Dryice container tests in 160w bench

[ilkkahy]

http://koti.mbnet.fi/ilkkahy/1.JPG

I decided to make thread about the previous dryice tests with the 160w bench ive been using to measure containers cooling efficiency to get some comparable results which kind of containers work well and which doesnt.

I would advise to make this sticky (and perhaps take off some of those noob-guides) as there arent really much (if any) this kind of information available at the moment. Im not saying this testing method is very accurate but at least it gives some direction (this doesnt tell how much there is fluctuation between idle/load).

Testers contact surface area is about 3,14cm^2 and used resistor is rated 330ohm which gives 160w in 230v line (230v^2/330ohm=160w). Thermal probe is inside the 14mm thick copper die. If i was to make another tester some day it had adjustable heatload and more advanced temperature measuring.

Anyway here are the results:

http://koti.mbnet.fi/ilkkahy/2.JPG
http://koti.mbnet.fi/ilkkahy/3.JPG
http://koti.mbnet.fi/ilkkahy/4.JPG
http://koti.mbnet.fi/ilkkahy/5.jpg
http://koti.mbnet.fi/ilkkahy/6.JPG

And last here are some pics of other containers that i havent tested or turned out to be flaws like the sandcast aluminium container in the picture (+6c due the wrong alloy type). The copper version flawed in the making process as there was suction and i also put the sand core "bit" on the other edge :D But that was my first own casting so its only reasonable... didnt have the intrest to try another time as there wasnt copper castings going on the next day at the foundry. There is also pic from r1150 heatpipe dryice container which didnt work as it supposed to (-22c). I think there was wrong pressure but idea was good anyway. Ive also tested forced convection in the bench but i havent got success yet (+3c with pathetic flow). I think there is cavitation with the pump when methanole comes really cold because flow rate seems to drop really low (just as dr.cooling told about year ago when i told i will test it). This could be avoided by using different type of pump or liquid (or both). Its very intresting and potential (not to mention how easy it would be with SLI-setups and such) cooling method but needs more time and energy than i will have in the near-future.

http://koti.mbnet.fi/ilkkahy/7.JPG

I hope someone gets something out of this :)

[Cpt.Planet]

Nice work. I personally have all of my pots tested at 400,500,and 600 watts load before they even get made through ansys. A friend of mine goes to college for conjugate heat transfer analysis so it helps me tell how the pot will perform before its even made.

It upsets me that so many people just do designs with out doing any testing or having any theorys. I prolly have 30 designs i havent tried because I am able to use it to optimize my pots.

Finally TBH 160 watts of load is not alot. Many of the current processors are putting out much more than that. Some like the quad cores are putting out 500 + watts.

[ilkkahy]

Thats true. I started testing these in A64 times so heatload is pretty low compared to current cpus but i didnt want to change it higher because then temps would not be comparable. I think results will correlate somehow to higher loads too. Some guys use to test cpu evaporators with higher wattage but the copper die size is usually a lot bigger than any cpu/gpu dies (like 16cm^2 or something) which isnt good because there should be some "horisontal" heat conducting properties from centre of container bottom to pipe too (which is ignored if heatload is put partly "sraight" to pipe part by using large contact surface area).

It would be great to see other peoples container researches in this thread too but honestly im not very optimistic well see any ;) Also anybody have information about thermodynamics 3d-modelling software that could be used with desinging these?

[Stapler]

I'm glad to see you post these Ilkkahy, it seems like a lot of people don't do any real testing of their pots before making them. I think 160w on your load tester is more stressfuly than a lot of the other load testers I'v seen, which are far larger than actual cpus.

[Stapler]

Btw, I think this should be stuck rather than a lot of the noob guides, they're really not that helpful.

[Nosfer@tu]

sticky :D

And lets have a user do some reviews with load tester.

Maybe Bazzx or what his name is, Cool guy :D

[Solarfall]

this was very interesting to read, thanks for posting this ilkkahy :up: i vote for a sticky
btw how do you thing a container would perform if it had a center pole made from silver.. could it give better temps or would it be just a waste of money ??

[Freddie123]

+1 for sticky

[@rne]

Thank you for your tests. It would be also interesting to know how a pot with copper base and alupipe performs in comparison to the others:)

[Gautam]

Very nice tests ilkkahy. :up:

I hope you guys don't mean what I think you by the "noob-guides" though.

[Cpt.Planet]

Thank you for your tests. It would be also interesting to know how a pot with copper base and alupipe performs in comparison to the others:)

Over 3 inches it doesnt much matter.

Just curious what are you using to get the temps? A probe in the pot or a probe on the resistor?

@ Guatum I think he means the non kingpin guides. The others dont give much info or are old an outdated.

Heres some of the pre production runs of my pots. This particular example is aluminum with a 500 watt heat load. The load simulator was a modeled core2duo/am2 chip and then imported and put under the pot.

The pot was filled with dryice/acetone at -70C about2 inches high because thats all that really matters. We were able to calibrate this with man real world runs and comparing to these runs and adjust the level to where it was most accurate. These arent 100% accurate but it really helps with pot design/optimization.. With 500 watts of load you see ~ 252K. The copper performed much better but cant seem to find the screen shots.

The one picture shows how the base is able to cool the chip down and remove the heat. The second shows how the heat travels up the container.

