a not so quick attempt at H2O cooling.

[Aedan]

I finally got around to putting this stuff together and seeing if I can make it work.

I decided that until I get around to water cooling the PSU, that I might as well make one fan count for as much as possible. Hence the combined Radiator-Fan-PS unit (RFPSU for short!).

http://www.hawks.dircon.co.uk/aoa/rad-fan-psu.jpg

Yes, it's a Maxxxpert MONO radiator. too.

Just for grins, the other half of the system looks like this...

http://www.hawks.dircon.co.uk/aoa/otherbits.jpg

It's not tidy yet, cuz it's a work in progress. The CPU block is an InnovaCOOL 3, the GPU block is a InnovaGraph-O-Matic, and the northbridge cooler is a Zern AT design. The pump is a Sicce IDRA, and I wouldn't recommend it to anyone.

The funny twisted ribbon cable is a Ultra 160 SCSI cable BTW. There's another hidden narrow SCSI cable elsewhere.

AidanII

[Random Nonsense]

plumbers nightmare!

Watercooling rules! Hey, this is the 1st pic you have posted of your rig ever isnt it?

[Daniel ~]

Surgery is seldom pretty!

Is there a significant fan noise increase?

[Aedan]

Originally posted by Random Nonsense
plumbers nightmare!

Watercooling rules! Hey, this is the 1st pic you have posted of your rig ever isnt it?

Err, actually, no. It's the first pic of it in this state though.

AidanII

[Aedan]

Originally posted by Daniel ~
Surgery is seldom pretty!

Is there a significant fan noise increase?

Well, the PSU fan doesn't exist any longer - the radiator fan is doubling up as the PSU fan. Taking the 92mm case fan and PSU fan out, as well as all those stupid little twirling things on stuff lik northbridge and GPU, and the result is that it's a lot quieter.

Now if only there was a good way to quieten down 10,000 RPM hard disks...

AidanII

[Daniel ~]

One less fan then, well done. The drive noise I could live with!

[Allan]

Nice! But this makes be realize I need a bigger case if I want to go with watercooling anytime soon!

[Holst]

Originally posted by Allan
Nice! But this makes be realize I need a bigger case if I want to go with watercooling anytime soon!

Just shove the watercooling stuff next to you PC like I do.

I alter my system far too much to keep it all in my case.

Nice system Aidan.

How do you get on with the waterblock, especially the mounting. Those inovatech blocks do look very nicely made.

Do you get much flow restriction through the northbridge cooler, im thinking about adding a watercooled northbridge although mine will be on a paralell loop to stop any flow restrictions.

[The Spyder]

Exclent job Aidain!!! Ive been trying to get some innovative stuff myself......

[drow_elf]

dumb question but wouldnt that pump look really nice infront of the "RFPSU" ??

[dimmreaper]

Originally posted by AidanII
Now if only there was a good way to quieten down 10,000 RPM hard disks...

Enclose them in a foam lined wood box and water-cool them . . .

Don't laugh, I've seen it done . . .

[Aedan]

My current thoughts are to:

Run a couple of copper pipes (plumbing) down the inside of the back of the PC. At strategic points, there'd be a T connector that would allow each waterblock to be connected seperately,and in parallel. However, this would require some way of controlling the flow of water into each circuit, and I've not seen anything small enough to look really good. It'd be even better if the valves could be automated!

The PSU is currently air cooled. As the air needs to be channeled, and heat removed, the PSU is in it's box, which is large (standard really!). Building two water blocks, one for the rectifiers and one for the switching transistors would allow me to reduce the high of the PSU significantly. In addition, I would be able to mount the PSU board and enclosure whereever. Another plus is that I can seperate the mains input from the PSU case, so that I can run a pump/fan controller, which allows the pump and fan to start first and stop last. This means a trip to the plumbing suppliers in order to pick up some 10mm copper tubing, and something to bend it with. I'll also need to pick up some replacement isolating washers, to keep things isolated. Yes, that stuff that looks like a thermal pad is actually to isolate the transistors/rectifiers rather than to help conduct heat!

The hard disks make too much noise. So, I'd like to put together some sideplates that are watercooled in order to remove the heat. Two 7200RPM and a 10,000RPM drive make a lot of noise! The existing case has a removable 3.5" drive holder, which makes it easier to remove the existing one and fabricate a new one that'll hang in the same space.

I also want to move the pump! However, I need some space for it, so the stuff above would have to happen first, esp the PSU mod.

And last, I need to build a controller for it all!

