80mm vs 92mm Fan?
 

Hi there,



I am trying to decide what size fan to buy for my current P4 build. I am
not sure whether a 80mm or 92mm is best. AT first I thought 92mm was the
way to go but after having read a few reviews around the web it seems like
the 80mm fans are giving better performance due to their smaller *hub* size,
which reduces the size of the dead-zone.



I am also trying to decide what Heat-Sink to use also, it's down to the
ThermalRight SP-94 Swiftech MCX478-V which are both good solutions, so I'm
not sure whether I need a *Tornado* Type solution or go with a quiet
version.



I intend to overclock my P4c 2.6GHz to 3.2GHz +



Anyone had experience of the different sized fans?



Despite the possible larger *Dead-Zone* of the 92mm fans, I wondered also if
their larger size could help keep the motherboard components near the CPU
(MOSFETs) a little cooler?



Thanks for any feedback. .
--
Wayne ][
Sign on door reads: Please Do No Disturb! Pentium 4 assembly in progress!

Wayne Youngman wrote:
Hi there,



I am trying to decide what size fan to buy for my current P4 build.
I am not sure whether a 80mm or 92mm is best. AT first I thought
92mm was the way to go but after having read a few reviews around the
web it seems like the 80mm fans are giving better performance due to
their smaller *hub* size, which reduces the size of the dead-zone.



I am also trying to decide what Heat-Sink to use also, it's down to
the ThermalRight SP-94 Swiftech MCX478-V which are both good
solutions, so I'm not sure whether I need a *Tornado* Type solution
or go with a quiet version.



I intend to overclock my P4c 2.6GHz to 3.2GHz +



An 80mm 55dB 80CFM 5700RPM is deafening, period.

Using the 92mm Vantec Tornado on the Thermalright SP-94, P4 2.6 at 3.4GHz,
this reviewer found 54C load and 39C idle temperatures.
http://www.3dgameman.com/vr/thermalr...eo_review.html

This next review uses the 92mm Panaflo H1 to cool a P4 at 3.18GHz to 47C
load.
http://www.gruntville.com/reviews/he...sp94/page4.php

I'd opt for the 92mm Panaflo
http://sidewindercomputers.com/pa92hisp.html
http://sidewindercomputers.com/pa92mesp.html


BTW, the stock heatsink can often keep a P4 at 3.2GHz under 50C(load) if
case airflow and more importantly, ambient temperature, is ~20C.

60mm, 80mm & 92mm fans usually have the same hub size.

Choose a good heatsink like the SLK, then either size fan.
However, keep cfm scaled proportional to thermal load:
o A 25cfm fan will cool 125W of CPU
o A 43cfm fan will cool 215W of CPU

So either is beyond the hottest P4-Prescott - and L spec Panaflo.

Instead ensure you are removing hot air from the CPU cooler area
rather than recirculating it. If your CPU-cooler recirculates 50% of
its air then a 75W CPU that requires 15cfm now needs 30cfm. That
is easiest with an exhaust fan near the CPU on the rear of the case.

You don't need screamers, just design a system level solution.
--
Dorothy Bradbury
www.stores.ebay.co.uk/panaflofan for fans, books & other items
http://homepage.ntlworld.com/dorothy...ry/panaflo.htm (Direct)

Dorothy Bradbury wrote:

60mm, 80mm & 92mm fans usually have the same hub size.

Choose a good heatsink like the SLK, then either size fan.
However, keep cfm scaled proportional to thermal load:
o A 25cfm fan will cool 125W of CPU
o A 43cfm fan will cool 215W of CPU

What are the rest of the parameters in those CFM calculations? Like
heatsink efficiency, ambient, and target temp?

So either is beyond the hottest P4-Prescott - and L spec Panaflo.

Instead ensure you are removing hot air from the CPU cooler area
rather than recirculating it. If your CPU-cooler recirculates 50% of
its air then a 75W CPU that requires 15cfm now needs 30cfm. That
is easiest with an exhaust fan near the CPU on the rear of the case.

