Water cooling - how many case fans would still be necessary? Tower case recommendations?

Andrew Hamilton wrote:
On Sat, 6 Jun 2009 00:29:18 -0500, "Ken Maltby"
wrote:

So, can anyone tell me if there are watercooling systems
with _ quiet _ radiator fans? I have the _ impression _ that at least
some radiator fans for water-cooling systems are quite noisy.

-AH

Zalman makes a passive cooling system that uses a pump to move the water
to an external reservoir. I believe there is a review of the system on
the Silent PC Review site I mentioned earlier. Zalman claims that the
total noise is below 20Db.

From a thread from three months ago:


One way to look at it, is that you only need waterblocks on
componets that need fans. (One exception is for devices that
are also cooled by the fans installed on adjcent componets,
like the voltage regulators next to the CPU. [ASUS sometimes
provides a special fan for use on the voltage regulators if you
are watercooling the CPU.] You can get waterblocks for the
voltage regulators, as well.)

In reading a LOT about air vs. water cooling, some people say that you
should consider water-cooling "even the MOSFETs." I know what the
acronym stands for, but not what this part actually does on a P6T
series motherboard.

Where is(are) the MOSFET(s) on this board?
http://usa.asus.com/product.aspx?P_I...NhO&templete=2
Select the left thumbnail image from the right side of this page.

Thank you.

-AH

On Thu, 24 Sep 2009 09:59:42 -0700, Andrew Hamilton
wrote:

From a thread from three months ago:


One way to look at it, is that you only need waterblocks on
componets that need fans. (One exception is for devices that
are also cooled by the fans installed on adjcent componets,
like the voltage regulators next to the CPU. [ASUS sometimes
provides a special fan for use on the voltage regulators if you
are watercooling the CPU.] You can get waterblocks for the
voltage regulators, as well.)

In reading a LOT about air vs. water cooling, some people say that you
should consider water-cooling "even the MOSFETs." I know what the
acronym stands for, but not what this part actually does on a P6T
series motherboard.

The MOSFETs being referred to are part of the voltage regulator
circuitry.

Where is(are) the MOSFET(s) on this board?
http://usa.asus.com/product.aspx?P_I...NhO&templete=2
Select the left thumbnail image from the right side of this page.

They are under the two copper-colored heat sinks above and to the
right of the CPU socket. It looks like that board uses a heat-
pipe arrangement to cool the MOSFETS and north/south bridge chips
using the airflow generated by the CPU cooler.

If you are planning to use a fanless CPU cooler of some sort
(water cooling or a large passive heatsink), you'll either need to
provide an alternate source of airflow over those copper heatsinks
or remove the heat-pipe system and water cool all four of the areas
covered by it.

Mike Paff wrote:
On Thu, 24 Sep 2009 09:59:42 -0700, Andrew Hamilton
wrote:

From a thread from three months ago:

One way to look at it, is that you only need waterblocks on
componets that need fans. (One exception is for devices that
are also cooled by the fans installed on adjcent componets,
like the voltage regulators next to the CPU. [ASUS sometimes
provides a special fan for use on the voltage regulators if you
are watercooling the CPU.] You can get waterblocks for the
voltage regulators, as well.)
In reading a LOT about air vs. water cooling, some people say that you
should consider water-cooling "even the MOSFETs." I know what the
acronym stands for, but not what this part actually does on a P6T
series motherboard.

The MOSFETs being referred to are part of the voltage regulator
circuitry.

Where is(are) the MOSFET(s) on this board?
http://usa.asus.com/product.aspx?P_I...NhO&templete=2
Select the left thumbnail image from the right side of this page.

They are under the two copper-colored heat sinks above and to the
right of the CPU socket. It looks like that board uses a heat-
pipe arrangement to cool the MOSFETS and north/south bridge chips
using the airflow generated by the CPU cooler.

If you are planning to use a fanless CPU cooler of some sort
(water cooling or a large passive heatsink), you'll either need to
provide an alternate source of airflow over those copper heatsinks
or remove the heat-pipe system and water cool all four of the areas
covered by it.

MOSFET water blocks. Acrylic top, copper bottom, O-ring seal.
"EK-Mosfet ASUS X58 KIT"

http://www.scan.co.uk/Images/Products/1032958-a.jpg

http://www.scan.co.uk/Images/Products/1032958-b.jpg

Example of original heatpipe assembly.

http://img357.imageshack.us/img357/7585/stocktopej0.gif

http://img357.imageshack.us/img357/6...kbottomvk4.gif

As for what MOSFETs do, see page 11 here for a sample schematic of a VCore.
VCore circuitry surrounds a CPU socket, and provides power at up to levels
of around 100 amps at a low voltage to the CPU core logic. This power design
might have been used on a P4 motherboard.

http://web.archive.org/web/200403310...5ADP3180_0.pdf

A conventional switching regulator consists of the "brains", a chip with more
pins at the bottom of page 11. They try not to make it dissipate too much power,
so the PWM1,PWM2,PWM3,PWM4 signals are buffered with 8 pin chips, before the drive
signals arrive at the MOSFETs. In that example design, there are three
MOSFET transistors per phase (Q1,Q2,Q3 belonging to one phase). The toroid coil
or square cube is L2. Q1 handles the "high side". Q2 and Q3 are "low side"
MOSFETs. Either Q1 is turned on, or Q2+Q3 are turned on. Since Q1,Q2,Q3 have
large gate capacitance (like 3000 pF), the tiny eight pin chip ends up pumping
current at the 1 amp level, in pulses, to charge the gate on each transistor.
It makes the whole process easier to cool, if the chips are split up, and the
8 pin driver is separate. The 8 pin driver should get a little warm on its own.

