Has anybody ever stacked two CPU coolers?

On 2020-07-27 4:32 a.m., Carlos E.R. wrote:

On 26/07/2020 16.20, Commander Kinsey wrote:
On Sat, 25 Jul 2020 23:01:26 +0100, Wolffan wrote:

On 25 Jul 2020, Commander Kinsey wrote
(in article op.0obycr1swdg98l@glass):

On Sat, 25 Jul 2020 21:52:31 +0100, Â* wrote:

On 25 Jul 2020, Commander Kinsey wrote
(in article op.0oboc2f2wdg98l@glass):


...

Doubling the size of the heatsink (and also the number of fans,
which won't blow towards each other) must make a considerable
difference. It will be twice the cooling effect, minus the lack
of conduction through the first heatsink, which I assume will
conduct well, being metal and all that.

You’re not going to get anywhere close to double the cooling. Instead of
kludging together two fans, just buy one of the properly-designed
dual fan units out there. (which don’t get twice the cooling of
similar single-fan designs from the same company, pros can’t get
double the cooling either. Not even close. Look at the specs for
yourself.) Or, if you’re serious about cooling, use liquid
cooling.

What you _will_ get with a dual-fan system is twice the noise.
Especially if you use small-diameter fans.

Dual fan is pointless.Â* Each fan is designed to push x amount of air,
adding another one in it's path won't speed that air up.Â* You only get
better airflow from faster fans, or fans in parallel.Â* Series is stupid.

Not really. Depends.

https://www.mouser.es/pdfDocs/theparallelandseriesoperation.pdf


There is at least one helicopter with "serial" blades (for another reason).



And various planes, for different reasons, see explantion.

https://en.wikipedia.org/wiki/Contra...ing_propellers

Rene

Paul wrote:
Chris wrote:
Commander Kinsey wrote:

Anyway I fixed it, by changing the fan, I'd forgotten that some fans are
really crap - they can be anything from 1100rpm to 4000rpm. I happened
to have a few 3000rpm fans sitting about, so I replaced what looks like
about 1100rpm with 3000rpm, and lost 20C of temperature (I don't care
about the noise, it's in the garage and operated remotely). What's odd
is I used to have one CPU at 65C and one at 75C (with identical coolers),
and now the 75C has dropped to 55C. I guess the physical position meant
cooler air for one of them.

Right. Fixing the simple things is always better than coming up with
elaborate and convoluted ideas. Occam's razor at work!

The CFM rating. Fans come in four classes.

Low, medium, high, ultra

35CFM 110CFM

A medium is a typical case cooling fan.
I have one ultra here, noisy, removed from machine.

Once the velocity of the air reaches 800 LFPM,
cooling is asymptotic. Jamming an Ultra in may not
be necessary. The Ultra also tends to be thicker
(37.5mm instead of 25mm for regular case fans).

Some setups are poorly designed, and air
leaks out the side. Study the airflow
carefully, to see if the air is being
well-used and guided.

Agreed. I had an old router with vents on the sides which got quite warm.
Once i mounted it sideways allowing the warm air to escape from the top and
cooler air to enter from the bottom it never got warm

As a dumb-ass example, I have some disk enclosures that
have exhaust fans... and no air intake holes :-/

Paul

On 27/07/2020 15.59, Rene Lamontagne wrote:
On 2020-07-27 4:32 a.m., Carlos E.R. wrote:

On 26/07/2020 16.20, Commander Kinsey wrote:
On Sat, 25 Jul 2020 23:01:26 +0100, Wolffan wrote:

On 25 Jul 2020, Commander Kinsey wrote
(in article op.0obycr1swdg98l@glass):

On Sat, 25 Jul 2020 21:52:31 +0100,
wrote:

On 25 Jul 2020, Commander Kinsey wrote
(in article op.0oboc2f2wdg98l@glass):


...

Doubling the size of the heatsink (and also the number of fans,
which won't blow towards each other) must make a considerable
difference. It will be twice the cooling effect, minus the lack
of conduction through the first heatsink, which I assume will
conduct well, being metal and all that.

You’re not going to get anywhere close to double the cooling.
Instead of
kludging together two fans, just buy one of the properly-designed
dual fan units out there. (which don’t get twice the cooling of
similar single-fan designs from the same company, pros can’t get
double the cooling either. Not even close. Look at the specs for
yourself.) Or, if you’re serious about cooling, use liquid
cooling.