You don't need screamers, just design a system level solution.

A 25cfm fan will cool 125W of CPU? Well so can a 5 cfm fan. Question is,
is it enough. It's pretty easy to figure out if your statement could
possibly be true.

The heat capacity of air (STP) is ~ 24 Watt seconds per cu. ft. per K, so a
1 k rise would require, for a 125 Watt CPU dissipation, 125/24 * 60 ~ =
300 cu. ft. per minute, or 30 cfm for a 10 K rise. But then there is the
thermal resistance of the spreader plate, the heatsink base, and the fin
interface... even pure copper is going to have a significant thermal
resistance for the die size and thermal spreader size of a Pentium 4
Prescott. So say case air temperature is a balmy 30 degrees C; add 15
degrees C for all thermal resistance, then the 10 degrees C for the air
temperature rise, and you get 30 + 15 + 10 = 55 C CPU temperature... not
what overclockers would be thrilled with. A dissipation level of 215 Watts
with the same die size and thermal spreader size would end up with a CPU die
temperature of more like 65 degrees C with the same input air temperature
and a 43 cfm fan. Within Intel specs, but on the edge of throttling, and
again, not thrilling to overclockers. Real world results will be even
worse. So I have to doubt the usefulness of your pronouncement.

--
Phil Weldon, pweldonatmindjumpdotcom
For communication,
replace "at" with the 'at sign'
replace "mindjump" with "mindspring."
replace "dot" with "."

"Dorothy Bradbury" wrote in message
news:8A19c.14$le4.5@newsfe1-win...
60mm, 80mm & 92mm fans usually have the same hub size.

Choose a good heatsink like the SLK, then either size fan.
However, keep cfm scaled proportional to thermal load:
o A 25cfm fan will cool 125W of CPU
o A 43cfm fan will cool 215W of CPU

So either is beyond the hottest P4-Prescott - and L spec Panaflo.

Instead ensure you are removing hot air from the CPU cooler area
rather than recirculating it. If your CPU-cooler recirculates 50% of
its air then a 75W CPU that requires 15cfm now needs 30cfm. That
is easiest with an exhaust fan near the CPU on the rear of the case.

You don't need screamers, just design a system level solution.
--
Dorothy Bradbury
www.stores.ebay.co.uk/panaflofan for fans, books & other items
http://homepage.ntlworld.com/dorothy...ry/panaflo.htm (Direct)

Interesting replies, but missing the context.

Lowest rule-of-thumb design:
o Basic guideline for cooling a 1500W heat source is 300cfm
o Lowest denominator guideline, cited by PAPST-EBM

Next level up of design:
o Taking the guideline, adjust it according to implementation
---- thermal input - what %age duty cycle the CPU does (assumed 100%)
---- air recirculation - what %age the CPU-cooler recirculates (assumed 0%)
o Taking the assumptions in Isolation obviously need adjustment
---- CPU duty cycle of 60-70% is more typical
---- air-recirculation of 30-40% is more typical (impingement heatsink)
o So the assumptions in Combination are somewhat moot
---- hence the guideline of 1500W heat source requiring 300cfm

If a CPU is overclocked, more consideration is needed.

