156w pelt yielding weak results... why?
 

K guys (and girls)... This article is gonna go up on ProCooling pretty soon so you will have a better idea then, but here's the deal... I have a 16v 156w peltier running at 12v. It is on its own 300w AT psu. So that's like 115w... When I have a duron at 950 1.85v, my temps are around ambient. This is w/ Overclock-Watercool's copper waterblock and a copper plate. I'm using arctic silver. I expected to get to around 950, but i'm unpleasantly surprised by the temperatures...

i hope to god that is on water coolign!

and if it is, i suspect that oyu are not radiating the heat away enough

you had better have a very large radiator: more than 5"x5"

if the water temp is over 35C you have a problem

Well, the water temps were definitely higher than room temp, but they were not that high. You could definitely be right, though. The radiator was kinda weak. My next project will have a big radiator :-).

i am using one from napa auto parts

its a 10x5 inch hayden radiator
i think it was like 35 bucks
awsome performance

and you dont have to wait for someone to ship it to you

I'm gonna go w/ the biggest of the black radiators at www.overclock-watercool.com

How thick is your cold plate? also, does your waterblock get hot to the touch(especially the base of the waterblock, is it hot feeling?)? your problem could be one of three things (i have the same issue with a 120w pelt on a duron 650, around 60-75F temps)

1. Your cold plate sucks
2. Your cold plate sucks
3. Your waterblock sucks

1 and 2 are most likely, and what i know for a fact is my bottleneck, you see, with the Duron chips, you have a teensy tiny core to radiate all that heat, so your coldplate needs to be extremely effective, i have a brass/bronze one now that's like 2.5mm thick and while it was fine on a nice big fat slug'd celeron 366 on a duron it just can't cut it. I'm looking for a source for a pure silver cold plate, something like 45x45x8mm

and yes, i have tried two waterblocks, both of which i know to be extremely high performers, with only a 4F change in temp www.anime-unlimited.com/tomservo <- the aluminum block is what i was using, now using the copper block of which i have no pic of the inside

PS: Kevin, do you buy painted heatsinks too? cause i can't think why anyone in their right mind would buy a painted radiator for a high performance application. Paint isn't very heat conductive as opposed to raw metal.

[This message has been edited by tentaclefiend (edited 11-27-2000).]

My cold plate is the one from Dangerden... extremely smooth and shiny. It does not suck, but it is perhaps too thick. My waterblock is awesome (overclock-watercool's copper swirly chamber block). As far as painted waterblocks go, I definitely do NOT use them :-). The radiator i referred is one of the highest performers around as it uses a different design than the standard radiators. The one i'm currently using is made of raw metal, but the water is warm so I think that the radiator is the culprit. The new one I get should solve problems for me.

It is not an enamel based paint on those radiators, its EXACTLY the same as the Anti-corrosion chemical applied to 99% of all of radiators, this one is Black though in order to help its heat transfer just as Alpha HSF's are black.


If you wanted to strip them you would need to bead blast em. ( which isn't such a bad idea... but be careful the thin Al. channels would erode away fast.( while sand blasting)

------------------
C-ya
Joe

I am fairly certain that heatsinks are anodized black for style ... the FOP38 used to be anodized black but is now just aluminum because it disipates heat easier, or so Millisec says.

Oh yea and if you ask stiltner over at the hardforums he will prolly be able to tell you how to strip the stuff off of the radiator if you need to know.

anodizing isn't as polished smooth looking as the units i've seen pictured, so maybe i've seen different units

Hi there Kevin thanks a lot for good article.
I think you might want to see this. I did peltier/watercooling experiment. Here are some of my temp resaults. When I am testing my peltiers I use 6mm coldplate (pure copper) And I flush my waterblock (pure copper also) with tap water @7-9c.

System:AbitKT7-R Duron 700 300w Aopen PSU. All temps are under full load(one hour), no spot readings.