What you _will_ get with a dual-fan system is twice the noise.
Especially if you use small-diameter fans.

Dual fan is pointless.Â* Each fan is designed to push x amount of air,
adding another one in it's path won't speed that air up.Â* You only
get better airflow from faster fans, or fans in parallel.Â* Series is
stupid.

Not really. Depends.

https://www.mouser.es/pdfDocs/theparallelandseriesoperation.pdf


There is at least one helicopter with "serial" blades (for another
reason).



And various planes, for different reasons, see explantion.

https://en.wikipedia.org/wiki/Contra...ing_propellers

True, I forgot them.


--
Cheers, Carlos.

On Sun, 26 Jul 2020 18:46:05 +0100, Paul wrote:

Commander Kinsey wrote:
On Sun, 26 Jul 2020 02:45:51 +0100, Paul wrote:

Rene Lamontagne wrote:
On 2020-07-25 5:33 p.m., Paul wrote:
https://www.quietpc.com/tnn500af

That case sure has a lot of good engineering put into it, I can see why
it has casters, it weighs about 70 lbs, I looked around a bit but
couldn't find a price on it yet.
It will be great for audio and such purposes.

I'm not in need of one because one of the perks of old age is a much
reduced sound perception (what did you say). :-)

Rene

That case was $1000, and was a popular item in audio
recording studios.

Just had a quick look, how on earth do you affix the coolers to the CPU,
GPU, etc? Heatpipes aren't flexible. And even worse if the CPU or GPU
isn't in the place the case designer thought it would be.

At this late date, you'd have to find a review where
they did the install. Then you would be able to see
why it works.

For the video card, there's probably a block to be fitted,
and RAMsinks for the memory. That sort of thing.

Presumably assuming very specific positions for each. Or maybe you can bolt the other end to the case in a few different places. For that price there are probably spare pipes of different lengths.

Commander Kinsey wrote:
On Sun, 26 Jul 2020 18:46:05 +0100, Paul wrote:

Commander Kinsey wrote:
On Sun, 26 Jul 2020 02:45:51 +0100, Paul wrote:

Rene Lamontagne wrote:
On 2020-07-25 5:33 p.m., Paul wrote:
https://www.quietpc.com/tnn500af

That case sure has a lot of good engineering put into it, I can see
why
it has casters, it weighs about 70 lbs, I looked around a bit but
couldn't find a price on it yet.
It will be great for audio and such purposes.

I'm not in need of one because one of the perks of old age is a much
reduced sound perception (what did you say). :-)

Rene

That case was $1000, and was a popular item in audio
recording studios.

Just had a quick look, how on earth do you affix the coolers to the CPU,
GPU, etc? Heatpipes aren't flexible. And even worse if the CPU or GPU
isn't in the place the case designer thought it would be.

At this late date, you'd have to find a review where
they did the install. Then you would be able to see
why it works.

For the video card, there's probably a block to be fitted,
and RAMsinks for the memory. That sort of thing.

Presumably assuming very specific positions for each. Or maybe you can
bolt the other end to the case in a few different places. For that
price there are probably spare pipes of different lengths.

There was at least one review that went into details.

I don't think the unit came with variant pipes.
Instead, it's just clever engineering, maybe the
pipe sticks out the side of the block a bit more
or a bit less. It did take a fair amount of "butter",
to butter up the kit while assembling it. The
pipes had to be slathered in material when
screwing down the lid on an assembly.

At one time, the holes for the GPU heatsink were
at known distances from the datum and radial
away from the datum. For cards which were not
bridged (bridge chip causes datum of GPU to shift),
this cooling scheme should have worked. It's
possible an AGP card (with PCIe GPU inside like
an HD3450), the pipes would not have aligned
properly.

Part of the design relied on "thermal seat cushions",
which were a curse. This was a material to allow
some vertical height differences, and there was adhesive
on the top and bottom of the seat cushion item. But
like RAMsink kits, usually the adhesive lets go at
the worst possible time.

