Underclocking

[_|_|_]

I'm considering buying a new CPU for an existing P4T-E motherboard,
but I'd like to get a faster CPU so I can swap motherboards later.
If I pick up a Northwood 3Ghz, 800mhz FSB I'd assume that this
could easily be throttled back to 2.6Ghz, 533mhz or 400mhz.
Any problems with this?

 

[Bob Knowlden]

Problems? One or two.

The multipliers have been locked on Intel CPUs since the days of the
"Deschutes" PIII CPUs.

A 3 GHz/800 MHz CPU runs at 15X200 MHz. (The front side bus is "quad
pumped".) I don't know what sort of FSB overclocking is possible with a
board that uses the original Intel 850 (not 850E) chipset, but the board
only officially supports 400 MHz FSB CPUs.

I doubt that you'd want to run a 3 GHz Northwood at 1.5 GHz, even if it
would otherwise run properly.

Sorry that I can't be more encouraging.


Address scrambled. Replace nkbob with bobkn.

 

[no_one]

I would also think that older voltage regulators might not support the newer
CPU's.

 

[_|_|_]

Problems? One or two.

The multipliers have been locked on Intel CPUs since the days of the
"Deschutes" PIII CPUs.

A 3 GHz/800 MHz CPU runs at 15X200 MHz. (The front side bus is "quad
pumped".) I don't know what sort of FSB overclocking is possible with a
board that uses the original Intel 850 (not 850E) chipset, but the board
only officially supports 400 MHz FSB CPUs.

I doubt that you'd want to run a 3 GHz Northwood at 1.5 GHz, even if it
would otherwise run properly.

Sorry that I can't be more encouraging.

I guess Intel never suspected that locking the multiplier might cost
them some CPU sales. Looks like I'd be better off with the existing
CPU (1.6ghz, 400Mhz). I hate to swap the motherboard, cause it
uses Rambus, and I've got 1.5Gigs in it already.

I was hoping to figure some way of incrementally increasing thruput
now and reusing parts in a new motherboard later. I guess it's not
possible.

Thanks

 

[_|_|_]

I would also think that older voltage regulators might not support the newer
CPU's.

Wouldn't older Northwood circuits still be compatible with newer
Northwoods? If not, then I can't even shop for a faster Northwood
400FSB CPU.

 

[Ed]

I guess Intel never suspected that locking the multiplier might cost
them some CPU sales. Looks like I'd be better off with the existing
CPU (1.6ghz, 400Mhz). I hate to swap the motherboard, cause it
uses Rambus, and I've got 1.5Gigs in it already.

I was hoping to figure some way of incrementally increasing thruput
now and reusing parts in a new motherboard later. I guess it's not
possible.

Thanks

Intel pretty much sucks when it comes to swapping parts, they are always
changing cpu/ram sockets, chipsets, cpu specs, same shit with their dual
cores, Intel CPUs should start coming with free motherboards!

 

[Paul]

I guess Intel never suspected that locking the multiplier might cost
them some CPU sales. Looks like I'd be better off with the existing
CPU (1.6ghz, 400Mhz). I hate to swap the motherboard, cause it
uses Rambus, and I've got 1.5Gigs in it already.

I was hoping to figure some way of incrementally increasing thruput
now and reusing parts in a new motherboard later. I guess it's not
possible.

Thanks

But you already have that solution. The 2.8GHz/FSB400/512KB
processor. (I checked in Google, and someone has run one of
those in a P4T-E with BIOS 1008.)

http://groups.google.ca/group/alt.comp.periphs.mainboard.asus/browse_frm/thread/5e9a87aefca1bc36

To a first order approximation, performance is proportional
to processor core clock rate. You get a smaller improvement,
from improving memory bandwidth.

I have a P4 1.8GHz running with PC133 SDRAM. Now, that is
an inadequate memory subsystem, and one I wouldn't consider
upgrading. Your Rambus board is the equivalent of a DDR
single channel board with DDR400 RAM in it, which should be
an adequate platform for a processor upgrade. The 3.2GB/sec
bandwidth of the processor FSB400 interface, exactly balances
the 3.2GB/sec of your dual channel PC800 16bit DIMMs.

Another area that limits older boards, is the method of connecting
the Northbridge and Southbridge. Chipsets that use the PCI bus
to connect those two chips, are sluggish when it comes to I/O. As
long as some faster bus is used between those two chips, you avoid
that limitation. Your hub bus runs at 266MB/sec, so again, the
board is modern enough to be useful.

Your board isn't any worse than say a P4P800S (848 single channel
based) or a P4PE (845 single channel based) board. There is still
life left in it, if you want to use it.

While the $200 for a new processor doesn't make this the cheapest
upgrade in the world, that processor is probably a lot less money
than you've already poured into providing the board with RDRAM
DIMMs.

If you just want to completely upgrade your system, be honest with
yourself, and admit you've got the upgrade bug Nothing wrong
with that.

