• Home
  • blog
  • Review: ActiveCool AC4G Thermoelectric cooler

Review: ActiveCool AC4G Thermoelectric cooler

blog image

With Intel CPUs soaring well over the 100W mark, questions about the efficacy of conventional heatsink and forced air cooling arise again. The AC4G system from ActiveCool employs one of the likely alternatives: Thermoelectric Cooling. Thermoelectric Coolers (TEC) are nothing new, but Active-Cool aims to put the technology to work in a new way, by varying the power of the TEC in accordance with demand to keep both the temps and the noise down. Our indepth-report on the ActiveCool AC4G.

April 24, 2004 by Russ
Kinder

Product
AC4G
Thermoelectric cooler
Manufacturer
ActiveCool
Market Price
~US$90

The ever increasing heat output of CPU’s has been a topic of much debate
of late. Each new chip introduction drives the wattage up another notch, only
to be followed soon after by yet another increase. The runaway heat race has
more and more people asking the questions, “How hot is too hot?”
and “Is this as hot as it can get?”

With each generation of chips someone theorizes that we’ve reached the end
of what can effectively be cooled with conventional air cooled heatsinks. While
we probably haven’t reached the maximum yet, clearly the cost and complexity
needed to keep the hottest of CPUs cool with conventional air cooling are greater
and greater. As heat output continues to rise, the need for a more effective
— and cost effective — cooling technique becomes more pressing. With
the Intel Prescott P4s at ~110W and climbing , the need is here and now.

The AC4G
system from ActiveCool
employs one of the likely alternatives to the conventional heatsink and fan:
Thermoelectric Cooling. Thermoelectric Coolers (TEC) are nothing new,
but Active-Cool aims to put the technology to work in a new way, to keep both
the temps and the noise down.

A PELTIER PRIMER

The concept of thermo-electric cooling dates back more than a century. Thermo-electric coolers operate on the
Peltier effect, first documented by Jean Peltier in 1834. Hence their common
name of “Peltier Coolers”. A TEC is essentially a solid-state heat
pump. When a current is applied to the unit heat is transferred from one
side to the other, resulting in a temperature gradient of as much 50°C from one side to the
other. For more technical information, Active-Cool has an excellent explanation of thermoelectric cooling on their site: Understanding
Thermoelectric Cooling

In theory, TECs have a couple of distinct advantages over conventional HSFs:

  • They can cool the CPU to below ambient temperature.
  • They allow the heatsink to be at a higher temperature than the CPU.

For our purposes, the second characteristic is particularly important. With a conventional
air cooler the heatsink is always cooler than the CPU. For a heatsink,
the heat transferred is proportional to the difference between the heatsink
temperature and the air temperature. All other things held constant, doubling
the differential doubles the heat transfer. In other words, getting the heatsink
hotter lets you move more heat (watts) with the same CFM (and noise), thus
making the heatsink perform more efficiently.

Sounds great, more heat transfer, same noise…. but there’s a catch. (Isn’t
there always?) TECs are not perfect conductors of heat. For every watt that
they move from the cold side to the hot side, they must also consume energy,
which is also released on the hot side. The measurement for this is referred
to as a TEC’s coefficient of performance (CoP). The CoP is defined as:
“the amount of heat energy being moved divided by the amount of supplied
electrical power”
. For a typical TEC, the coefficient of performance is
between 0.4 and 0.7. That means that to move 60 watts from the cold side you
will be releasing between 85 and 150 watts from the hot side. That extra 25
to 90 watts works against the efficiency gains made by increasing the temperature
of the heatsink, and there is also the issue of producing that wattage. It has
to be produced by the PSU (with its accompanying efficiency heat loses), and
then be removed from inside the case.