Temp resaults:

Air Cooling:Coolermaster hsf cpu=46c @1,94v/1000mhz (case open)

Water cooling: Tap water @9c cpu=20c @2,15v/1100mhz

Water cooling: water@24c cpu=33c @2,15v/1100mhz

Single peltier 68w water@9c cooled: cpu=13c @1,7v/800mhz

Single Peltier 68w water@9c cooled:cpu=30c @1,98v/936mhz

Dual Peltiers 2x68w water@9 cooled:cpu=9c @1,7v/800mhz

Dual Peltiers 2x68w water@9 cooled:cpu=23c @1,98v/1050mhz

Dual Peltiers 2x68w water@9 cooled:cpu=32c @2,15v/1100mhz

Peltiers were run at 10-16v side by side,on a separate BIG 35A psu.. 12v gave best 'cooling'.

I think these resaults are in phase with your readings. You might get lower temps by using bigger cooler and higher voltage for your peltiers, but you will not see any dramatic temp droppings with 156w unit.

Even a regular watercooling outperforms my 136w dual peltier setup, the reason is that they do not transfer enough heat. Only at 700mhz and 1.65v did I see any condensing on my coldplate, And at 1100mhz 2.10v my coldplate was a lot hotter than my waterblock. Bigger peltiers are needed for a Duron at this voltage/speed! In my opinion you could possibly get below 0c by using at least two 156w units..Hmm then you need some good psu I guess (26A@12V)

Thanks for those results. For my next project, I'm going to have much better radiators, the voltage hack, and a water chiller w/ between 180 and 210w of total power coming from 3 side-by-side pelts.

Check this out!!

I have the solution and I will give you all the resutls in my upcoming article.

I used the four holes in the socket A mobo to use 1in square tubing to mount my waterblock on. I then used a 118 W pelt with a copper cold plate. This yielded me about the same as yoiu but a bit better. I am actually below ambient with a modded voltage mobo running 26C under load at 2.1 vcore (Its a KT7 RAID)

I used my regular power supply to power the 118W and then hooked up an old AT PSU to power a inline peltier which cools the water. The thing to pay attention to is that you dont want a really powerful pump. you want to keep the water in so it will cool as it passes through the waterblock with the pelt on it. I superglued a few of them on there, but ended up ruining too many waterblocks doing that, so I have a simple attachment mechanism which clamps the pelt on the waterblock. this stabilizes my temperatures, which really helps keep it under control under a load.

you can check out my little chapters of socket A watercooling at http://www.athlonoc.com which will soon have another segment when I take this 600@1134 Duron out and put in a 750 blue core T-bird that has already hit 1200 with air alone.

BTW The 1000 I had at 1345 was crushed during preliminary tests of my waterblock attachment mechanism, RIP 1ghz core.

GOTTA LEARN SOMEHOW!!

I have a couple 24v 10A lambda PSU's for my .8 24v 172W drift pelt that i'll be putting on soon!! YEEHAW!!

------------------

Hey Kev, I have found that with AT and ATX PS's you have to have a load on the 5v rail or you are only getting 10V ( I originally thought this only happened in ATX PSU's)

------------------
C-ya
Joe - Owner/Editor
www.ProCooling.com
Where the Completely Addicted Come to Cool Off

Somebody set up us the bomb.

Joe which ones only give you 10V ? I have all my plug connections cut off except those for the peltier and I get 12V..

Just Curious http://www.procooling.com/ubb/smile.gif

------------------

Yeah... a lot of 'em need a dummy load. As far as the last project though... I'm now almost positive that it was the water temperature that screwed me over...

TEC mfgrs recommend 150 to 300 psi clamping pressure, FAR beyond what an Athlon CPU (or even all 3 lugs) can tolerate.

A separate mechanical clamping sys is req between the cold plate (oversize) and the waterblock (again, must be larger than the TEC). The screws must be thermally insulated from the cold plate.
Yes, that's how its done.

Incomplete post.
An alternative is to adhesively bond the assy together, Artic Silver would be good. The question is if it would stay together, thermal cycling is VERY tough.

be cool

[This message has been edited by BillA (edited 01-15-2001).]

BillA is doing some major retraining here..=)

matrickage.... cool.. (literally)
but these PSU cost so damn much..

and since somebody (BillA http://www.procooling.com/ubb/smile.gif) helped me out of the dream to use a 20A 13.8V PSU I'm looking for a 25A version that can do 12.9V... (2x156W pelts)


------------------
OC'ing is my middlename

ck out www.herbach.com
on pg 38 they had a 12v @ 25A for $69.95 w/2 fans even
you'll need about 3A or so on the 5V output

be cool