P4P800-E Deluxe: BIOS Chipset settings

[Peter Wagner]

Hi

I've got
an Intel(R) Pentium(R) 4 CPU 3.00GHz Northwood,
2 Kingston DDR-SDRAM PC3200 à MBytes

CPU-Z reports:
Timings table
Frequency (MHz) 166 200
CAS# 2.0 2.5
RAS# to CAS# delay 3 3
RAS# Precharge 3 3
TRAS# 7 8

The Chipset settings are 'Auto'.

Should I set the settings 'Manual'?

TIA
p

 

[Pete D]

Looks like you have a problem Peter, they should be the same for both memory
sticks. I would manuallt set them to the settings on the right and see if
that works. With the Northwood P4 they should both be at 200Mhz.

 

[SpongeBob]

No, all is allright.

The data of CPU-Z should understand in this way:
With 166 MHz the Kingston RAM runs with 2/3/3/7 and
with 200 MHz the Kingston RAM runs with 2.5/3/3/8.

That's O.K and logical!

CPU-Z reports the same for my MDT-RAM.

BTW I also use the 'Auto'-settings without any problems - my system with a
P4P800 is rock stable.

 

[SpongeBob]

No, all is allright.

The data of CPU-Z should understand in this way:
With 166 MHz the Kingston RAM runs with 2/3/3/7 and
with 200 MHz the Kingston RAM runs with 2.5/3/3/8.

That's O.K and logical!

CPU-Z reports the same for my MDT-RAM.

BTW I also use the 'Auto'-settings without any problems - my system with a
P4P800 is rock stable.

 

[Pete D]

Will that still run in dual mode properly?

No, all is allright.

The data of CPU-Z should understand in this way:
With 166 MHz the Kingston RAM runs with 2/3/3/7 and
with 200 MHz the Kingston RAM runs with 2.5/3/3/8.

That's O.K and logical!

CPU-Z reports the same for my MDT-RAM.

BTW I also use the 'Auto'-settings without any problems - my system with a
P4P800 is rock stable.

 

[Frank Mennig]

Yes, of course.

Will that still run in dual mode properly?

 

[Peter Wagner]

No, all is allright.

The data of CPU-Z should understand in this way:
With 166 MHz the Kingston RAM runs with 2/3/3/7 and
with 200 MHz the Kingston RAM runs with 2.5/3/3/8.

That's O.K and logical!

CPU-Z reports the same for my MDT-RAM.

BTW I also use the 'Auto'-settings without any problems - my system
with a P4P800 is rock stable.

Mine is rock stable too, but can 'Auto' recognize the lowest values or will
it be better, one keys in lower values?

p

 

[Pete D]

Of course? Shouldn't they be matched?

Yes, of course.

 

[Wim Zefat]

Op Wed, 22 Sep 2004 23:37:46 +0200 schreef Peter Wagner:

Mine is rock stable too, but can 'Auto' recognize the lowest values or will
it be better, one keys in lower values?

Since your machine uses a clock speed of 200 Mhz (unless you overclock),
you should use the values stated for that speed: 2.5/3/3/8.

Take a look at the Tab Memory in CPU-Z, which tells you the timing your
memory actually uses (as opposed to the timing it *should* use, which you
can see in the Tab SPD, which is the one you reported here). If the timing
you see here is slower then 2.5/3/3/8, you can try setting the timing
manually. Otherwise just use Auto. Don't forget to test the memory
thoroughly though, for instance with memtest86, if you decide to change the
settings!

Mine ran at 2.5/4/4/8 using Auto setting, even though the RAM's SPD said it
should work at 2/3/3/6, so I set those values manually. Machine has been
rock solid at this setting for months now.

Wim

 

[Peter Wagner]

Op Wed, 22 Sep 2004 23:37:46 +0200 schreef Peter Wagner:

Wim

Thank you for your answer.

Since your machine uses a clock speed of 200 Mhz (unless you
overclock), you should use the values stated for that speed:
2.5/3/3/8.