ACTIVE-COOLs unique proposition is that by instantly adjusting the power to the TEC in direct response to demand, even very hot processors can be cooled effectively without creating constant excess heat. Unlike a conventional TEC, the AC4G does not stay at full power all the time, but only when needed with high CPU load. It combines this thermally-controlled variable power TEC with a sophisticated fan controller as well. If everything is executed well, the AC4G promises effective cooling of the hottest processors without the price of constant high fan noise.

Let’s find out how close the Active-Cool AC4G gets to this ideal.

The AC4G-B comes in a nice full-color retail cardboard box, with just enough
techno-marketing-babble on the back to get you interested.

Removed from the box we see the components of the Active-Cool system.

The kit is composed of 2 components, the heatsink/fan assembly and the power supply/control
unit. The kit comes complete with a packet of generic silicon goop and an
AC cord for the power unit.

THE HEATSINK

The heatsink itself is a fairly standard-issue extruded aluminum fin affair,
topped by a 15x70mm fan. According to Active-Cool the fan is rated at 36CFM.
The fins are thick, short, and densely packed. The mounting mechanism is a
standard captive 6-lug clip.

The peltier element can been seen above. Between the peltier
and the CPU is the coldplate, which would otherwise be called the HS base. Its function is to spread the heat from the
CPU over a larger area of the Peltier element. It consists of a fairly beefy
aluminum plate, with a copper slug immediately over the CPU. Bored into the
coldplate is a temperature probe, which feeds its signal into the power/control
unit.

The fit and finish of the heatsink unit is unremarkable. Nothing horrible,
but nothing great either. About on par with a typical stock HSF. The underside
could use with some more finishing to improve its matting surface.

The heatsink has a bundle of wires coming from it that connect to the power/control
unit, and a separate RPM wire to connect to your motherboard’s CPU fan header.
This RPM connector doesn’t actually provide any useful information to the
motherboard; it simply transmits a constant 4600 RPM signal to prevent the
motherboard from protesting when the fan RPM’s are reduced.

POWER & CONTROL UNIT

This is the part of Active-Cool’s system that really separates it from other
peltier coolers. It combines two functions into one PCI-based module: It’s
a PSU to provide power to the TEC, and it contains a controller chip which
regulates the power supplied to both the TEC and the heatsink fan. It can
also control a case fan, through its onboard Molex connector. It uses a separate
AC power connection as a power source for the Peltier unit, to avoid adding
additional load to the PC’s PSU.

Unlike a typical TEC, which runs at full power all the time, the AC4G monitors
and controls the temperature of the cold plate. The danger with a conventional
TEC is that while the CPU is at idle or under low load the temperature of
the cold plate can drop to the point where condensation occurs on it, a potentially
dangerous and damaging situation. The AC4G controls the voltage being fed
to the Peltier unit, ensuring that the cold plate temp never drops below 28°C,
removing the risk of condensation.

The second major function of the control unit is to adjust the speed of the
heatsink fan. The unit starts the CPU fan at 6 volts, stepping it up to 8
or 12 volts if the coldplate plate temperature rises above a critical value.

The functions of the unit are well described by Active-Cool:

The Power and Control Unit is a PC card containing an AC/DC switching
mode power supply and a microprocessor controller. The power supply, which
receives AC input directly from the electric network (through a plug in
the bracket of the card), provides the power to the thermoelectric unit.
The power supply is controlled by the microprocessor.

The microprocessor controller receives input from the ambient temperature
sensor and the CPU temperature sensor. The microprocessor samples the temperature
more than 40 times per second, and adjusts the cooling power of the thermoelectric
unit and the speed of the CPU and case fans accordingly.

To reduce computer noise, the microprocessor normally runs the CPU and
case fans at 6 volts (½ power, drastic reduction in noise). When the temperature
of the ambient air rises, the microprocessor operates the PC case fans at
higher power until temperature is reduced. If the CPU temperature rises,
the microprocessor initiates a carefully orchestrated reaction. First, power
is increased to the (virtually noiseless) thermoelectric unit. If this is
insufficient, then additional power can be supplied both to the thermoelectric
unit and to the CPU fan. When the thermal load is reduced, the fans can
return to quiet operation. During short bursts of processor load, the fans
are often not needed.