Q1,Q2,Q3 are the MOSFETs. They're supposed to operate saturated. In other
words, when commanded to conduct, the channel resistance drips to Rds_on,
and a lot of current flows. If Rds_on is low enough, the I squared R loss is
pretty low, so the MOSFET shouldn't get hot. You can make MOSFETs
"bigger", but then the gate capacitance goes up, and the eight pin
driver would start to fry from the load. So the choice of the number
of MOSFETs, their RDS_on, is a trade off between "hot MOSFETs" and
"hot driver chip". The price of the MOSFETs, also plays a part in their
selection.

L1 and C1 thru C6, are the input side filtering. They help prevent the
pulsed operation of the Vcore circuit, from being felt by the power
supply. On the output side, C21 thru C28 is the bank of capacitors
for the output. Capacitors filter the AC component coming from
L2, L3, L4, and ampere level currents will be flowing through C21 thru C28.
The caps have to be rated for some level of "ripple current", to operate
for thousands of hours trouble free. Multiple capacitors are required, to get
enough total ripple current rating. Otherwise, one fat ass capacitor
could have been used, if all they wanted was the capacitance itself.
The capacitance is not quite as critical, as the ripple current rating.

I still haven't found any oscilloscope signal pictures, to demonstrate
what it is doing, so I'll leave the rest of the explanation to the
fine datasheet from Analog Devices.

On page 19, you can see a drawing of the layout of the circuit. The
three MOSFETs on each phase, haven't been place optimally for usage
with a rectangular cooler to fit over the top of them. So the
assumption in that picture, is likely air cooling coming from the
"spill air" from the CPU cooling. You'd need to line the MOSFETs up
in a neat row, to place a cooler over them.

You'll notice, in the figure on page 19, the mention of a thermistor.
On my P4C800-E, Asus didn't include that. If they'd included the
thermistor, the Vcore would have been able to better handle
transitions from "busy" to "idle". The Vcore voltage is off by 0.05
volts for a few seconds, while the copper coils cool off. If
the thermistor had been included, the circuit would have been
temperature compensated, and the voltage error would have been
minimal. So even when you pay for a "Deluxe" motherboard, the design
isn't always as "Deluxe" as it could be. That is what I like, about
digging up these datasheets.

HTH,
Paul

"Paul" wrote in message
...
Mike Paff wrote:
On Thu, 24 Sep 2009 09:59:42 -0700, Andrew Hamilton
wrote:

From a thread from three months ago:

One way to look at it, is that you only need waterblocks on
componets that need fans. (One exception is for devices that
are also cooled by the fans installed on adjcent componets,
like the voltage regulators next to the CPU. [ASUS sometimes
provides a special fan for use on the voltage regulators if you
are watercooling the CPU.] You can get waterblocks for the
voltage regulators, as well.)
In reading a LOT about air vs. water cooling, some people say that you
should consider water-cooling "even the MOSFETs." I know what the
acronym stands for, but not what this part actually does on a P6T
series motherboard.

The MOSFETs being referred to are part of the voltage regulator
circuitry.

Where is(are) the MOSFET(s) on this board?
http://usa.asus.com/product.aspx?P_I...NhO&templete=2
Select the left thumbnail image from the right side of this page.

They are under the two copper-colored heat sinks above and to the
right of the CPU socket. It looks like that board uses a heat-
pipe arrangement to cool the MOSFETS and north/south bridge chips
using the airflow generated by the CPU cooler.

If you are planning to use a fanless CPU cooler of some sort
(water cooling or a large passive heatsink), you'll either need to
provide an alternate source of airflow over those copper heatsinks
or remove the heat-pipe system and water cool all four of the areas
covered by it.

