How come theres so few Pelts around?

Associate
Joined
24 Feb 2009
Posts
1,758
Location
Teesside
Pretty much as the title says really, how come so few people use them?

Easy enough to find a 140W one for around £10. Thats not too powerful so should be able to work along with most 600+ Watt PSUs I'd imagine. Just a little bit extra work but would it be worth it?

I dont know much about them lol, so I may have it completely wrong though :p
 
Because for most people there's no need and if it fails, you've basically got no cooling whatsoever.

Yea, I read about them not being conductive so if they fail the heat cant move away from the CPU.

I was just looking into it as a temporary thing, just wack it on for a few hours to see how far I can clock and get some good benchmarking in. £10 for the pelt and a bit of work with neoprene and various other inssulating things seems alright for the extra cooling I may get...

I dunno, get some more opinions first :p
 
As you say, it could be interesting for a quick benchmarking run - maybe a day to check it's prime stable or whatever... but gurusan's very right, it's a lot of wattage to be pulling 24/7.
 
I see that fitting one to a processor is hazardous if it fails, I also see that they generate a lot of heat themselves. I don't see how this prevents them being of use though.

In use, I take it you need one capable of cooling the load placed on it. You also need to be able to power it, and your loop must be able to handle the additional heat dropped into it. Anything further?
In these days of very power hungry graphics cards it's hard to believe that a loop can't cope with the heat output of a peltier device. So the standard use would be to get a cooler processor at the cost of warming up the rest of your loop, correct?

There are other uses for this though. If you strap a waterblock to each side of one then you have a way of running two watercooling loops, one at below ambient and the other significantly hotter.

If you strap a waterblock to one side and a really big heatsink to the other, something along the lines of an entire case designed for the purpose, then you have potential for a sub ambient no fans system. Heatsink would clearly have to be very large.

I think these things are interesting, but haven't come up with a sensible use for one yet
 
Last edited:
So the standard use would be to get a cooler processor at the cost of warming up the rest of your loop, correct?
Entirely correct

There are other uses for this though. If you strap a waterblock to each side of one then you have a way of running two watercooling loops, one at below ambient and the other significantly hotter.
That is pretty interesting as a concept

If each side was a complete loop, it would give the potential for sub-ambient, but the failure of the pelt would mean that your CPU was still cooled by a single loop.

Basically, treating the pelt more like a radiator than a part of the waterblock... I've certainly never seen it done.
 
There are other uses for this though. If you strap a waterblock to each side of one then you have a way of running two watercooling loops, one at below ambient and the other significantly hotter.

I may be wrong but I thought the cold side had to have a 'load' e.g. a hot processor or they are uneffective/fail? Sounds interesting though
 
I may be wrong but I thought the cold side had to have a 'load' e.g. a hot processor or they are uneffective/fail? Sounds interesting though

7c is a load, it has heat, a pelt say can reduce water temp by 20C for example, doesn't matter if its cooling something 70c down to 50c, or 10c down to -10C.

The problem however is a 140W pelt ISN'T remotely close to enough. hell, 140W would be the bare minimum I'd put on something like a 280gtx.

You have to remove the cpu heatload AND the heat the pelt creates. In reality you need something MASSIVELY more powerful than the thing you're cooling to do so effectively. IE a 100W cpu(overclocked 150W or more easily when overvolted) you'd want at a bare minimum a 227W pelt, maybe quite a bit more. People used to use 227W ones on single core older cpu's, not massively lower loads to be honest, but a quad core under full load, not sure a 227 would cut it.

This is ignoring the fact most pelts are rated at 14v, at 12v you get a pretty decent drop in capacity, you can use a bigger pelt and run undervolted at 12v, but even then its a massive single load to put on a single line on your psu which will be under constant heavy load. 18amps is what most single rail's on psu's will run at, which is barely capable of running a 227W pelt on its own.

So most people are buying something like a meanwell(last i looked years ago) dedicated variable voltage psu, talking something like £100 for a high quality psu just for the pelt, which will set you back £40, which will consistantly require 250W of power from the wall to run all day every day regardless of cpu load.

IN other words, they cost a bomb, cost a lot to run, are inefficient, a pain to set up and frankly won't give you a "huge" drop in temps/increase in overclock compared to plain watercooling.

Also need to cool the pelt evenly rather than the more direct/concentrated cpu waterblocks that tend to focus removing heat from the centre of the cpu, and not many waterblock makers focusing on pelt blocks anymore(if any?).

In other words, simply not worth it anymore.

About the most I'd bother with these days, is a 80-120w pelt on a gpu, even thats just really not worth it these days. For the price of pelt, cost to run, cost to insulate and voiding warranty by covering the pcb in grease, buy a 2nd gfx card.
 
Exactly why I don't like pelts.

About 2 years ago I did some research into them as sub-ambient temps looked really attractive..

...but using 227 additional watts to cool a 150W cpu is just retarded.
 
How does the rating system on pelts work, answered. Two ratings, with inevitable temperature dependance etc. One i found was 40mm square, draws 140W and pumps at most 80W. Cost a tenner delivered, which is frankly rather cheap. Characteristic curves are difficult to find, but as a crude approximation four of these will eat 600W and drain 300W from the loop. 600W is reasonable, and I think 300W is a lot to be taking from a loop.