Paul

On Sun, 26 Jul 2020 19:50:34 +0100, Carlos E.R. wrote:

On 26/07/2020 16.34, Commander Kinsey wrote:
On Sun, 26 Jul 2020 12:27:38 +0100, Carlos E.R.
wrote:

On 26/07/2020 00.47, Commander Kinsey wrote:
On Sat, 25 Jul 2020 23:33:54 +0100, Paul wrote:

...


This is why we buy a larger heatsink, like a Noctua.
Instead of impishly welding stuff together, the
heatpipes on the properly-designed unit, have been
selected for the perfect location to get the
best effect. While the unit may be designed with
two fans, it works quite well with only the
central fan fitted.

Heatpipes have non-linear behavior. If you pump too
much heat into them, the liquid-vapor phase transfer
stops because all the materials remain in the vapor
phase. The manufacturer will usually suggest a max
power as a practical limit (before the pipes saturate
and the degree-of-cooling ends up actually reduced).
If you put six pipes in parallel, as soon as one
pipe saturates, the other five carry the load and
then they drop in quick succession.

What temperature should the CPU be for optimal heatpipe cooling?

In theory, a heatsink is designed to work best with a certain cpu.

Strange, all the ones I've seen advertised are for any CPU in multiple
types of socket. There must be an optimal temperature that heatpipes
work at.

No, heat pipes work best at certain design temperature, depending on the
boliling point of the fluid. And the temperature specs changes on each
CPU family.

Too hot, and all the fluid is gas: bad transfer.
Too cold, and all the fluid is liquid: bad transfer.

I've always bought generic coolers that say they work with any CPU. Anyway, a certain CPU does not work at a specific temperature, that's up to you and your case cooling. I run mine at 60-70C.

On Mon, 27 Jul 2020 00:26:59 +0100, Wolffan wrote:

On 26 Jul 2020, Commander Kinsey wrote
(in article op.0odr2pn5wdg98l@glass):

On Sun, 26 Jul 2020 21:14:06 +0100, wrote:

On 26 Jul 2020, Commander Kinsey wrote
(in article op.0odggwqpwdg98l@glass):

On Sun, 26 Jul 2020 16:38:33 +0100, wrote:

On 26 Jul 2020, Commander Kinsey wrote
(in article op.0oc9vhsewdg98l@glass):

On Sun, 26 Jul 2020 15:40:34 +0100, wrote:

On 26 Jul 2020, Commander Kinsey wrote
(in article op.0oc80rgowdg98l@glass):

On Sun, 26 Jul 2020 00:41:40 +0100, wrote:

On 25 Jul 2020, Commander Kinsey wrote
(in article op.0ob3ngs7wdg98l@glass):

On Sat, 25 Jul 2020 23:08:09 +0100,
wrote:

On 25/07/2020 23:01, Wolffan wrote:
On 25 Jul 2020, Commander Kinsey wrote
(in article op.0obycr1swdg98l@glass):

On Sat, 25 Jul 2020 21:52:31 +0100,
wrote:

On 25 Jul 2020, Commander Kinsey wrote
(in article op.0oboc2f2wdg98l@glass):

Has anybody ever stacked two CPU coolers? First one fitted
normally,
then
the
second welded on top of it. Theoretically, you have twice the
cooling,

no, you won’t.

if it
conducts through the first fast enough. I'm referring to the
type
with
the
fan on the side.

there are numerous sites out there which cover this; I first
encountered
this
idea in 2001. (Yes, 2001. Not a typo. And it wasn’t a new idea
then,
it
was
just new to me.) It turns out that doubling the fans gives from
-5%
(that’s
_negative_ five percent, it can make things worse) to +65%
airflow,
depending
on numerous factors. (Size of fan, are the fans spinning in the
same
direction or in opposite directions, fan speed, how fans are
placed
relative
to the heat exchanger, space inside the computer’s case,
location
of
CPU
relative to sides of computer case, location of other items
including
hard
drives, RAM, power supplies, video and/or sound cards, type of
case
especially with respect to vents, internal cabling, more.)
However,
airflow
and cooling do not necessarily scale together.. Merely increasing
the
airflow
does not necessarily do much for the cooling, for that you need
heat
exchangers attached to the CPU, heat exchangers attached to
another
heat
exchanger don’t do much, and airflow is a square effect while
heat
exchangers are a cube effect. Airflow maxed out at around +65%
for
two
fans;
cooling maxed out at about +45%. And it took considerable effort
to
get
that
+45%, most systems were considerably lower than that. There would
be
a
_reason_ why liquid cooling is popular. It’s actually simpler
and
easier
to
do. Most modern CPUs generate a whole lot less heat than CPUs
dating
from
2001 did, so a good way to get a cooler-running system is toi
use a
newer
CPU. And to stay way from the heat-producing monsters, notably
IBM
Power
CPUs, which have improved no end since then but which are still
devices
to
convert electricity into heat, something anyone who ever went
near
a
G5-
or
a
G4- powered Mac could tell you. (One G4 Mac was called the
Windtunnel
because
of the noise level from the fans to keep the CPU cool; one G5 Mac
shipped
with liquid cooling because Apple didn’t want to ship
Windtunnel
Mk
2...)
IBM is up to Power 8 or 9 or some such now, where G5s were Power
5s,
and
they’re much better behaved, but still...

