P4P800-E Deluxe - Apacer 2x512MB PC-3200 CL2.5 (or not really CL 2.5!) - a word of warning + memtest

[Technik]

I've replaced my good and trusty ASUS P4PE/Intel 2.4GHz (533MHz) with
P4P800-E Deluxe/3.0MHz(800MHz FSB).
Now, don't get me wrong the system is running as smooth as baby's
behind, but ;o( it was freezing once in a while...
- it froze upon restore from suspend.
- then it froze when I was surfing the net - in Firefox!
- then it froze...
It was freezing on average 2 times a day. So it occured to me it
could be a problem with RAM timing. System runs okay then it just
locks up for no apparent reason. So I re-booted and entered BIOS
settings (after upgrading BIOS to 1007).
I've tinkered with chipset settings. Turned off RAM timing settings
by SPD - entered settings as prescribed by Apacer. No luck!
Then I increased first seting (CAS I believe) to 3 instead of
original 2.5 (as set by SPD on Apacer memory module).
Moral of the story - SPD is nice; but one cannot always trust it.

My system:
Case: ASUS TA-231
PSU: Enermax Whisper 350W
M/B: ASUS P4P800-E Deluxe H/W rev.1.02
CPU: Intel (Prescot) 3.0GHz (1MB cache) 800MHz FSB
RAM: 2x Apacer 512MB CL 2.5 (not!!!) PC-3200 <-- trouble (?!)
HDD: WDC 160GB PATA, WDC 160GB SATAII, 2xMaxtor 80GB SATA as RAID 0
FDD: Mitsubisihi LS-120 EIDE
Sound: Creative SB Live! 5.1 24-bit
Video: BFG 6600GT 128MB (nVidia)

BTW: why is memtest86 v3.2 always freezing on Memory Size Probe at
50%? Has anyone else seen this happening?

 

[Paul]

I've replaced my good and trusty ASUS P4PE/Intel 2.4GHz (533MHz) with
P4P800-E Deluxe/3.0MHz(800MHz FSB).
Now, don't get me wrong the system is running as smooth as baby's
behind, but ;o( it was freezing once in a while...
- it froze upon restore from suspend.
- then it froze when I was surfing the net - in Firefox!
- then it froze...
It was freezing on average 2 times a day. So it occured to me it
could be a problem with RAM timing. System runs okay then it just
locks up for no apparent reason. So I re-booted and entered BIOS
settings (after upgrading BIOS to 1007).
I've tinkered with chipset settings. Turned off RAM timing settings
by SPD - entered settings as prescribed by Apacer. No luck!
Then I increased first seting (CAS I believe) to 3 instead of
original 2.5 (as set by SPD on Apacer memory module).
Moral of the story - SPD is nice; but one cannot always trust it.

My system:
Case: ASUS TA-231
PSU: Enermax Whisper 350W
M/B: ASUS P4P800-E Deluxe H/W rev.1.02
CPU: Intel (Prescot) 3.0GHz (1MB cache) 800MHz FSB
RAM: 2x Apacer 512MB CL 2.5 (not!!!) PC-3200 <-- trouble (?!)
HDD: WDC 160GB PATA, WDC 160GB SATAII, 2xMaxtor 80GB SATA as RAID 0
FDD: Mitsubisihi LS-120 EIDE
Sound: Creative SB Live! 5.1 24-bit
Video: BFG 6600GT 128MB (nVidia)

BTW: why is memtest86 v3.2 always freezing on Memory Size Probe at
50%? Has anyone else seen this happening?

Is that the "USB Legacy Support" [Enabled] bug ? If that setting
is enabled, I think it causes problems for earlier versions of
memtest. Try disabling it.

In any case, there are plenty of shiny new versions of software
on www.memtest.org. memtest86+ is currently at version 1.6 (and
was based on the 3.2 code you are using, plus some bug fixes).

Increase Vdimm to at least 2.6V, and preferably, give the RAM
a touch more, like 2.7V. That may help a bit.

You are not finished testing, until you've tried the Prime95 "Torture
Test" (mersenne.org) and tried some 3D games for a few hours. If
you don't own any games, download 3DMark2001SE and run the demo loop
overnight - see if it is still running the next day.