Next level up of design:
o Id thermal solution objective requirements
---- id thermal load (100w)
---- id maximum temp delta (max-desired-temp - max-ambient-temp)
-------- eg, 52oC max, 35oC ambient, required temp-delta is 17oC
o Id thermal solution subjective requirements
---- overclocking = risk of hot-spots for CPU heatspreaders/cores
---- avoid hotspots = all copper base for low thermal gradient (Flower)
---- get heat to fins = heatpipes to fins using phase-change (SLK)
o Id heatsink meeting objective requirements for a given P-Q fan characteristic
---- Sufficiently low oC/W to not exceed your temp-delta given wattage
---- eg, 0.17oC/W for a 100W load with 17oC max rise
o Id estimate of heatsink air recirculation
---- heatsink data is for free air, integrated heatsinks recirculate air
-------- eg, impingement heatsinks recirculate 40%, but can be more or less
---- id your heatsink air recirculation
-------- eg, case-mounted fan behind heatsink reduces air recirculation
-------- eg, PSU intake port behind heatsink reduces it somewhat less
---- id resulting adjustment factor to use
o Id appropriate fan solution for revised oC/W
---- heatsink oC/W for CPU use is quoted with a particular fan cfm
-------- ideally showing relevant P-Q fan characteristic curve
---- adjust fan choice in light of required oC/W
-------- with view to testing resulting choice empirically :-)


Progression is from rules-of-thumb thro to spreadsheet modelling, with
more detailed meshing of assumptions by simple model & optimisers.

Next level up of design:
o Id resistance to airflow
---- in terms of Pa or mm-H2O to airflow to intended airflow range
---- eg, airflow resistance rises rapidly with airflow cfm
o Id required fan to maintain cfm against resistance
---- from fan characteristic curve re P-Q data
---- intending to keep the fan in the optimum position of the curve

Progression now is into empirical testing.
Here too, some basic rules-of-thumb apply - the ATX specification case
has 82-84% of the entire case resistance as the PSU exhaust fan grill.


The trend is clear - migration upwards to system level cooling design.


Reality is different - none of the above are really used now.
o System level thermal design is done at as early a level as possible
---- virtual modelling, minimal empirical testing
---- avoid design changes after component/layout in progress
o Design in reliability by predicting reliability of new tech
---- allows adoption of new tech, closer design to limits
---- MIL spec out, IEEE-1998 & 1999 reliability prediction stds in

So reality is post-it-note # design concepts, some rules of thumb
to narrow-down the likely design component choices, real-time assess
each design with a simple 2D block-blob finite network model, then
on to a more detailed 3D CFD/FEA model & assess hot-spots, shadowing,
refine selection of components & minimise costing thro to target cost.
Reason is watts - from PCB vias to MCM, BGA & compact thermal models.


Individual PC component buyers are OTS, Off-The-Shelf, integrators
who must create a thermal solution from very disparate components.
As such, 300cfm will cool 1500W with estimate of overdesign needed.
That overdesign becomes more detailed as one moves into specific
application needs - such as lower noise levels or overclocking.

For example, an 80mm fan may cool a heatsink well for overclocking.
However, in overclocking one has a system chain to consider:
o CPU VRM modules run hot and require some chassis cooling
---- hence P4-Prescott in *open air* had VRM capacitors at 235F
o Overclocking usually involves higher Icore & Vcore than normal
---- which involves more watts not just on the CPU, but VRM set
o Capacitors may be rated to 105oC, but not always low E-S-R
---- so VRM cooling can be important also

Hence a larger fan blowing over associated components can help,
but at the same time you can lose pressure over the CPU heatsink
as well as airflow - so impact on cooling there deleteriously.

Takes a bit of effort. Hub size is more critical in terms of the heatsink design
under it - generally the heatsink should be height layered down into hub centre,
helps a bit - cost $$. So if it's a choice of say heatpipes or layering, former wins.

That's a quickie.
--
Dorothy Bradbury
www.stores.ebay.co.uk/panaflofan for fans, books & other items
http://homepage.ntlworld.com/dorothy...ry/panaflo.htm (Direct)

A quickie, opaque, and off point.

--
Phil Weldon, pweldonatmindjumpdotcom
For communication,
replace "at" with the 'at sign'
replace "mindjump" with "mindspring."
replace "dot" with "."

"Dorothy Bradbury" wrote in message
news:4zm9c.1118$Pd7.946@newsfe1-win...

A quickie, opaque, and off point.

No, that would adequately qualify your reply above.
--
Dorothy Bradbury