Do Ya Need A Pump....eeeerrr NOPE

[Tony]

Convection.. within a loop its the movement of water by heat alone. Im still designing a new loop for my HTPC but a quick and dirty test has my mouth watering with the fact that i can have my cake and yes..I can eat it to

No pump, no failure worry and complete silence.

At this time im using my ubber loop in the workshop, I will conduct more tests with the crossflow rad I have in mind and bring more info to the table then but for now here are some results.

Q9650 @ stock speeds and voltage. EIST works but the VID does not drop on the P5Q Deluxe so I will have to play around with the Asus EPU six engine...so anything I quote at Idle is 6x333 and load is 9x333 with 1.2375V vid.
Please note it can take a few mins for the convection to get moving, temps initially rise to around 66C load but then once the water is moving they settle to what you see me quote here. Keep this in mind if you decide you want to play also



Im using a Hydroflow block, with around 75LBS of mounting pressure here, coupled with Freeze compound.

Idle is sitting between 21 and 26C:



Convection temps:
Flowing in:



Flowing out:



This was taken after a 4 hr load run and then left to idle for 20 mins, next up load temps:



This was around 2HRs into a 5 loop Linpack run using max stress with 16GB of ram...ouch!!

Water temps:
Flowing in:



Flowing out:



Now this is not all that bad...firing the pump on drops the temps by a huge amount...but that's not the plan here...the plan is no pump ever in the loop if I can.

All I need now is test the rad with a cpu mounted in case etc...At this time the CPU is horizontal in the board on a bench (not good in itself for convection)

Overall im quite pleased with this, as long as the rad I choose cools the water in the loop pretty efficiently the project should be good to go soon.

[Tony]

Installed the Asus Six engine and started a fresh Linpack run, seems testing half my ram applies more load to the CPU and load temps have gone up 4C..Im gutted !! LOL

If this Asus app is to be believed full load is 60W 1.2375V, Idle was 4.6W and 1.1V...which again points to the question why would you need a pump while surfin the net etc running low load? you are just wasting energy!

I may design a loop with a couple of Y's to form a bypass and add a pump thats CPU temp controlled...so i get a pump only when i need it

[skinnee]

What size Res are you using? My thinking there is you might be getting some passive cooling in the res due to the volume of water and the water trying to equalize to room temperature.

[Tony]

Res is massive, pipework to the res is long also, it takes quite a while for the heat to move round the loop. For 1C increase in res temp it took around 30 mins on load.

I was just thinking about this and my initial thoughts would be:

1 there is no need for a crossflow rad, you could use a standard rad mounted on edge horizontally.

2 mount a large res on top of the case with the rad right below it...this way cold water flows from the bottom of the res thru the rad before hitting a heat source.

[dopestuff]

facinating.....
good job tony

[Tony]

Q9650 Idle using Asus Six engine = 4.3W....is this real?? 6x333 at 1.1V

A tripple rad would lose 5W with no fans easy...im wondering now if I could add something like a 4850 and a chipset into the loop and still be convection only?

[Bojamijams]

I guess you'll have to try it and see ... this is quite fascinating.. keep us posted please.

[Slovnaft]

This just might convince me to go WC. An absolutely silent rig sure is enticing.

I'm wondering how big of a rad this technique can accomodate. This certainly changes the dynamic of a loop, as the goal would be to get as little differential vertical as possible. Seems perfect for multi-loop systems!

[Tony]

Remember guys this is not for massive overclocking, saying that as I hinted before you could add 2 Y's to form a Bypass and add a small pump (needs to be 6W or less) fed off the CPU fan header, bios would then switch the pump on and off dependent on CPU Tcase temps ot you could go with Tbalancer (which is what I am going to use) with speedfan or similar plugin to use diode temps on the CPU. Pump only pumps when its needed, at idle you use convection and sit in silent bliss as even the fans could be shut off bar say 1 just cooling the water as it enters the rad.

[skinnee]

Res is massive, pipework to the res is long also, it takes quite a while for the heat to move round the loop. For 1C increase in res temp it took around 30 mins on load.

5 gallon? How much are we talking here? I might have to try this scenario out on the liquid test bench...as I am waiting for pumps.

2 mount a large res on top of the case with the rad right below it...this way cold water flows from the bottom of the res thru the rad before hitting a heat source.

This is what I was thinking, if the Res is large enough you will get cooling due to the water in the res trying to reach equalibrium with the ambient temp of the room, then water flowing through the rad would get that extra little drop before going to a heat source.

