Might have a use for pelts after all....

[maxSaleen]

The pentium M has been getting alot of attention lately on the desktop. The DFI 855 MB allows for some pretty serious overclocking which can result in some pretty serious performance. I'd imagine that DFI (or some other board vendor) will realease a 915PM based board. Put a 2 ghz Pentium M the revision with 2mb cache 533 fsb) onto that chipset and you could be in for some serious OC fun.

I mentioned pelts...

I'd imagine that even at 3 ghz you'd be hard pressed to get the said proccessor to output more than 80w of heat, though I'm only speculating. Slap a 100w pelt (common, is 100w really that much?) on and you could be headed for some seriously low temperatures.

Maybe its time to break out that meanwell psu...and a suitable pelt....and some foam, dielectric grease... hell I might have to make my own block for this one.

[Butcher]

100W pelt on an 80W cpu is going to be hotter than just straight watercooling on the CPU.

[AngryAlpaca]

80W? Aren't they specced for 25W? And doesn't the heat increase linearly with the clock speed and with the square of the voltage?

[maxSaleen]

80W? Aren't they specced for 25W?

I said I was just throwing out a number....50 percent increase in speed and maybe an additional .2v.... maybe more like 50w :shrug:

100W pelt on an 80W cpu is going to be hotter than just straight watercooling on the CPU.

Why would this be? I'm aware of the fact that the temperature of the water cooling loop will be raised substantially. I was under the impression that pelts were like heat pumps: they are able to move X amount of heat from one side to the other. If you have a pelt that is rated to move more than heat of the processor outputs, shouldn't you be able to achieve sub ambient temperatures? Maybe I'm missing something here?

[maxSaleen]

Just thought of something....

http://www.cluboc.net/reviews/super_...atingrange.jpg

According to club OC, the 50w pelt on the MCW50-T can hold a 35w GPU at 0 degrees C. Why, again, wouldn't a 100w pelt be able to hold a 80w CPU at a similar temperature?

[Since87]

Originally Posted by maxSaleen

According to club OC, the 50w pelt on the MCW50-T can hold a 35w GPU at 0 degrees C. Why, again, wouldn't a 100w pelt be able to hold a 80w CPU at a similar temperature?

The TEC supplied with an MCW50-T is an 80 Watt TEC.

The graph you linked was generated with a coolant temperature of 25C.

[pHaestus]

It SHOULD be possible to run a 226W pelt at a pretty low voltage and get better efficiency at desired final temps though. Right?

[maxSaleen]

I apologize for my mistake for the MCW-50. Still, shouldn't this scale in a linear fashion? If we say that an overclocked Pentium M puts out 50w, shouldn't a 100w pelt be more than enough to take the processor below ambient?

The graph you linked was generated with a coolant temperature of 25C.

Are you only stating this for our information or do you mean to say that it is an unreasonably low temperature? No offense here I just can't quite decide what you meant. If it was the later I would disagree with you.

[Since87]

Originally Posted by pHaestus

It SHOULD be possible to run a 226W pelt at a pretty low voltage and get better efficiency at desired final temps though. Right?

Hi pH,

Yeah. Assuming you can provide whatever voltage you want, a 226 Watt TEC running somewhere between 6V and 9V is a lot more efficient method for getting 50 Watt heatload to 0C than a 100W TEC is.

[Les]

Originally Posted by Since87

Hi pH,

Yeah. Assuming you can provide whatever voltage you want, a 226 Watt TEC running somewhere between 6V and 9V is a lot more efficient method for getting 50 Watt heatload to 0C than a 100W TEC is.

Good to see you again Sean.
Maybe your "Peltcalc" program is useful to illustrate the 226's performance.

[Since87]

Originally Posted by maxSaleen

I apologize for my mistake for the MCW-50. Still, shouldn't this scale in a linear fashion? If we say that an overclocked Pentium M puts out 50w, shouldn't a 100w pelt be more than enough to take the processor below ambient?

Below ambient? Or to 0C? That's a pretty big gap when you are talking about TEC's. No doubt you can get below ambient with the 100W TEC.

Originally Posted by maxSaleen

Are you only stating this for our information or do you mean to say that it is an unreasonably low temperature? No offense here I just can't quite decide what you meant. If it was the later I would disagree with you.

Just information. Whether it is an "unreasonably low temperature" depends on what the local ambient is and what is used for a rad and fan(s). Keep in mind that the heatload on the radiator may be more than quadrupled when the heat added by the TEC is taken into account.

Also, I imagine that the graph you linked was generated with BillA's heat die. (well insulated from the rest of the world) Keep in mind that as the temperature of the CPU is taken below the temperature of the mobo, additional heat (from vcore regulator etc.) is going to flow from the mobo to the cold side of the TEC, increasing the heatload.

[Since87]

Originally Posted by Les

Good to see you again Sean.
Maybe your "Peltcalc" program is useful to illustrate the 226's performance.