Active-Cool advertises that the AC4G can be installed in any PC “in
90 seconds or less”. While I made no attempt to make a race out of it,
they are probably not far off. The heatsink mounts easily, you slide the power
unit into a vacant PCI slot, connect the wiring harness, plug in the AC cord
to the back, and you’re done.

TESTING

The proof is in the test results. The AC4G was tested in two different configurations; one with the stock 70mm
fan, and again with the SPCR reference standard 80mm L1A.

The testbed is as follows:

  • XP2100+ T-bred (1733Mhz) 62.1 watts max
  • Gigabyte GA-7VM400M motherboard with onboard VGA
  • 512 megs of PC2100 RAM
  • Whatever HDD and PSU happened to be lying around. For all subjective
    noise comparisons the HDD was allowed to power down before the analysis.

Notes on Testing Conditions

– Ambient temp was within 0.5° of 22° for the testing period.
– Temps were recorded from the internal diode with Motherboard Monitor 5.
– CPUBurn was used to achieve load temps, idle temps were read while at the
Window’s desktop.
– A standard multimeter was used to measure the voltage being fed to the CPU
fan, the case fan Molex, and to the Peltier unit itself.
– A Kill-A-Watt meter was connected to the Power unit to report on its AC
draw.

Stock Fan (Full voltage)
CPU Temp
°C/W
Power Unit AC Draw
CPU Fan voltage
Case Fan voltage
Peltier voltage
Idle
33°C
9 watts
6.0 v
6.3 v
6.5 v
Load
38°C
0.26
39 watts
6.0 v
6.3 v
14 v
Stock Fan (@ 5v max*)
Idle
35°C
26 watts
6.0v (2.5v@fan)
6.3 v
12.2 v
Load
47°C
0.40
64 watts
12.0v (5.0v@fan)
12.2 v
19.2 v

*For the reduced voltage testing, a Zalman fanmate was placed between
the CPU fan and the power unit and adjusted to the level that would provide
a maximum of 5 volts to the fan. The number in parentheses reflects the voltage
measured
after the Fanmate. The CPU fan must be connected to the power
unit, if it isn’t the unit goes into failure mode and refuses to allow the
machine to boot.

The results at full speed are impressive. The 0.26°C/W is top tier
for conventional CPU coolers, and the fact that the CPU fan is still at
6 volts shows that the AC4G isn’t even breathing hard yet. The problem is
the noise: It is simply out of the question for SPCR. Even at idle the noise is
way beyond what any of us would consider quiet.

The noise has two sources; besides the obvious CPU fan there is also a tiny
fan inside the power/control unit. At 6 volts, the CPU fan is significantly
louder than a Panaflo L1A at 12v. Its noise is raucous and whiny, with a noticeable
clicking component. But it’s really the fan inside the power unit that produces
the most annoyance; full of high pitch whine, rattles, and wind turbulence.
It was so bad that I felt compelled to disassemble the power unit just to see
what was making such a racket.

With the cover off we can see the guts of the Power & Control unit. Essentially
it’s a neatly packaged PSU. The offending fan is quickly identified: A
a 35mm centrifugal blower which exhausts out through the PCI slot. Its purpose
is to keep the PSU components cool by drawing air in under the plastic cover
and across the aluminum heatsinks attached to the MOSFETs. For the remainder
of the testing the fan was disabled (by jamming a plastic cable tie into
its blades) and the plastic cover was left off the power unit to allow the
internals to cool themselves via convection. No instability or overheating
problems were seen, even under the most strenuous of testing.

NOTE: Leaving the cover off during operation is NOT a recommended mod, especially for those with errant
fingers. The heatsinks attached to the MOSFETs are most definitely live, very
much full of electricity that can ZAP you.