[Tony]

Res here is around 12 gallons with around 11 actually installed. March 230V pump and 4 blocks...I usually test 4 boards all at the same time.

im thinking of making a miniloop with a 120.3, no res and a T bar to fill it...maybe 18x2 inches of pipe from the rad to the block to mimic what a case layout would be and see what its like without a res...then maybe add a res (as this would also add head pressure) and see if it improves things enough for it to be deemed a necessity.

[Bojamijams]

You sir... have a fun job.

[Tony]

You would think that but it does get tedious after a while, especially when you know boards have bad bios files...

Anyway back to watercooling ...Asus Six engine lowers the FSB and i have no way of bringing it back up...46MHZ loss just so I get Vid change during EIST is not acceptable...DFI boards Always worked with EIST and OCing i may try the new DKX48 next.

You know someone needs to design a case that has a top compartment, large enough for a tripple rad on edge and a res that holds 1 gallon or so of coolant......would be a little top heavy though me thinks ...LOL

[crazy1323]

This is because viscous liquids defuse rather quickly. When you are heating the water near the back the water not only defuses fairly quickly but the heat is distributed well within the loop.

Look at it this way. If you stand near a heater, you heat up too and you heating up could warm someone next to you. Now think of the people as molecules of water. The heat is transered very well from molecule to molecule.

[HESmelaugh]

Wow, that's really amazing! I would never even have considered trying to run a watercooling loop with no pump. Really fascinating stuff.

[Eldonko]

Interesting concept. What are the chances of the flow stopping unexpectedly?

Can we see the whole set up?

[BlueAqua]

Reminds me of nuclear reactors using natural circulation. Natural circulation can work great, but the loop has to be designed around that. You need to have very little resistance in the tubing, that means no tight bends, no elbows, no multiple blocks. Having large runs of tubing on each side running vertically will help too. The larger the distance between the rad and the cpu block the better. I don't think a huge res is necessary as long as the system works.

The main point is that you have to setup your thermal driving head. The colder you can make the water coming out of the radiator the better. The greater the difference of temperatures (delta T) across each loop the more flow you'll have.

This is something I've been wanting to try for some time now. Heck, I might just now do it.

[dopestuff]

here's an whole pc case from zalman build totaly on the convection principle (only no watercooled though but heatpipe)
http://www.zalman.co.kr/ENG/product/...ad.asp?idx=116

[Tony]

here's an whole pc case from zalman build totaly on the convection principle (only no watercooled though but heatpipe)
http://www.zalman.co.kr/ENG/product/...ad.asp?idx=116

Yup I have one of these also.. TNN500A prototype.

[Tony]

Interesting concept. What are the chances of the flow stopping unexpectedly?

Can we see the whole set up?

The loop is to huge and being honest I don't want people seeing huge sections of my workshop.

When i get the small loop together with a rad etc i will run some pics and test results off. As long as I can moderate the voltages and CPU speeds accaurately in relation to load i think it may work quite well. I for 1 always though Q9650's were 100W+...seems Intel has us fooled

I will use a thermochill 120.3, its popular and I have 3 floating around doing nothing at the moment.

[axis]

Would be nice to see a setup that is more "normal". A 12gal reservior is hardly normal nor feasable for the average user. I'm sure you get some cooling from the massive reservior. Do you have a way to check flow speed? I would also assume that the hotter the water gets, the faster it should move, thus increasing cooling through the rad. While you're at it, see if you could add a a single VERY SLOW moving fan to the rad to see how much, if any difference it makes. I'm thinking maybe a YL low at stock speed and dialed down.

Axis

[dopestuff]

if i understood correctly he is already using fans on his rad only not using a pump but he do want to try running whitout a fan when the pc is at idle (and the fans kicking in at a certain temp.. )

[Eldonko]

Ok no worries, I was just curious what the res looks like and what you made it from.

The loop is to huge and being honest I don't want people seeing huge sections of my workshop.

When i get the small loop together with a rad etc i will run some pics and test results off. As long as I can moderate the voltages and CPU speeds accaurately in relation to load i think it may work quite well. I for 1 always though Q9650's were 100W+...seems Intel has us fooled

I will use a thermochill 120.3, its popular and I have 3 floating around doing nothing at the moment.

[Tony]

No fans no pump at the moment, but im adding a rad tomorrow and will test closed loop just like in case. Idle CPU 5W (so asus tells me) so im pretty confident.

[Tony]

Ok no worries, I was just curious what the res looks like and what you made it from.

Central heating header tank...big plastic reservoir 18x12x18high