MOSFET water blocks. Acrylic top, copper bottom, O-ring seal.
"EK-Mosfet ASUS X58 KIT"

http://www.scan.co.uk/Images/Products/1032958-a.jpg

http://www.scan.co.uk/Images/Products/1032958-b.jpg

Example of original heatpipe assembly.

http://img357.imageshack.us/img357/7585/stocktopej0.gif

http://img357.imageshack.us/img357/6...kbottomvk4.gif

As for what MOSFETs do, see page 11 here for a sample schematic of a
VCore.
VCore circuitry surrounds a CPU socket, and provides power at up to levels
of around 100 amps at a low voltage to the CPU core logic. This power
design
might have been used on a P4 motherboard.

http://web.archive.org/web/200403310...5ADP3180_0.pdf

A conventional switching regulator consists of the "brains", a chip with
more
pins at the bottom of page 11. They try not to make it dissipate too much
power,
so the PWM1,PWM2,PWM3,PWM4 signals are buffered with 8 pin chips, before
the drive
signals arrive at the MOSFETs. In that example design, there are three
MOSFET transistors per phase (Q1,Q2,Q3 belonging to one phase). The toroid
coil
or square cube is L2. Q1 handles the "high side". Q2 and Q3 are "low side"
MOSFETs. Either Q1 is turned on, or Q2+Q3 are turned on. Since Q1,Q2,Q3
have
large gate capacitance (like 3000 pF), the tiny eight pin chip ends up
pumping
current at the 1 amp level, in pulses, to charge the gate on each
transistor.
It makes the whole process easier to cool, if the chips are split up, and
the
8 pin driver is separate. The 8 pin driver should get a little warm on its
own.

Q1,Q2,Q3 are the MOSFETs. They're supposed to operate saturated. In other
words, when commanded to conduct, the channel resistance drips to Rds_on,
and a lot of current flows. If Rds_on is low enough, the I squared R loss
is
pretty low, so the MOSFET shouldn't get hot. You can make MOSFETs
"bigger", but then the gate capacitance goes up, and the eight pin
driver would start to fry from the load. So the choice of the number
of MOSFETs, their RDS_on, is a trade off between "hot MOSFETs" and
"hot driver chip". The price of the MOSFETs, also plays a part in their
selection.

L1 and C1 thru C6, are the input side filtering. They help prevent the
pulsed operation of the Vcore circuit, from being felt by the power
supply. On the output side, C21 thru C28 is the bank of capacitors
for the output. Capacitors filter the AC component coming from
L2, L3, L4, and ampere level currents will be flowing through C21 thru
C28.
The caps have to be rated for some level of "ripple current", to operate
for thousands of hours trouble free. Multiple capacitors are required, to
get
enough total ripple current rating. Otherwise, one fat ass capacitor
could have been used, if all they wanted was the capacitance itself.
The capacitance is not quite as critical, as the ripple current rating.

I still haven't found any oscilloscope signal pictures, to demonstrate
what it is doing, so I'll leave the rest of the explanation to the
fine datasheet from Analog Devices.

On page 19, you can see a drawing of the layout of the circuit. The
three MOSFETs on each phase, haven't been place optimally for usage
with a rectangular cooler to fit over the top of them. So the
assumption in that picture, is likely air cooling coming from the
"spill air" from the CPU cooling. You'd need to line the MOSFETs up
in a neat row, to place a cooler over them.

You'll notice, in the figure on page 19, the mention of a thermistor.
On my P4C800-E, Asus didn't include that. If they'd included the
thermistor, the Vcore would have been able to better handle
transitions from "busy" to "idle". The Vcore voltage is off by 0.05
volts for a few seconds, while the copper coils cool off. If
the thermistor had been included, the circuit would have been
temperature compensated, and the voltage error would have been
minimal. So even when you pay for a "Deluxe" motherboard, the design
isn't always as "Deluxe" as it could be. That is what I like, about
digging up these datasheets.

HTH,
Paul

Very interesting, Paul. I wonder if motherboard OEMs actually
do all that math. I suspect not, they probably use recommended
reference values, or use software that you just plug your design
parameters into to get the component values.

The missing thermistor, probably a 25 cent item, shows how cheap
these manufacturers are. Personally, I wouldn't mind spending an
extra 10 bucks, if it ensured all component design specs were fully
met. I think nowadays, that besides moderate increased performance
specs, deluxe just means more unneeded bells and whistles, rather
than increased quality.

Andrew Hamilton wrote:
From a thread from three months ago:


One way to look at it, is that you only need waterblocks on
componets that need fans. (One exception is for devices that
are also cooled by the fans installed on adjcent componets,
like the voltage regulators next to the CPU. [ASUS sometimes
provides a special fan for use on the voltage regulators if you
are watercooling the CPU.] You can get waterblocks for the
voltage regulators, as well.)

In reading a LOT about air vs. water cooling, some people say that you
should consider water-cooling "even the MOSFETs." I know what the
acronym stands for, but not what this part actually does on a P6T
series motherboard.

Where is(are) the MOSFET(s) on this board?
http://usa.asus.com/product.aspx?P_I...NhO&templete=2
Select the left thumbnail image from the right side of this page.

Watercoling the mosfets is generally not needed. Thay have a far
better temperature tolerance than the other semiconductors. Hwever
you may want too cool the filter capacitors that are in close
proximity. This can basically be done with air. The second problem
is that most mosfet "coolers" sit on the top of the mosfets, which
is thermally insulated by epoxy resin. The heat oes out the copper
ins on the bottom.

As a consequence the only good waty to watercool mosfets would be
to dismount them from the board and mount them on a copper plate
or the like with their own copper mounting area. This is not
really possible.

To sum up: mosfet coolers are ineffective. However good airflow
on the area lengthens the lifetime of the capacitors there.

Arno