Doubling the size of the heatsink (and also the number of fans,
which
won't
blow towards each other) must make a considerable difference. It
will
be
twice the cooling effect, minus the lack of conduction through the
first
heatsink, which I assume will conduct well, being metal and all
that.

You’re not going to get anywhere close to double the cooling.
Instead
of
kludging together two fans, just buy one of the properly-designed
dual
fan
units out there. (which don’t get twice the cooling of similar
single-fan
designs from the same company, pros can’t get double the cooling
either.
Not even close. Look at the specs for yourself.) Or, if you’re
serious
about cooling, use liquid cooling.

What you _will_ get with a dual-fan system is twice the noise.
Especially
if
you use small-diameter fans.

Why are you talking down to a Computer Science graduate, Wolffan?

Apple *FAN* BOISE have no experience of such matters. :-P

You are way out of your league here.

Are you not an Apple fan (can't call you a boy at your age) yourself?

he’s an idiot.

Apart from believing in god and using Apples, no he isn't.. He's just
down
the
road from me (well 500 miles).

He attempted to post authoritatively about IPs, when he not merely
didn’t
know the difference between IPv4 and IPv6, he _thought that x.x.x.x was
a
valid IP address_. Yes, really. He’s an idiot.

I guess using Apples he never encounters IPs. All the important stuff is
hidden from you on the mickey mouse machines.

Nope. Ips are quite visible in the Network pane of System Preferences.
There
are panes for setting your IPv4 and v6 addresses, for DNS, proxies,
802.1X,
even for playing with the MAC. Ethernet has one set of panes, 802.11
wireless
another, there are others for other types of network connections. He’s
sufficiently idiotic and pig-ignorant that he he’d bother to go to his
Apple menu, select System Preferences, click on Networks, click on the
network connection of his choice, and have a look. And, when his idiocy
was
pointed out to him, he denied it, and when evidence of his idiocy was
provided (a MID and a quote from his post, IIRC) he ran away. Nor is this
the
only example of his idiocy, just an example easily shown. Most good news
clients have a built-in search engine; in the Hog, I can search for and
find
any string in any post on any newsgroup in my system list. Finding
‘x.x.x.x’ in a.c.w, or any other newsgroup, is trivial. I don’t know
if
your client has similar abilities. He has perpetuated many idiocies over a
prolonged period. He’s an idiot.

And so are you for taking so much time proving he's wrong.

nope.
Get a hobby.

this is a hobby.

Do I have to spell it out for you?

you can try.
Get a useful or interesting hobby

this is a useful, to mer, and amusing, to me, hobby. Part of its use and part
of its amusement comes from the reactions of troll-boy’s fan club. I find
_that_ to be worth all the trouble and more.
instead
of being a sad lonely stalker.

Som, I don’t have to stalk hium, I just have to be reading any of the
newsgroups where he makes a spectacle of himself. And then to just hit save
when I see a particularly idiotic post. and not just one of troll-boy’s
posts.

Just so you know, I just hit save. You’re funny, boyo.

Get a life.

On Mon, 27 Jul 2020 10:14:48 +0100, Paul wrote:

Chris wrote:
Commander Kinsey wrote:

Anyway I fixed it, by changing the fan, I'd forgotten that some fans are
really crap - they can be anything from 1100rpm to 4000rpm. I happened
to have a few 3000rpm fans sitting about, so I replaced what looks like
about 1100rpm with 3000rpm, and lost 20C of temperature (I don't care
about the noise, it's in the garage and operated remotely). What's odd
is I used to have one CPU at 65C and one at 75C (with identical coolers),
and now the 75C has dropped to 55C. I guess the physical position meant
cooler air for one of them.