And in the RAM business, nothing should surprise you. If the chips
have real manufacturer part numbers printed on them, you can look
them up, and see if the actual chips are rated for use on that
DIMM or not. If the chips have nothing printed on them, or some
bogus manufacturers name (the "happy ram" company), we call that
generic roulette.

Paul

 

[ñíñjà¤têç]

bump...
;-)

I've replaced my good and trusty ASUS P4PE/Intel 2.4GHz (533MHz) with
P4P800-E Deluxe/3.0MHz(800MHz FSB).
Now, don't get me wrong the system is running as smooth as baby's
behind, but ;o( it was freezing once in a while...
- it froze upon restore from suspend.
- then it froze when I was surfing the net - in Firefox!
- then it froze...
It was freezing on average 2 times a day. So it occured to me it
could be a problem with RAM timing. System runs okay then it just
locks up for no apparent reason. So I re-booted and entered BIOS
settings (after upgrading BIOS to 1007).
I've tinkered with chipset settings. Turned off RAM timing settings
by SPD - entered settings as prescribed by Apacer. No luck!
Then I increased first seting (CAS I believe) to 3 instead of
original 2.5 (as set by SPD on Apacer memory module).
Moral of the story - SPD is nice; but one cannot always trust it.

My system:
Case: ASUS TA-231
PSU: Enermax Whisper 350W
M/B: ASUS P4P800-E Deluxe H/W rev.1.02
CPU: Intel (Prescot) 3.0GHz (1MB cache) 800MHz FSB
RAM: 2x Apacer 512MB CL 2.5 (not!!!) PC-3200 <-- trouble (?!)
HDD: WDC 160GB PATA, WDC 160GB SATAII, 2xMaxtor 80GB SATA as RAID 0
FDD: Mitsubisihi LS-120 EIDE
Sound: Creative SB Live! 5.1 24-bit
Video: BFG 6600GT 128MB (nVidia)

BTW: why is memtest86 v3.2 always freezing on Memory Size Probe at
50%? Has anyone else seen this happening?

Is that the "USB Legacy Support" [Enabled] bug ? If that setting
is enabled, I think it causes problems for earlier versions of
memtest. Try disabling it.

In any case, there are plenty of shiny new versions of software
on www.memtest.org. memtest86+ is currently at version 1.6 (and
was based on the 3.2 code you are using, plus some bug fixes).

Increase Vdimm to at least 2.6V, and preferably, give the RAM
a touch more, like 2.7V. That may help a bit.

You are not finished testing, until you've tried the Prime95 "Torture
Test" (mersenne.org) and tried some 3D games for a few hours. If
you don't own any games, download 3DMark2001SE and run the demo loop
overnight - see if it is still running the next day.

And in the RAM business, nothing should surprise you. If the chips
have real manufacturer part numbers printed on them, you can look
them up, and see if the actual chips are rated for use on that
DIMM or not. If the chips have nothing printed on them, or some
bogus manufacturers name (the "happy ram" company), we call that
generic roulette.

Paul

ñíñjà¤têç

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

 

[Technik]

On Sun, 21 Aug 2005 22:05:09 GMT, (e-mail address removed) (Paul) wrote:

Is that the "USB Legacy Support" [Enabled] bug ? If that setting
is enabled, I think it causes problems for earlier versions of
memtest. Try disabling it.

This was exactly it! But I need USB Legacy Support to operate
memtes86+... I've had to get my old trusty ps/2 keyboard to do that.

In any case, there are plenty of shiny new versions of software
on www.memtest.org. memtest86+ is currently at version 1.6 (and
was based on the 3.2 code you are using, plus some bug fixes).

I've got memtest86+ v3.2 - it worked great ~8 hours running tests -
not a single problem. BTW: I've found out that PAT is enabled. It's
not suppossed to work for 865 chipset... ASUS vs. Intel... ASUS won
they made it again with help of a little "cheat" circuit.

Increase Vdimm to at least 2.6V, and preferably, give the RAM
a touch more, like 2.7V. That may help a bit.

Don't seem to find these settings in BIOS - I haven't seen a jumper to
adjust Vdimm... I'm not too keen on modifying motherboard either -
it's still under warranty.