At 2.5 volts, the CPU fan spins smoothly, and actually has less clicking that it did without
the Fanmate. Perhaps the Fanmate’s circuitry softens the signal wave from
the PWM? At load, the CPU fan is running at a Fanmate restricted 5V,
and has nearly the same noise characteristics as it does at 6V stock.
Still significantly louder than an Panaflo L1A is at 12 volts. The 0.40°C/W
at load is quite good, but the noise level is still unacceptable.

Of particular
note is the AC load that the unit is drawing from the outlet. At 64 watts
it is nearly as much as the rest of the entire system draws. (80-85 watts)
It is important to remember that in a normal case environment that a large
portion of that extra 64 watts of heat would be being dumped inside
the case, and would result in higher interior ambient temps. Adding 64 watts
is like adding a second CPU, or a very high-end graphics card.

With a Panaflo 80L1A Swap

For this series of tests the SPCR reference Panaflo L1A fan was mounted in place
of the stock 70mm. The same method of using the Fanmate to reduce its voltage
was employed. Allowing the power unit to have full control over
the L1A produced the best results, achieving a balance
between noise and cooling.

Unfortunately the power unit’s use
of PWM to control the fan speed soured the mix by introducing a very noticeable
clicking to the Panaflo. But reducing the noise to even this level is a dangerous
tradeoff. The temperature of the heatsink itself reached over 100°,
well beyond the 80-85° generally cited as the max safe operating temperature
for a peltier element. Continuous use at that load would almost definitely
shorten its life.

L1A reference Fan (Full voltage)
CPU Temp
°C/W
Power Unit AC Draw
CPU Fan voltage
Case Fan voltage
Peltier voltage
Idle
34°C
20 watts
6.0 v
6.3 v
10.1 v
Load
43°C
0.34
64 watts
8.7 v
9.0 v
19.2 v
L1A reference Fan (@ 5v max*)
Idle
50°C
64 watts
12.0v (5.0v@fan)
12.2 v
19.2 v
Load
68°C
0.74
64 watts
12.0v (5.0v@fan)
12.2 v
19.2 v

CONCLUSIONS

The Active-Cool
AC4G is interesting product, and a clear leap forward in the conventional thinking about
peltier cooling… but its noise performance is not yet up to SPCR standards. The AC4G is hampered primarily by its heatsink and
fan. A higher performing HSF combination would surely reduce both temps
and noise. A quieter way to cool the Power & Control unit is also necessary before PC Silencers can fully adapt this product.

An improved Active-Cool package could become successful as Intel’s super-hot >100W Prescott-core P4 processors come on-stream. For Prescott-based PCs, especially in small packages, it may be one of the few practical cooling options.

Personally, I think an interesting experiment would be to attach the huge HeatLane
Zen
heatsink in place of the AC4G’s stock heatsink and fan. The fact
that the aluminum fin and heatpipe structure can be seperated from the copper
base means the peltier element can be easily inserted into place. The increase
in temperature differential would improve both convection cooling capacity and
the conductivity of its heatpipes. (Editor’s Note: Sounds like an itch Russ
has to scratch. We’ll have to find him a Zen HS.)

Pro:
  • Easy installation
  • Interesting adaptation of peltier technology
Con:
  • Expensive. You can achieve similar (or better) performance from a
    conventional HSF, at 1/3rd the price
  • Noisy stock fans.

Much thanks to Active-Cool
for the review sample.

* * * * *

Discuss this this article in the SPCR Forums.

Leave a Comment

Your email address will not be published. Required fields are marked *

Silent PC Review has been providing expert advice and detailed reviews of PCs and peripherals since 2002. Our technical advice has been featured on publications such as: New York Times, O’Reilly, PCMag, Popular Mechanics, Forbes, etc. plus countless trade shows and industry articles. We’re dedicated to providing top-notch advice and reviews for choosing your next PC build.