Right. Fixing the simple things is always better than coming up with
elaborate and convoluted ideas. Occam's razor at work!

The CFM rating. Fans come in four classes.

Low, medium, high, ultra

35CFM 110CFM

Most I see are quoted in rpm, but I guess that's a good guide.

A medium is a typical case cooling fan.
I have one ultra here, noisy, removed from machine.

Once the velocity of the air reaches 800 LFPM,
cooling is asymptotic. Jamming an Ultra in may not
be necessary. The Ultra also tends to be thicker
(37.5mm instead of 25mm for regular case fans).

Some setups are poorly designed, and air
leaks out the side. Study the airflow
carefully, to see if the air is being
well-used and guided.

As a dumb-ass example, I have some disk enclosures that
have exhaust fans... and no air intake holes :-/

What I do with my tower cases is have all the fans as intake, and leave off the covers at the back where the expansion cards aren't, to allow the hot air out. Way better cooling, since the fans are blowing onto the hot CPU etc instead of just feebly pulling air away from them. Plus an exhaust fan and an intake fan achieves nothing, they're both moving the same air.

On Mon, 27 Jul 2020 19:12:00 +0100, Chris wrote:

Paul wrote:
Chris wrote:
Commander Kinsey wrote:

Anyway I fixed it, by changing the fan, I'd forgotten that some fans are
really crap - they can be anything from 1100rpm to 4000rpm. I happened
to have a few 3000rpm fans sitting about, so I replaced what looks like
about 1100rpm with 3000rpm, and lost 20C of temperature (I don't care
about the noise, it's in the garage and operated remotely). What's odd
is I used to have one CPU at 65C and one at 75C (with identical coolers),
and now the 75C has dropped to 55C. I guess the physical position meant
cooler air for one of them.

Right. Fixing the simple things is always better than coming up with
elaborate and convoluted ideas. Occam's razor at work!

The CFM rating. Fans come in four classes.

Low, medium, high, ultra

35CFM 110CFM

A medium is a typical case cooling fan.
I have one ultra here, noisy, removed from machine.

Once the velocity of the air reaches 800 LFPM,
cooling is asymptotic. Jamming an Ultra in may not
be necessary. The Ultra also tends to be thicker
(37.5mm instead of 25mm for regular case fans).

Some setups are poorly designed, and air
leaks out the side. Study the airflow
carefully, to see if the air is being
well-used and guided.

Agreed. I had an old router with vents on the sides which got quite warm.
Once i mounted it sideways allowing the warm air to escape from the top and
cooler air to enter from the bottom it never got warm

I bought a 1kW 12V supply to power 4 large GPUs. It came with a 40mm very loud fan on the back, which lasted 2 months, probably because it was stupidly extracting the hot air, therefore melting its own bearings. I sawed a great big hole in the top of it and fitted a 120mm intake fan. Silent and cooler.

On Mon, 27 Jul 2020 08:05:23 +0100, Chris wrote:

Commander Kinsey wrote:
On Sun, 26 Jul 2020 11:16:11 +0100, Chris wrote:

Commander Kinsey wrote:
On Sat, 25 Jul 2020 23:33:54 +0100, Paul wrote:

Commander Kinsey wrote:
Has anybody ever stacked two CPU coolers? First one fitted normally,
then the second welded on top of it. Theoretically, you have twice the
cooling, if it conducts through the first fast enough. I'm referring to
the type with the fan on the side.

Call me, when your solution uses heatpipes.

It does. Two coolers, each with their own 4 heatpipes.

Look at the Zalman completely-passive cooling
computer case, if you want to see how to do it.
It moves heat from one place to another with
heatpipes. The exterior of the case is the
final destination.

https://www.quietpc.com/tnn500af

No no and no. I once bought a "silent power supply" that threw it's heat
out the back with a sticky out heatsink. It went bang after a few months.

Only with heatpipes, can you move heat from
the origin, to the second heatsink, and expect
to get improved performance.

And both the heatsinks have heatpipes running through them.

That's why heatpipes are used, between baseplate
and top-of-fins, so that an excessively-long fin
does some work for you.

