Overclocking a... peltier?

[satanicoo]

Lets just imagine, that i buy some peltier (example: an swiftech mcw50-t)
Imagine that it runs at 12v and it is 80 watts.

If i run it at 17, what are the chances of it burning out?
if 17 volts is too much, how much should it be?
And, finally, is it even possible\worth?

[BalefireX]

I'm more interested in "underclocking" peltiers - using lots of them at 3v or 5v for a waterchiller, so I thought I'd throw in this question as it is on a similar subject: how do you figure out how many, what type, and at what voltage you need to remove a set number of watts?

[bigben2k]

Odds of a burnout are high. Better to use a bigger pelt, and undervolt it.

Pelt efficiency in a chiller, guesstimating at 60% (ha!) will determine the number and size.

[satanicoo]

hmmm.... everyone calls me crazy for running a WC barton at 2.2 volts...
I think i will be crazy to test this.

who knows, it can even be a pleasant surprise ( like the kryo on the P4... )

EDIT: pelt eficiency is better at high or low V?

[Groth]

The max voltage rating on a pelt refers to the voltage at which it transfers the maximum heat. If you go higher, the additional resistive heating will overwhelm the additional cooling; more power spent doing less net work.

[satanicoo]

oh my i guess ill chill my thinking then.

[HAL-9000]

Originally Posted by satanicoo

Lets just imagine, that i buy some peltier (example: an swiftech mcw50-t)
Imagine that it runs at 12v and it is 80 watts.

If i run it at 17, what are the chances of it burning out?
if 17 volts is too much, how much should it be?
And, finally, is it even possible\worth?

The Swiftech MCW50-T's peltier actually runs at 15.2V Umax and 8.1amps Imax to generate the Q value advertised at 80 watts. The 80 watt spec quoted on the page is actually a little bit of hyperbole. If you run it at twelve volts, even at 8 amps (even though it won't support that current at the lower voltage) you would actually get ~60 watts of effective thermal transfer.

Look at Swiftech's own chart for the T's performance at 12 volts...it doesn't even spec a 80 watt heatload because that pelt is a heater at 12 volts with a 80 watt heatload. At 25c where the DT between the sides is zero, maximum Qmax is reached at ~54 watts.

If you "overvolted" the MCW50-T to ~15 volts, it would perform better than what's indicated on Swiftech's chart, assuming the waterblock itself can hold the hot side to 25c. The hot side temp is important with pelts, because Qmax is achieved when hot and cold sides of pelt are identical. More realistic figure for hot-side in real world w/c apps is more like 35c.

[satanicoo]

Oki, thanks for the clarification.
Guess i'll just stick another psu to my machine, and use the 3.3 volt rail with the 12 volt from the other to make 15.3 volts. Evil!

[Butcher]

Maximum efficiency for eplts is usually at around 75% of Vmax.

[satanicoo]

Guess ill just stick with 12 volts then??

EDIT:
nvm... ill just test it with 12, 15.3 and 17 volts, and then put the results.
burnt pelt anyone?

[HAL-9000]

Originally Posted by Butcher

Maximum efficiency for eplts is usually at around 75% of Vmax.

Maximum efficiency for pelts is around 75% Imax, or the current. Not the voltage. Current is driven by DT temps, hence whatever config you run regarding hot vs cold side is critical for temps on the chip. CPU chips have gotten hotter, to the point where it is no longer cost effective to go pelt for extreme cooling. Too much electricty needed and cost needed for too little of a real-world result in temps. The same thing is happening with GPU's. They are getting so hot as to make a peltier unappealing from a electrical consumption point of view, just like cooling a CPU with pelts has become over the past two years.