Take a look at the Tab Memory in CPU-Z, which tells you the timing
your memory actually uses (as opposed to the timing it *should* use,
which you can see in the Tab SPD, which is the one you reported
here). If the timing you see here is slower then 2.5/3/3/8, you can
try setting the timing manually.

They are the _same_.

But Kingston says:
HyperX Features:
Description Settings

400MHz (PC3200) 2-3-2-6-1 (CAS Latency 2)


What does this means?
Correlates this to 2/3/2/6/1?
But I think 1 instead of 8 can't be right?!
What is the correct order of this numbers?


In an other paper says Kingston:

Programmable Burst length (2, 4, 8)
Programmable Burst type (sequential & interleave)
Timing Reference: 2-3-2-6-1 at +2.6V

Otherwise just use Auto. Don't
forget to test the memory thoroughly though, for instance with
memtest86, if you decide to change the settings!

Mine ran at 2.5/4/4/8 using Auto setting, even though the RAM's SPD
said it should work at 2/3/3/6, so I set those values manually.
Machine has been rock solid at this setting for months now.
Fine.


Wim

p

 

[Wim Zefat]

Op Fri, 24 Sep 2004 08:19:43 +0200 schreef Peter Wagner:

They are the _same_.

But Kingston says:
HyperX Features:
Description Settings

400MHz (PC3200) 2-3-2-6-1 (CAS Latency 2)

What does this means?
Correlates this to 2/3/2/6/1?
But I think 1 instead of 8 can't be right?!
What is the correct order of this numbers?

Read as 2-3-2-6 and forget the 1. Manufacturers use 5 numbers in the full
technical specification of memory chips, but only the first 4 are used for
setting the timings, which is why the last one is often left off.

In an other paper says Kingston:

Programmable Burst length (2, 4, 8)
Programmable Burst type (sequential & interleave)
Timing Reference: 2-3-2-6-1 at +2.6V

This probably means yours are 'selected' DIMM's. The numbers programmed
into the memory's SPD-chip represent 'safe' timings, which should always
work. You can sometimes get away with faster timings - and sometimes not.
'Selected' simply means that the manufacturer has tested the DIMM's at
faster timings and found that they can handle the speed.

(Yup, just looked at Kingston's site - see the bottom of the page with the
specifications: "All Kingston products are tested to meet our published
specifications. Some motherboards may not work at the published HyperX
memory speeds and timing settings." In other words: "It might work at the
speed we advertise, or it might not. If not, tough luck.")

It's up to you to decide if you want to try the faster settings. If it
works, you gain a little speed (don't expect more than something in the
order of 1 to 2 percent increase though). If it doesn't work at all and the
machine hangs, just turn it off and back on - the BIOS *should* notice the
machine did't start right the last time and automatically reset itself to
it's default settings, leaving you back where you were. (If it doesn't, you
may have to clear the CMOS to get the machine to start up again.)

If you do try it and it does work, it's advisable to run a memory tester
(memtest86 or similar) for an hour or so, to make sure the machine really
is stable at the new settings.

Wim

 

[Peter Wagner]

Op Fri, 24 Sep 2004 08:19:43 +0200 schreef Peter Wagner:

Wim

Thank you very much for your comprehensive explanation.

It's up to you to decide if you want to try the faster settings. If it
works, you gain a little speed (don't expect more than something in
the order of 1 to 2 percent increase though).

So little?!

If it doesn't work at
all and the machine hangs, just turn it off and back on - the BIOS
*should* notice the machine did't start right the last time and
automatically reset itself to it's default settings, leaving you back
where you were. (If it doesn't, you may have to clear the CMOS to get
the machine to start up again.)

To open the case, remove the battery or to make a short? :-(

If you do try it and it does work, it's advisable to run a memory
tester (memtest86 or similar) for an hour or so, to make sure the
machine really is stable at the new settings.

One can't detect errors w/o tester?

Wim

p

 

[Wim Zefat]

Op Fri, 24 Sep 2004 16:21:27 +0200 schreef Peter Wagner:

So little?!