It's also why a 50mm tall 40mm square heatsink,
is no more effective than a 25mm tall 40mm square heatsink.
Those are used on chipsets. The upper portion of the
taller unit, has insufficient thermal conduction
for the top of the fins to remove heat. It means
making fins longer and longer (without heatpipes),
is asymptotic. At some point, a longer fin adds
nothing to the solution.

I can't believe that, metal is a very good conductor.

But none is superconducting so there'll always be efficiency losses over
distance. Not all metals are the same which is why sinks are a mix of
copper and aluminium and not made of steel. Any joins also reduce
efficiency. It all adds up to a bad idea.

It was a very short distance from the heatpipes of the bottom one to the
heatpipes of the top one.

The gap could be micrometers, but if it's effectively an insulator (e.g. an
air gap), then there's no benefit.

The gap was horizontal, and made of a metal heatsink fin.

The top of tall ones can still be hot to touch.

Once you lace the parts together with heatpipes,
then we'll talk. There's a limit to how many
heatpipes can be fitted, and some thought has to be
put into where the heatpipe is moving the heat.

I would have 4 heatpipes running through the bottom half, and 4 through
the top half. There would only be a short interface of normal metal to join the two.

You place the heatpipe at the source, as only if
the heatpipe is in intimate contact, does its 1000x
better performance, work. Just ramming a pipe into
a block with 3 inches of aluminum between the
baseplate and the pipe, that would be useless too.
As the 3 inches of aluminum is a huge thermal
resistance, and the zero resistance of the heatpipe
on top cannot then help. The heatpipes always start
in the baseplate, touching the CPU.

Doesn't have to. The top of the existing heatsink is too hot to touch.
So the heatpipes on the new heatsink will take heat away from there..

Compare it to the temperature of the CPU. It's probably already lost 30-40
°C. Another sink would lose that again.

I touched the block at the bottom where the heatpipes come from and the
temperature was very similar.

Anyway I fixed it, by changing the fan, I'd forgotten that some fans are
really crap - they can be anything from 1100rpm to 4000rpm. I happened
to have a few 3000rpm fans sitting about, so I replaced what looks like
about 1100rpm with 3000rpm, and lost 20C of temperature (I don't care
about the noise, it's in the garage and operated remotely). What's odd
is I used to have one CPU at 65C and one at 75C (with identical coolers),
and now the 75C has dropped to 55C. I guess the physical position meant
cooler air for one of them.

Right. Fixing the simple things is always better than coming up with
elaborate and convoluted ideas. Occam's razor at work!

You should have seen my maths exam papers. I could solve a problem in 50 steps that everyone else only took 5.

On Mon, 27 Jul 2020 10:32:44 +0100, Carlos E.R. wrote:


On 26/07/2020 16.20, Commander Kinsey wrote:
On Sat, 25 Jul 2020 23:01:26 +0100, Wolffan wrote:

On 25 Jul 2020, Commander Kinsey wrote
(in article op.0obycr1swdg98l@glass):

On Sat, 25 Jul 2020 21:52:31 +0100, wrote:

On 25 Jul 2020, Commander Kinsey wrote
(in article op.0oboc2f2wdg98l@glass):


...

Doubling the size of the heatsink (and also the number of fans,
which won't blow towards each other) must make a considerable
difference. It will be twice the cooling effect, minus the lack
of conduction through the first heatsink, which I assume will
conduct well, being metal and all that.

You’re not going to get anywhere close to double the cooling.. Instead of
kludging together two fans, just buy one of the properly-designed
dual fan units out there. (which don’t get twice the cooling of
similar single-fan designs from the same company, pros can’t get
double the cooling either. Not even close. Look at the specs for
yourself.) Or, if you’re serious about cooling, use liquid
cooling.

What you _will_ get with a dual-fan system is twice the noise.
Especially if you use small-diameter fans.

Dual fan is pointless. Each fan is designed to push x amount of air,
adding another one in it's path won't speed that air up. You only get
better airflow from faster fans, or fans in parallel. Series is stupid.

Not really. Depends.

https://www.mouser.es/pdfDocs/theparallelandseriesoperation.pdf

In a PC, the pressure is virtually zero. So you're not helping by putting two in series.

There is at least one helicopter with "serial" blades (for another reason).