Good to see you as well Les. I've been doing some lurking lately and noticed the the discussion of using C/W at a specific hydraulic power as a single point 'figure of merit' for waterblocks. I appreciate your linking my earlier thoughts on that subject.

I've been plugging some numbers into my "Peltcalc" spreadsheet today. I wouldn't recommend it for those who don't have a reasonable understanding of the Kryotherm software, and are aware of the ways in which the "Peltcalc" spreadsheet falls short. That said, it does let people play with numbers for TEC's which the Kryotherm software doesn't support. (e.g. 226W)

[Cathar]

Hmmm, my experience with both Kryotherm, as well as actual playing around, seems to suggest that the hot-cold side delta of a TEC is somewhat proportional to Vin/Vmax * dTmax, even when given near perfect insulation and thermal transfer efficiency (99.99 Rins and 0.001 for Rh and Rc).

As you lower Vin, you are also lowering the maximum delta from the hot to cold sides.

This was something that I especially learned when playing with the TEC chiller was that while 5v Vin being fed to a 226W TEC mean that the TEC still moved a LOT of heat, its heat moving power dropped away very rapidly at anything greater than an (estimated) ~15C difference between the pelt hot and cold sides.

In essence, while undervolting a TEC results in much higher efficiency, it also comes at the expense of the achievable delta between the hot and cold sides. So while running a 226W TEC at 5V on a 50W heat source can be easily managed by the TEC, expecting it to get much more than 10C below ambient is likely to result in disappointment.

[maxSaleen]

Using Peltcalc, I'm getting 15.9 degrees C for a heat load of 50w with a 226w pelt running at 9v and coolant loop with a C/W of .25. Not bad, it would be interesting to see what sort of an OC you could get out of a Pentium M under such conditions.

[Since87]

Originally Posted by Cathar

Hmmm, my experience with both Kryotherm, as well as actual playing around, seems to suggest that the hot-cold side delta of a TEC is somewhat proportional to Vin/Vmax * dTmax, even when given near perfect insulation and thermal transfer efficiency (99.99 Rins and 0.001 for Rh and Rc).

Hmm, the attached Kryotherm screen shows > 72% of dTmax at 50% of Vmax.

However, you are right, as voltage goes down, both Qmax@v and dTmax@v go down, and at low voltages you have to accept either modest dT's or a very large ratio between the total Qmax of the TEC(s) and the actual heatload.

[Since87]

Originally Posted by maxSaleen

Using Peltcalc, I'm getting 15.9 degrees C for a heat load of 50w with a 226w pelt running at 9v and coolant loop with a C/W of .25. Not bad, it would be interesting to see what sort of an OC you could get out of a Pentium M under such conditions.

That C/W 0.25 is probably very high for reasonable watercooling of the TEC hotside. The TEC has so much more surface area than a CPU that something like 0.1 or less would be more accurate.

On the other hand, peltcalc doesn't take into account the heat being pulled from other places. (e.g. mobo) So it's temperatures are going to be overly optimistic in that regard.

[Butcher]

Generally the difficulty with TECs is keeping the hotside cool enough. If you have a 40C hotside, it starts to become quite difficult to get the coldside to stay cold. Even at 9V you're talking 100W for the TEC alone (at 9V a 226W TEC draws ~10.5A).

[andy497]

Originally Posted by Butcher

Generally the difficulty with TECs is keeping the hotside cool enough. If you have a 40C hotside, it starts to become quite difficult to get the coldside to stay cold. Even at 9V you're talking 100W for the TEC alone (at 9V a 226W TEC draws ~10.5A).

I can verify this in practice. I built a pelt/water cooling system in the early days (circa 2000), and I speced out the parts with software from melcor.com I think. (or maybe tecooling.com or something, it was a long time ago) I ended up going with 4 172 watt pelts powered in series by a custom 0-120 V DC @ 25 amp switching power supply.

It was sick. Those pelts drew ~10 amps at 100 volts (~24 x 4) at peak power. Add that up and you'll see ~1000 watts plus whatever's pumped getting dumped into the radiator (dual heater cores and 120mm fans). Of course, I didn't see that very often, because the 15 amp house circuit breaker was always blowing. By cranking the power supply to max, it would blow a 25 amp breaker pretty quickly as well (switching power supply wasn't all that efficient). So much for practical.

Anyway, long story short, since the power supply was continuously variable, I could play with the input voltage for peak efficiency/min temperature. The radiators would not adequately dump any amount of heat without the T-hot side jumping way up. (Neither would a later-constructed water bong) I actually found that the system worked pretty well with the power supply turned of, i.e. just a plain water-cooling setup. That is just lame.

It was a helluva lot of money, but it was still fun. Nothing beats cranking the front rheostat and watching all the water lines literally wilt, followed by the house going dark.

[Butcher]

The 10.5A draw was measured from my bench PSU with a 226W TEC attached, it's not a theoretical number.
Luckily I can draw 3000W+ on a single line without blowing any fuses/breakers.