I've put a bit more thought into the previous suggestion. I still don't think the traditional use of them is wise these days, however using them to couple two independent loops may work well.

Commercial water blocks are not suitable, but those which would be are basic enough that manufacture of them isn't unreasonable. Windows is currently unavailable so I can't offer a picture, think of a squat square box divided into two. Separate loops flow through each half, the tec is mounted in the wall dividing them. Care is taken to otherwise thermally insulate the two loops. This also suggests using multiple smaller tecs beside each other, which power supplies might get on better with. Not stacking them. Essentially using the tec to make an active heat exchanger.

Intended effect would be a cpu/gpu loop, with pump, passing through one side of the exchanger. A separate loop passing through the other half is sent through radiators.
You drop an extra 200W or so (due to the tec) (apparently rather more, call it 4-600) into the second loop. You maintain the cpu loop at whatever temperature your tec is capable of (at 600W, I think 300W drawn from the cpu loop). Now, as long as the tec is rated high enough it will beat normal water cooling. Potentially conclusively. If it dies, you still have water circulating. A radiator in this loop is safer, but will hurt temps if system goes sub ambient.
The side effect is that the radiator loop will run hotter than in a conventional system. Radiators work more efficiently at higher temperatures, so as long as the loop is cool enough not to damage tubing/pumps, I think this is a solid idea. The electrical efficiency will be horrific, the rate at which heat is passed to the environment pretty good.

Thoughts?

rewrote the above after a fair bit of research. xtreme systems for the win.

Conclusion I'm reaching at present is that sub ambient is a bad call but that pelts are still worth a look. Main appeal is that they allow the loop temperature to be significantly higher, it's possible this means fewer radiators at the cost of more electricity.
It should be possible to hold the cold side of the plate at ambient despite the processor bleeding heat into it, and counterbalance this with higher loop temperature. Radiator efficiency improves with loop temp, and I think there may be a balance to strike there. Problem remains that dead pelts don't conduct heat worth a damn.
 
Last edited:
There are a couple of pelt based water chillers around the net, I built a simple 160w one a few years back, very inefficient. As for pelt failure causing cpu melt down, this was a problem in the days before thermal shut-down, these days your cpu should be safe.

If the idea is to chill the water/loop then the best solution is a commercial water chiller but why stop there?

Commercial chillers use a compressor to reduce the temperature, so why not go for a full phase change system?

Whenever phase change is mentioned at least one of the three following posts mentions noise and power consumption. Well, compared to the constant power draw and cooling requirements of a pelt system I don't think you have much to lose :)
 
You say "water chiller" and on comes the lightbulb.... external radiator box with pelts being used to cool the main loop was exactly where I was going with this idea. So much for an original idea. And you're quite right, processors turn themselves off these days.

Power consumption can go hang, I'll heat my room with a computer rather than a fan heater. Noise is more of an issue but is flexible. For one thing, a water chiller can be turned off and an overclock dialled down to watch films. I doubt I personally will attempt this but we'll see.

How do I work out the cooling requirements for a given pelt system? My guess is
Assume the heat to dissipate = pelt wattage + wattage drawn through pelt + pump
Let ambient =25 degrees. Calculate or estimate flow rate, then find performance graphs from radiators used?

For that matter I really should work out some watercooling calculations in general. I've gone with adding another radiator to my loop to see if it helps temperatures. This is because it fits neatly in my case as much as anything else, but I really should apply thermofluids to wc at some point.

and the calculations are horrible...
 
Last edited:
Jeeez, I went out last night and coming back to this now a little hung over and its got a lot more technical and my head cant cope haha :p

Drunkenmaster, your post was pretty helpful, thanks :) Kinda puts me off them a bit really, but I still want to mess round with some sort of unusual cooling. JonJ678 has some pretty cool ideas, using water on both sides of the pelt... I reckon that would work pretty good!!
 
i had a rig running off a mini fridge (second cpu block bolted to pet of the fridge) and it worked quite well, that was untill one of the hoses on my thermaltake bigwater burst spraying most of the case with coolant, i was doing really well in a call of duty match at the time as well then it when't all slow and funny....... anyway long story short there's me with my first sli rig and brand new athlon x2 4200 running at 2.7 with coolant all over it, i thought it was all dead anyway decided to clean it up and hope for the best, replaced the water and replaced the pipe switched it on and was totaly shocked when it fired up and run fine, my mate's son is still useing it today

anyway if you were wondering, cooling the water is possable and using a second loop could work but as i said i was using a mini fridge and i think pet's work better if the hot side can be cooled well enough
 
Just do a search in this forum, eveything you can think to do, or plan to has been done before (good and badly).

Search for 'TEC' (thermoelectric cooler) as thats the correct name for 'peltier' (this is the effect).

I used them through from a PII333 -> celeron2 well onto Athlon's. Even with 'stacking' a couple too, but by nature TEC builds are very ghetto.

Fact is Phasechange is far more efficent, and by probably cheaper ;)
 
Phase change is very unlikely to be cheaper if I machine the pieces for the peltier chiller myself
edit; will with luck write a better response once sober
 
Last edited:
Back
Top Bottom