Paul

 

[_|_|_]

But you already have that solution. The 2.8GHz/FSB400/512KB
processor. (I checked in Google, and someone has run one of
those in a P4T-E with BIOS 1008.)

http://groups.google.ca/group/alt.comp.periphs.mainboard.asus/browse_frm/thread/5e9a87aefca1bc36

To a first order approximation, performance is proportional
to processor core clock rate. You get a smaller improvement,
from improving memory bandwidth.

I have a P4 1.8GHz running with PC133 SDRAM. Now, that is
an inadequate memory subsystem, and one I wouldn't consider
upgrading. Your Rambus board is the equivalent of a DDR
single channel board with DDR400 RAM in it, which should be
an adequate platform for a processor upgrade. The 3.2GB/sec
bandwidth of the processor FSB400 interface, exactly balances
the 3.2GB/sec of your dual channel PC800 16bit DIMMs.

Another area that limits older boards, is the method of connecting
the Northbridge and Southbridge. Chipsets that use the PCI bus
to connect those two chips, are sluggish when it comes to I/O. As
long as some faster bus is used between those two chips, you avoid
that limitation. Your hub bus runs at 266MB/sec, so again, the
board is modern enough to be useful.

Your board isn't any worse than say a P4P800S (848 single channel
based) or a P4PE (845 single channel based) board. There is still
life left in it, if you want to use it.

Good points, well presented. In fact, I have been tempted to upgrade
to a faster 400fsb CPU just to give the system some extra snap. The
problem is that, even with 1.5gb Rambus, I'm running short on ram on
this particular machine. It serves primarily for I/O: Upload/
download, email, burning DVD's and CD's, running web browsers,
networked printers, scanners... I also run all the virus, spyware
checkers on this machine. IOW, a zillion mem-hungry apps. I often
push it over the edge and crash it (maybe a thorough memory test is
in order, too, but it seems very stable with just a few apps running).

I run all the I/O on this machine to effectively keep the development
machines isolated. Ironically, the development machines (KVM'd to the
same monitor) don't have such bottlenecks.

Given the number of USB, Firewire, network and other assorted boards,
I'm also out of card slots. A new motherboard would give me most of
those back.

You have a good point about the efficiency of this particular
motherboard, but by the time I buy more Rambus + CPU, it may end up
expensive. I was looking for an interim fix while I'm throwing more
$$ at the development stations. It doesn't look like the prices of
'old' Northwoods or Rambus are dropping much either.

While the $200 for a new processor doesn't make this the cheapest
upgrade in the world, that processor is probably a lot less money
than you've already poured into providing the board with RDRAM
DIMMs.

I'm trying to remember what I just paid for 2GB of RDRAM. I think it
was $280 for 2GB of DDR400 CAS2.5 Corsair. To even get up to
2GB on my current motherboard I'd probably have to spend more than
that for two 512MB RDRams.

If you just want to completely upgrade your system, be honest with
yourself, and admit you've got the upgrade bug Nothing wrong
with that.

I wish that were the case, Paul. I'm trying to avoid crashes.

Now, for the development stations, I'll own up to it. <g>

LL

 

[_|_|_]

If you just want to completely upgrade your system, be honest with
yourself, and admit you've got the upgrade bug Nothing wrong
with that.

Paul

PS: I think one of the reasons for the crashes is driver stability.
I'd also like to move up to a newer gen of sound hardware. Also need
to do video capture but I'm out of card slots. Just added another
printer, so even the last USB2 port is taken. Now that I think about
it, if I can't reuse the CPU in another board later, it may be the end
of the road for this box.

LL

 

[Paul]

PS: I think one of the reasons for the crashes is driver stability.
I'd also like to move up to a newer gen of sound hardware. Also need
to do video capture but I'm out of card slots. Just added another
printer, so even the last USB2 port is taken. Now that I think about
it, if I can't reuse the CPU in another board later, it may be the end
of the road for this box.

LL

Maybe you should do some testing of your existing board.
Memtest86+ from www.memtest.org can be used to test the
memory. You use the program to format a floppy - it places
a self-booting test program on the floppy, so no OS is needed.
The floppy boots the machine, and tests all of the memory -
the program even moves itself out of the way and "tests
underneath". That is why this program is better than the
average free memory test program.

The second program to use is Prime95 (mersenne.org). That
program runs in Windows, and if you use the "torture test"
option, the processor is run at 100% load. A calculation
with a known answer is done, and if there are any problems
with processor, memory, or a weak power supply, this program
will help find it. It found memory problems here, on a machine
that passed memtest86+.

As for getting another motherboard, I still don't see any
great advantage to socket 775 systems. You can certainly get
faster processors for that socket, but the price of those
processors is outside most people's budget range.

In your case, the S775 systems for the most part, have
PCI Express, and that will hurt your expansion options.
There aren't enough PCI Express cards around yet, to make
PCI Express slots on a motherboard a viable proposition.

