Peltier Stacking

SUMMARY: Stacking peltiers is fraught with complications and may not be a practical solution to supercooling high-wattage CPUs.

A number of attempts have been made to supercool CPUs by stacking peltiers – placing one unit on top of another – to get very low temperatures. Most who try this are shocked to find that CPU temps are, in fact, high and climbing when this is attempted.

As background on this, first consider the following excerpts from Tellerux, a peltier manufacturer:

How cold can these devices get?

That depends upon a great many things—ambient temperature, the nature of the thermal load, optimization of current delivery to the TE device, optimization of heat sinks etc. It is theoretically possible to get a Delta T (hot side to cold side) of around 75° C working against a THot of 35° C (note that this is the temperature drop across the TE device, itself, and does not include system losses such as the hot side’s temperature rise above ambient).

However, this theoretical maximum only occurs if there is no thermal load—which is not going to happen in a ‘real’ system. In a typical application, you will achieve about half of the theoretical maximum using a single-stage TE device. In order to reach colder temperatures, a multi-stage approach must be employed, either by using multi-tiered Peltier devices, or by using other technologies to create part of the desired Delta T.

For example, you might use a compressor-based system to provide a below-ambient condition for the hot-side of a TE device, then employ Peltier cooling to further reduce the temperature of your load—this is sometimes done to get down into cryogenic levels. It should be noted, however, that TE devices become less efficient at colder temperatures and the ratings for Delta T will be markedly reduced when you operate under extremely cold conditions.

Furthermore, even though multi-stage Peltier devices can achieve greater Delta T’s, they have much less cooling capacity (in terms of watts pumped) than their single-stage counterparts and are far more expensive to produce.

Can I physically stack TE devices to get a greater Delta T?

Yes . . . but it is not so simple as merely stacking two identical Peltier devices, one on top of the other. The critical reality here is that the second device must not only pump the heat from the thermal load plus its own internal power dissipation (I2R), but it also must remove the heat dissipated within the first TE device.

It is usually most sensible from a cost standpoint, therefore, to employ a much smaller device on the first stage than the second. Be aware in stacking modules, however, that the overall heat-pumping capacity (in watts) of the stack will be limited to the throughput of the smallest device.

Source: Tellerux FAQs.

Now let’s consider some peltier math:

Toby at BX Boards featured the following equation:

dTload = (1 – (heat load/max cooling power)) * max temp difference

where heat load = CPU cooling requirement and max cooling power = peltier rating and max temp difference = peltiers max dT under no load.

Let’s see what a single peltier setup for a 60 watt CPU and 72 watt peltier with a Delta T of 65 C can deliver:

This means that with the above setup, the peltier can deliver 10.8 C in cooling. Now don’t take this to mean you’ll drop CPU temps by 10.8 C under ambient. To determine what this will finally do for you, some more math:

Th = Tamb + (C/W)(Qh)
Qh = Pin + Qc
dT = Th – Tc

Th = Peltier hot side temperature
Tamb = Ambient temperature inside your system’s case
C/W = Heatsink efficiency (Degrees C/Watts)
Qh = Total heatsink load (watts)
Pin = Peltier power input (watts)
Qc = CPU heat load (watts)
dT = Total temperature difference between the peltier’s cold and hot side.
Tc = Peltier cold side temperature

Substituting for an ambient temp of 25 C and a heatsink C/W of 0.3 (air-cooled Alpha class):

Th = Tamb + (C/W)(Pin + Qc)
Th = 25 C + (0.3 C/W)(72 + 60 Watts)
Th = 25C + 39.6C = 64.6 C

This is the peltier’s hot side temp; the cold side is then:

Tc = Th – dT
Tc = 64.6 C – 10.8 C = 53.8 C

Hmmm…not what we wanted, is it? This setup is worse than a good air-cooled solution alone! This is the problem facing folks who are trying to cool the AMD beast – the CPU is radiating way more heat than any Intel CPU. In frustration, some have tried peltier stacking (or cascading) expecting twice the cooling. Not so fast.

First off, lets assume two peltiers. What dTload can we expect??

Peltier 1: dTload = ( 1 – (60/72) ) * 65 = 10.8 C
Peltier 2: dTload = ( 1 – (132/72) ) * 65 = -54.2 C

Uh Oh…not good. The second peltier is actually overwhelmed by the first one, because it has to cool the CPU AND the first peltier. This is why many folks are seeing temps rise when they attempt to stack peltiers.

Is there an alternative? Here’s some options:

1. Use a very high watt single peltier – something like 150 watts.
2. Use cold water (0 C) to cool the peltier’s hot side.
3. Use peltiers side by side rather than stacked.

This last option could be real interesting. Because you reduce the heat load on any single peltier, the dTload goes up. The most intriguing setup would be four peltiers side-by-side. This entails some custom work as far as the mechanics, but the payoff could be substantial.

For example, four peltiers along the lines outlined above would give the following dTload:

dTload = (1 – (heat load/max cooling power)) * max temp difference

dTload = ( 1 – (60/288)) * 65 = 51.5 C

Let’s say you use water cooled to 0 C in a water cooling setup with a C/W of 0.1:

Th = 0 C + (0.1 *( 288 + 60)) = 34.8 C
Tc = 34.8 C – 51.5 C = -16.7 C

Now that’s more like it! With a cold side temp of about -17 C, maybe we can get an Athlon to do something extraordinary, although I think all this pays off for Intel’s CPUs a lot better.

Email Joe