulator circuit.
The linear regulator drops the voltage from 3.3V to the 2.55-2.85
range. Asus likes to use a 14 pin quad opamp for their linear
regulating functions, with MOSFETs used as the series pass element.

On my P4C800-E, the load on 3.3V is about 14.4 amps or so. That is
with 4 sticks, DDR400 2-2-2, PAT enabled, running memtest86+. With
one stick of RAM, and sitting in the BIOS screen, the current draw
drops to 8.1 amps from 3.3V. That is the range of current consumption.
And the 5V consumption by the motherboard itself is less than 1 amp,
which means you will not even be close to the combined power limit
for +3.3V and +5V output windings.

I would check the Asus download page, and see if there are any
BIOS releases that mention "improved memory stability" or the
like. For example, on my P4C800-E, using the 1014 BIOS that my
board came with, the BIOS memory timings displayed on the screen,
are not the settings actually being used by the RAM. I had to use
CPUZ (www.cpuid.com) to verify the RAM timings, and they were all
over the place. Also, if you are offered a setting like "Turbo",
don't use it - setting things up manually makes a lot more
sense.

You know what, I did update the bios around when all this started to
happen. I came home and flashed it prior to a new install. I took the
install off the raid and put it on the ide. Initially, I thought my
raid0 install was causing all the funk. So at least it seems that this
problem was prevalent from the last bios to the current one. I will
definatley have to put the Balistic dimms back in to see what shows up
in the bios settings...just to be sure...

It could be that your problems would clear up, with a particular
BIOS release. I don't really see the hardware being responsible,
unless:

1) Northbridge is fried (and your 3-3-3-8 RAM test proves that is
not the case).
2) Vdimm or Vterm are out of spec. Again, your 3-3-3-8 RAM test
proves that is not the case.
3) Ballistix is out of spec. Testing the Ballistix on another
board will prove or disprove that.

I can set the intel board(with the balistic memory @332mhz) to 400mhz
for testing and it should run at least spd spec with 2.65v..? But the
proc is only a 533mhz. Mind you, running memtest86, it shouldn't
matter. Additionally, I would boot to windows and install the intel
bios tweaker to read the memory voltages. I'll try that to test
Balistic modules @ DDR400. For optimal system spec, Intel does not
recommend that. As long as runs a test, I won't worry about it...

I would try Googling or using the private forums, to see if any
BIOS releases for your board, have had problems with the setup
of memory timing. Try to find a BIOS that other people say works
for them, then repeat your Ballistix testing. And verify via
CPUZ, that the BIOS is doing what it says it is doing - my
BIOS lied to me.

My board is rev 1.02, perhaps I will find a compatibility problem.
Thanks to you, I do check the timings with cpuZ...(and memtest86).

As for Prime95, there is an option called "Torture Test", and
that is what you use for testing. The mixed test does some
big and some small arrays - the big array tests memory, the
small array stays in cache and makes the processor hotter.

I'm on it, thanks again for all your help / suggestions!
;-)

HTH,
Paul

Regards,

ñíñjà¤têç

---END---PGP---SIGNATURE---

 

[ñíñjà¤têç]

The Balistic memory has been getting problematic on the intel board
running at DDR333 (533fsb cpu). This is upsetting, it must be
something with the dimms and the power. The symptoms are the same as
it seems to start, run fine, then just gets worse. I'll have to
contact crucial now as I'm pretty much out of tests...
:-(

ñíñjà¤têç

---END---PGP---SIGNATURE---

 

[Paul]

The Balistic memory has been getting problematic on the intel board
running at DDR333 (533fsb cpu). This is upsetting, it must be
something with the dimms and the power. The symptoms are the same as
it seems to start, run fine, then just gets worse. I'll have to
contact crucial now as I'm pretty much out of tests...
:-(

ñíñjà¤têç

---END---PGP---SIGNATURE---

Yup. Return them and get what you paid for.

Paul