You are not finished testing, until you've tried the Prime95 "Torture
Test" (mersenne.org)

8 hours straight - no problem.

and tried some 3D games for a few hours. If
you don't own any games,

6 hours straight of FarCry and CounterStrike: Source - demo loop and
me playing - no problem.

download 3DMark2001SE and run the demo loop
overnight - see if it is still running the next day.

And in the RAM business, nothing should surprise you. If the chips
have real manufacturer part numbers printed on them, you can look
them up, and see if the actual chips are rated for use on that
DIMM or not. If the chips have nothing printed on them, or some
bogus manufacturers name (the "happy ram" company), we call that
generic roulette.

Apacer chips and Apacer modules. I really should've purchased OCZ
2x512MB PC-3200 CL 2.5 kit - it's on sale now ;o(

Paul

Thanks Paul for all the advice - now I have more confidence in my new
system!

 

[Technik]

First of all - thanks to Paul for great advice - now things are
looking even better.

On Sun, 21 Aug 2005 22:05:09 GMT, (e-mail address removed) (Paul) wrote:

Is that the "USB Legacy Support" [Enabled] bug ? If that setting
is enabled, I think it causes problems for earlier versions of
memtest. Try disabling it.

This is exactly it - unfortunately without USB Legacy Support Enabled
keyboard is not working once memtest86+ boots up.

In any case, there are plenty of shiny new versions of software
on www.memtest.org. memtest86+ is currently at version 1.6 (and
was based on the 3.2 code you are using, plus some bug fixes).

memtest86+ tested out okay - I've had it looping all night ~ 8 hours.
Not a single error.

Increase Vdimm to at least 2.6V, and preferably, give the RAM
a touch more, like 2.7V. That may help a bit.

Did not try this one - don't have that option in BIOS - don't see a
jumper that would do that either. I don't want to modify motherboard
just yet - it's still under warranty - but since things are much more
stable now - I don't think this should be required.

You are not finished testing, until you've tried the Prime95 "Torture
Test" (mersenne.org)

I've run Prime95 Torture Test for more than 12 hours. No errors at
all.

and tried some 3D games for a few hours.

Counter Strike: Source and Far Cry - consecutively running for more
than 6 hours - looped demo and me playing. Not a glitch. Only
temperature of CPU and GPU rose few degrees ;o)

If
you don't own any games, download 3DMark2001SE and run the demo loop
overnight - see if it is still running the next day.

And in the RAM business, nothing should surprise you. If the chips
have real manufacturer part numbers printed on them, you can look
them up, and see if the actual chips are rated for use on that
DIMM or not. If the chips have nothing printed on them, or some
bogus manufacturers name (the "happy ram" company), we call that
generic roulette.

The chips are made by or for Apacer. This is what's printed on them.

Paul

On top of everything I've done some extra reading and probing. It
turned out that P4P800 Deluxe by ASUS has PAT enabled by default even
though Intel claims it's only possible for 875 chipset. ASUS defeated
that by building a special circuit which allows PAT to be enabled on
865.

Thanks again Paul.

 

[Roland Scheidegger]

Apacer chips and Apacer modules. I really should've purchased OCZ
2x512MB PC-3200 CL 2.5 kit - it's on sale now ;o(

In my experience, apacer modules are of quite good quality. They may not
be as flashy as some overclockers module, but at least in the past were
built with decent ram chips (I've seen modules with infineon and samsung
chips). In fact, 2 sdram modules I have are rated cl 3 only and the spd
programmed accordingly, whereas the chips are rated for cl 2 indeed.
Though if these modules now use chips which are marked apacer,
everything is possible I guess. Looking at shops today still lists
modules as apacer/infineon or apacer/samsung though, so I'm not sure
what's going on. In general, apacer may not be a well recognized brand
in europe/usa, but afaik the definitely have brand recognition in asia.
So, imho, apacer really doesn't produce "generic ram", I'd rate them
about the same as kingston value (i.e. not for overclockers, but with
decent warranty, and usually flawless).
(in fact, I prefered apacer to kingston value - kingston lost my respect
when they used x4 chips in their value line, something which just
doesn't work usually with more than 1 module, regardless if the
manufacturer is kingston or not, apacer never (afaik) produced such
crappy modules. I think though they've stopped that stupidity some time
ago, and I bought some modules from them for the last pc - if only
because their availability is much better than apacer here.)

Roland