That leaves mixed technology boards like the P5P800 - it
has a S775 socket, but the rest of the board uses a previous
generation chipset. The result is you get to reuse your AGP
card and get to buy cheaper DDR memory. There are 5 PCI slots
on that one.

In the S478 boards, there are a whole variety of P4P800 and
P4C800 family boards. Typically there are five PCI slots on
those. The P4G800-V actually has six PCI slots, presumably to
make up for the crappy chips they included on the board. I
think this board was designed for an OEM, and a retail version
was "made from the same mould". This board doesn't appear
to be too common.

http://usa.asus.com/prog/spec.asp?m=P4G800-V<s=01

I would have recommended one of these options earlier, but for
one issue. I don't know if you have heard of the Southbridge
latchup problem, but ICH4/ICH5/ICH5R motherboards are failing
left and right, due to what appears to be static discharge into
their USB ports. What happens, is the Southbridge appears to go
into latchup - a lot of current flows inside the Southbridge
and it gets so hot, it burns up. This is the only public
recognition of the problem - this would apply to your P4C800-E
Deluxe, for example. Hot-plugging USB devices into the
motherboard USB ports is when the problem happens:

http://tw.giga-byte.com/Motherboard/Support/FAQ/FAQ_456.htm

Pictures of what latchup can do to the Southbridge...
http://www.abxzone.com/forums/showthread.php?t=84122&highlight=usb+port

Latchup has even happened to the P5P800 boards, as they use
the ICH5 Southbridge.

What this means, is if you buy a motherboard with a ICH5 or
ICH5R on it, you should continue to use the USB card that
you've got in the P4T-E. By using a separate PCI USB card,
that avoids the risks with using the USB ports on the new
motherboard. I believe there is more to this issue than
we know about, but only time will tell if any of the parties
(Intel or Asus), will give us more information about what the
real problem is. It is unclear to me, whether the motherboards
being repaired under warranty, are still prone to the same
failure mechanism or not, because Asus hasn't admitted there
is a problem on any of their web pages (like their FAQ pages).
I have a P4C800-E Deluxe myself, so this bugs me too.

Paul

 

[_|_|_]

Maybe you should do some testing of your existing board.
Memtest86+ from www.memtest.org can be used to test the
memory. You use the program to format a floppy - it places
a self-booting test program on the floppy, so no OS is needed.

Good idea. Thanks for the reference, Paul.

 

[_|_|_]

Maybe you should do some testing of your existing board.
Memtest86+ from www.memtest.org can be used to test the
memory. You use the program to format a floppy - it places
a self-booting test program on the floppy, so no OS is needed.
The floppy boots the machine, and tests all of the memory -
the program even moves itself out of the way and "tests
underneath". That is why this program is better than the
average free memory test program.

Good idea. Thanks for the reference, Paul.

The second program to use is Prime95 (mersenne.org). That

As for getting another motherboard, I still don't see any
great advantage to socket 775 systems. You can certainly get
faster processors for that socket, but the price of those
processors is outside most people's budget range.

After researching this a bit, I agree. The compelling reason to go
with socket 775 is DDR2, but I don't think that's paying off yet.
Aside from that, I've heard only of the bugs/compat problems
that usually accompany any new chipset.

I think DDR2 will pay off eventually, but for now, I'd rather stay
with 875 chipset.

That leaves mixed technology boards like the P5P800 - it
has a S775 socket, but the rest of the board uses a previous
generation chipset. The result is you get to reuse your AGP
card and get to buy cheaper DDR memory. There are 5 PCI slots
on that one.

I couldn't think of the advantage over the P4C800-E Deluxe.

I don't know if you have heard of the Southbridge
latchup problem, but ICH4/ICH5/ICH5R motherboards are failing
left and right, due to what appears to be static discharge into
their USB ports. What happens, is the Southbridge appears to go
into latchup - a lot of current flows inside the Southbridge
and it gets so hot, it burns up. This is the only public
recognition of the problem - this would apply to your P4C800-E
Deluxe, for example. Hot-plugging USB devices into the
motherboard USB ports is when the problem happens:

http://tw.giga-byte.com/Motherboard/Support/FAQ/FAQ_456.htm

Pictures of what latchup can do to the Southbridge...
http://www.abxzone.com/forums/showthread.php?t=84122&highlight=usb+port

Damn, that's not a subtle failure, is it! I've always been hesitant
to hot-plug those little USB-powered 2-1/2" notebook drives, but this
looks like it goes beyond that. The chips must be very delicate.

And I thought I had found the right motherboard for the next year
or two.

What this means, is if you buy a motherboard with a ICH5 or
ICH5R on it, you should continue to use the USB card that
you've got in the P4T-E. By using a separate PCI USB card,
that avoids the risks with using the USB ports on the new
motherboard.

Thanks for the tip. I don't stay up on this stuff, and probably
would not have noticed the reports. At least the P4C's onboard
Gigabit and audio will give me back a couple slots.

I'll probably hold off on buying another P4C800 until that is known
to be resolved. I hope Intel/Asus own up and fix it.

LL