[Eldonko]

Great testing, thanks for sharing!

[@rne]

Over 3 inches it doesnt much matter.


Ah ok, thank you:)

[Cpt.Planet]

Ah ok, thank you:)

Check the edit for picture of why.

[ilkkahy]

Thermal probe is placed inside the copper die which is between resistor and container. Measuring container bottom is not the right way to go which can be seen from those ansys pictures too ;)

Is that red area the dryice level which is why upper from there container starts to get warmer or whats the reason for it?

And about silver container, i dont think its worth a try. Any inpurities will often result in loss of heat conducting properties and practically this means youll propably need to have very pure silver to get pass 99,99 copper. Im not sure how does silver heat conducting improve in cold temps but copper has this property too. I would definetly concentrate on designing copper containers instead of changing material to silver :)

Have you made comparisons with same size old school wide flat bottom and cone/stepper type bottoms with that program Cpt.Planet?

Nice to notice that some people still finds this stuff intresting!

[Cpt.Planet]

Thermal probe is placed inside the copper die which is between resistor and container. Measuring container bottom is not the right way to go which can be seen from those ansys pictures too ;)

Is that red area the dryice level which is why upper from there container starts to get warmer or whats the reason for it?

And about silver container, i dont think its worth a try. Any inpurities will often result in loss of heat conducting properties and practically this means youll propably need to have very pure silver to get pass 99,99 copper. Im not sure how does silver heat conducting improve in cold temps but copper has

this property too. I would definetly concentrate on designing copper containers instead of changing material to silver :)

Have you made comparisons with same size old school wide flat bottom and cone/stepper type bottoms with that program Cpt.Planet?

Nice to notice that some people still finds this stuff intresting!

Its not measuring the base its actually measuring very close core temps. For some reason thats how the cfx picture shows it.

The red area is about an inch above where the acetone/dry ice mixture.

Yes I actually have. I have done it with 2 cones, 1 wide old school flat, and one that was rounded(filleted) with the same internals. The one with the tallest cone was the middle performer but it is able to dissapate the heat better and keep the core under constant temps going from idle to load and so on. The old school just flat/no cone was the coldest at the core but suffered from having to much mass. The rounded base was the worst performer. We are talking about differences of 1C however.

[SF3D]

I have tested five of these tubes in real world usage. They are working great and temperature differences are quite small.

Tomorrow I will test my QX6800 with that solid copper pot. I will use LN2 this time.

[luihed]

+1 for sticky...

Im one of those that made pots without testing but I have no access or even the knowledge to test so I have no choice.... Ive only made 4 pots ( one still not finish) and I think all of them has done well.... I just make what I think will work with what I have and hope it works....:D

ilkkahy, I love the design of your pots with the drill holes..... I drew a similar one that I want to use as a base for DICE but I wasnt sure how easy it will be to make... this piece goes in a copper pipe...

H= 40mm
D= 50mm
bottom of holes = 7mm ( from base )
top of holes= 13mm ( from base )
Big hole (D) =10mm
Small holes ( D ) = 5mm
pole ( Dx H )=14mm x 27mm

http://ocidb.com/albums/userpics/10062/normal_dicebase.jpg

[ilkkahy]

That Sampsas solid container (-38c one) had center pole when i designed it but i wasnt able to do it with the lathe blades i have. Its much more clever to make just 5-10cm long base since it requires less material and machining is easier (less vibrations due shorter "arm" of the blade).

I know that design is not very easy to make (at least if holes have to go 18cm deep like in those my containers) but its doable and i think it would kick ass (even though i would go bigger diameter than 50mm).

[luihed]

Yeah, 50cm is pretty narrow but thats all they have at the recycling place where I get my copper:( ... Big difference paying $7 per inch ( of round copper ) compared to $35/ inch if I buy from store.... Got lucky once and found one thats 3" wide but it was only 3.5" high...

What ive always wondered is how the thickness of the base affects the pots cooling ability.... Is it as simple as thinner will be colder but wont hold temp at load?

I think this one will be easier to make.... the outer ( bigger ) holes are lower than the ones in the center so the dice surrounds the core area....

http://ocidb.com/albums/userpics/10062/normal_dicepole.png

[Cpt.Planet]

luihed what are you using to design those? Also if your using dry ice only that center post can hurt performance because you need to keep a snowy mix orelse dry ice gets clumped up above it and doesnt help.

[luihed]

Its a program from emachineshop.com , its pretty neat cos you can get a quote from them to make your piece... But its never reasonable lol...

As for the pole, I had good results in dry ice with it so I thought its good for DI... The way I look at it is for pots that doesnt have alot of copper ( braze ) it might be better cos it gives more mass... solid container doesnt have that problem so it might make it perform worse like you said.... having just a lil pice of 10mm copper at the base doesnt look too effective... Thats just my opinion so LMK if Im totally wrong hehe.. Also I crush the Dice to snow when I bench so no prob with them not reaching the bottom...


Here's the 2nd pot I made, 5 piece ( 7 if you include solder and epoxy lol )... up there for one of the ugliest but it took my E6600 to 5gs:D
http://www3.telus.net/luis99/luihed1.2.JPG

http://www3.telus.net/luis99/oc/inside.JPG