See http://www.tomshardware.com/motherboard/20040119/index.html
for a comprehensive test of DIMM's with different timings. According to
this, the difference will hardly be noticeable (for a P4 that is, it seems
to make slightly more of a difference, though still not much, for the
Athlon 64).

To open the case, remove the battery or to make a short? :-(

Both, according to the manual. See section 2.6.1 - Clear RTC RAM. First
remove the battery, then move the jumper from pin 1-2 to 2-3. Wait ten
seconds, then put the jumper back the way it was and put the battery back.

One can't detect errors w/o tester?

Depends on how bad the errors are If it doesn't work at all, you'll
notice immediately - the computer won't even boot. But if you run the RAM
at the very edge of what it's capable of, it might produce ocaasional
errors - say just 1 or 2 bits wrong in an hour, or even less. That's enough
however to cause all sorts of mayhem - Windows sometimes hanging
unexpectedly or behaving strangely, applications and games which work most
of the time, but crash every once in a while without apparent reason and so
on. But since Windows and applications can crash for all sorts of reasons
(power fluctuations, hardware problems, driver problems and so on) - and
tend to crash on their own for no apparent reason once in a while anyway,
it can be hard to tell whether or not it's the memory which is at fault.

So it's always advisable to run a memory testing program for a while - that
way you'll know for sure whether or not the memory works perfectly. And
besides, memtest86 is free (you can download it from www.memtest86.com, or
a slightly different version from www.memtest.org), so all it'll cost you
is a bit of time. Just leave it running for an hour or so (or maybe even
overnight), then check whether it's noticed any errors. If not, you know
for certain the memory works A-OK.

Wim

 

[Peter Wagner]

Op Fri, 24 Sep 2004 16:21:27 +0200 schreef Peter Wagner:

See http://www.tomshardware.com/motherboard/20040119/index.html
for a comprehensive test of DIMM's with different timings. According
to this, the difference will hardly be noticeable (for a P4 that is,
it seems to make slightly more of a difference, though still not
much, for the Athlon 64).

Thank you for the link.

BTW
Do you know how much the performance will rise, if a machine with a P4-3000
runs in dual-channel instead in single-channel modus?

[...]

Depends on how bad the errors are If it doesn't work at all, you'll
notice immediately - the computer won't even boot. But if you run the
RAM at the very edge of what it's capable of, it might produce
ocaasional errors - say just 1 or 2 bits wrong in an hour, or even
less. That's enough however to cause all sorts of mayhem - Windows
sometimes hanging unexpectedly or behaving strangely, applications
and games which work most of the time, but crash every once in a
while without apparent reason and so on. But since Windows and
applications can crash for all sorts of reasons (power fluctuations,
hardware problems, driver problems and so on) - and tend to crash on
their own for no apparent reason once in a while anyway, it can be
hard to tell whether or not it's the memory which is at fault.
OK.


So it's always advisable to run a memory testing program for a while
- that way you'll know for sure whether or not the memory works
perfectly. And besides, memtest86 is free (you can download it from
www.memtest86.com, or a slightly different version from
www.memtest.org), so all it'll cost you is a bit of time. Just leave
it running for an hour or so (or maybe even overnight), then check
whether it's noticed any errors. If not, you know for certain the
memory works A-OK.

Thanks for the hints.

Wim

p

 

[Wim Zefat]

Op Sat, 25 Sep 2004 17:21:28 +0200 schreef Peter Wagner:

Thank you for the link.

BTW
Do you know how much the performance will rise, if a machine with a P4-3000
runs in dual-channel instead in single-channel modus?

Again, not as much as you might think. Memory bandwidth goes way up with
dual channel, but on the other hand, latency gets worse. Although the speed
increase is quite large in theory, the actual increase you can expect to
see in everyday use is a lot less. The concensus seems to be that you can
expect somewhere between 2.5 and 5 percent increase in speed of 'real
world' applications. See for instance:

http://www.gamepc.com/labs/view_content.asp?id=ddr400myths&page=5&cookie_test=1

Wim

 

[Peter Wagner]

Op Sat, 25 Sep 2004 17:21:28 +0200 schreef Peter Wagner:

Wim

Thank you.

Again, not as much as you might think. Memory bandwidth goes way up
with dual channel, but on the other hand, latency gets worse.
Although the speed increase is quite large in theory, the actual
increase you can expect to see in everyday use is a lot less. The
concensus seems to be that you can expect somewhere between 2.5 and 5
percent increase in speed of 'real world' applications. See for
instance:

http://www.gamepc.com/labs/view_content.asp?id=ddr400myths&page=5&cookie_test=1

Thanks.
I've asked because my WinXPpro runs in the single CPU mode instead in the
multiprozessor Mode.
The reason is that I've replaced a PIII motherboard by the P4P800-E Deluxe.
WinXP couldn't recognize my mb as a dual-channel.
A new installation of WinXP would be necessary. But this isn't in my
interest. To much work to install all my programms again.

Wim

p

 

[Wim Zefat]

Op Sun, 26 Sep 2004 09:52:48 +0200 schreef Peter Wagner:

I've asked because my WinXPpro runs in the single CPU mode instead in the
multiprozessor Mode.
The reason is that I've replaced a PIII motherboard by the P4P800-E Deluxe.
WinXP couldn't recognize my mb as a dual-channel.
A new installation of WinXP would be necessary. But this isn't in my
interest. To much work to install all my programms again.

Single/dual channel and hyperthreading (multiprocessing) are two different
things.

Single/dual channel has to do with memory usage only. As the name implies,
in single channel you have only one 'channel' between the processor and the
memory. In dual channel (which only works if you have 2 (or 4) identical
DIMM's) you have two, which allows the processor to shovel (nearly) twice
as much data to/from memory in the same time. This obviously makes the
machine faster, though the 'real world' increase in speed is limited.

Take a look at the Memory Tab in CPU-Z. In the top right, after 'Channels'
you can see whether your machine runs in single or in dual channel mode.

Hyperthreading allows a P4 to behave as if you had two processors. The
advantage isn't quite as much as having two *real* processors, but it can
still make a lot of difference. Just how much advantage you get depends
mainly on the applications you use. Some don't benefit at all from
hyperthreading, while others can work up to 30 percent faster with
hyperthreading enabled.

You can tell whether your machine uses hyperthreading by opening Device
Manager and looking under Computer. If it says ACPI Multiprocessor PC,
hyperhreading is enabled. Or look under Processors - if it shows two P4's,
hyperthreading is on. You can also open Task Manager (press Ctrl-Alt-Del)
and look under Performance: if there are two processor-graphs,
hyperthreading is on. If there's just one, hyperthreading is off.

It's generally advisable to do a complete reinstall of Windows if you
change the motherboard, but if you don't want to do this, it should still
be possible to get hyperthreading to work. I've never done this, but I've
seen messages here from people who've gotten it to work. If I remember
correctly, all you have to do is to update the HAL from Device Manager. See
if you can find something on this subject in this group, or ask a new
question, such as 'how to get hyperthreading to work' and somebody will
probably be able to tell you in more detail.

Success!

Wim

 

[Peter Wagner]

Single/dual channel and hyperthreading (multiprocessing) are two
different things.

Single/dual channel has to do with memory usage only. As the name
implies, in single channel you have only one 'channel' between the
processor and the memory. In dual channel (which only works if you
have 2 (or 4) identical DIMM's) you have two, which allows the
processor to shovel (nearly) twice as much data to/from memory in the
same time. This obviously makes the machine faster, though the 'real
world' increase in speed is limited.

Take a look at the Memory Tab in CPU-Z. In the top right, after
'Channels' you can see whether your machine runs in single or in dual
channel mode.

It says 'Dual'.
Performance Mode: 'disabled' ->what's this and there's a way to switch to
'enabled'?

Hyperthreading allows a P4 to behave as if you had two processors. The
advantage isn't quite as much as having two *real* processors, but it
can still make a lot of difference. Just how much advantage you get
depends mainly on the applications you use. Some don't benefit at all
from hyperthreading, while others can work up to 30 percent faster
with hyperthreading enabled.

You can tell whether your machine uses hyperthreading by opening
Device Manager and looking under Computer. If it says ACPI
Multiprocessor PC, hyperhreading is enabled.

Unfortunately it says only 'ACPI-PC'. :-(

Or look under Processors
- if it shows two P4's, hyperthreading is on.

It shows 2 P4's, but I thin HT is off, because...

You can also open Task
Manager (press Ctrl-Alt-Del) and look under Performance: if there are
two processor-graphs, hyperthreading is on. If there's just one,
hyperthreading is off.

It's just _one_. :-(

It's generally advisable to do a complete reinstall of Windows if you
change the motherboard, but if you don't want to do this, it should
still be possible to get hyperthreading to work. I've never done
this, but I've seen messages here from people who've gotten it to
work. If I remember correctly, all you have to do is to update the
HAL from Device Manager.

Yes, I've tried this, but there's only a KB article for Win2000 and not for
WinXP.
One should copy DLLs for Multiprocessors, but not all mentioned DLLs exists
for WinXP?!
And I don't found a KB for an exchange of HAL for WinXP.

See if you can find something on this
subject in this group, or ask a new question, such as 'how to get
hyperthreading to work' and somebody will probably be able to tell
you in more detail.

I've done.
Thank you for the hint.

Success!

I hope so.

Wim

p

 

[Wim Zefat]

Op Sun, 26 Sep 2004 22:11:25 +0200 schreef Peter Wagner:

It says 'Dual'.
Performance Mode: 'disabled' ->what's this and there's a way to switch to
'enabled'?

OK, so your memory does work in dual channel mode.
Mine says 'enabled' - so it is possible

I can't find any information on what this is supposed to mean, but I
suspect it shows whether hyperpath is enabled (also called Performance
Acceleration Technology or PAT). This is a feature of the Intel 875
chipset, which squeezes just a bit more performance out of the memory. The
P4P800-E uses the slightly cheaper 865 chipset, which officially doesn't
have this feature. Unofficially, it's still there though. Intel just
disabled it - but Asus found a way to switch it back on. They're not
allowed to call it Performance Acceleration Technology, so they call it
Memory Acceleration Technology (or MAM) instead.

MAM is disabled by default. Look for Memory Acceleration Technology or MAM
under the advanced chipset-settings in the BIOS. This can be set to Auto
(which means disabled!) or Enabled. Again, if you switch it on, I'd
recommend running memtest86 for a while - you're squeezing more performance
out of the memory and it might not like that.. It's on on my system though
and I've never had a problem with it.

Unfortunately it says only 'ACPI-PC'. :-(

Yup, sorry, that means hyperthreading is off.

It shows 2 P4's, but I thin HT is off, because...

Strange, that. I'd have expected just one to show. Apparently Windows does
recognize the 'second' (virtual) P4, even though it doesn't use it.

Yes, I've tried this, but there's only a KB article for Win2000 and not for
WinXP.
One should copy DLLs for Multiprocessors, but not all mentioned DLLs exists
for WinXP?!
And I don't found a KB for an exchange of HAL for WinXP.

I think you should be able to select ACPI Pc in Device Manager, go to the
Device driver tab, select Update driver and select the right HAL (ACPI
Multiprocessor). Windows should do the rest. Still, if I were you, I'd wait
and see if you get an answer to your new post from someone who has actually
done this - this is something you want to do right the first time, as you
can probably totally screw up Windows with this if you get it wrong ;-(

Good luck!

Wim

 

[Nom]

I think you should be able to select ACPI Pc in Device Manager, go to
the Device driver tab, select Update driver and select the right HAL
(ACPI Multiprocessor). Windows should do the rest. Still, if I were
you, I'd wait and see if you get an answer to your new post from
someone who has actually done this - this is something you want to do
right the first time, as you can probably totally screw up Windows
with this if you get it wrong ;-(

Swapping the HAL within device manager, almost always results in
BlueScreenOfDeath when you next boot the PC - meaning you have to reinstall.

There is a way to do it though, Google for the instructions - I can't
remember from the top of my head.