AidanII

[Aedan]

Originally posted by Holst
How do you get on with the waterblock, especially the mounting. Those inovatech blocks do look very nicely made.

Do you get much flow restriction through the northbridge cooler, im thinking about adding a watercooled northbridge although mine will be on a paralell loop to stop any flow restrictions.

I quite like the mounting actually. It's tri-hooked on both sides, and the bar at the top acts as a lever against the waterblock. Installing it is seriously easy, just have to remember that the levered hold down is capable of applying a pretty large amount of pressure. The blocks look pretty well made!

However, the interior of the GPU cooler is basically a flat copper plate. There doesn't appear to be any flow control or turbulators in there, so it's not as efficent as it could be. On the other hand, the GPU doesn't generate that much heat, and just about anything is better than the gainward heatsink that came with the card! (The stock heatsink was built by gluing Al vanes on top of the Al baseplate, and then putting a fan in the middle. Not really what I'd call a heatsink!)

None of the waterblocks are stock, as the Innovatek blocks are fitted with 8mm (I think) barbs with a compression fitting to hold the hose. I've replaced them with 10mm barbs without the compression fitting, mainly cuz I already had the barbs and 10mm silicon tubing.

In terms of flow restriction, the Maxxxpert mono radiator is probably the most restrictive part. The copper tubing is 8mm. On the other hand, should I ever add another radiator, then it's easy to run them in parallel!

One of the things I'm considering is a pressure relief valve across the pump. That way the pump can operate at full efficiency, but the flow is still controlled. This is really just a thought, and I suspect nothing will come of it. However, if I ever run things in parallel, then it will probably be included!

[Winkie]

One of the best ways to quieten hard drive noise is rubber washers


also, foam boxes would work

[Aedan]

Originally posted by Winkie
One of the best ways to quieten hard drive noise is rubber washers

I'm avoiding rubber washers at the moment. I remember reading a Seagate whitepaper about such mounting methods and their effect on hard disks.

One of the assumptions made by the software is that the case of the hard disk will remain in the same position, regardless of the moments of force acting upon it. (from the HDD's point of view!)

When the drive comes to make a seek, it starts the heads moving towards the track wanted. Depending on the distance, the acceleration may be quite high. During the acceleration phase, the case experiences the effects of the change of momentum. This causes the heads to accelerate slightly more slowly than normal. As the software is basing the inital seek on a computation of where it is, where it needs to be, the weight of the heads and various other factors (Yes, it's making a guesstimate of how far/fast to move the heads!), the software's model now no longer corrisponds to the real world.

Exactly the same thing happens when the drive deaccelerates the heads towards the end of the seek. The case once again moves slightly, leaving the software's model out of line with the real world.

Of course, once the software realises that over the wrong track, it repeats the seek. As the seek is now just a few tracks, there's very little movement, so little effect upon the case.

On the other hand, the drive has had to make multiple seeks in order to find a track that would ordinarily only take a single seek. The result is that the access time of the drive goes up.

Yes, Seagate did test it in practice as well as in theory. I can't remember how much seek time went up, but it did go up, especially for long seeks.

Alas, I've never been able to find the paper again, nor can I remember it's title. Otherwise I'd point you towards it, and let you read it yourself.

AidanII

[Kaitain]

dumb question but wouldnt that pump look really nice infront of the "RFPSU" ??

Probably, but centrifugal pumps such as that one typically need to be at the lowest point in the system - this being due to their inability to "suck". Unless it's a perfectly closed system, the water entering the pump won't move fast enough to prevent cavitation/pressure loss without the assistance of some gravity.

That does look to be a perfectly closed system though... (no header tank AidanII? Not expecting the temperature to rise much?)

However, this would require some way of controlling the flow of water into each circuit, and I've not seen anything small enough to look really good.

I believe you can get gate-valves that are set using a little gromet-screw and don't add much to the local pipe diameter. Whether your local plumbing shop has them is another matter. If money's no object then I'd strongly recommend Swagelok fittings. I used them all year in the lab, and they're excellent for all but the most stupidly extreme conditions. Starting at £7 per fitting, they bloody well ought to be

This means a trip to the plumbing suppliers in order to pick up some 10mm copper tubing, and something to bend it with

If you can't find pipe-springs that small, then just plug one end of the tube with a rag, pack it with sand and bend it around your favourite vice/knee/local's head. You can get very accurate bends without much special equipment, and even less pipe-deformation than most mechanical benders (!)

One of the things I'm considering is a pressure relief valve across the pump.

Since you're running a closed system and are going to move to solid tubing, consider also a system pressure relief valve that vents to a drainage point outside the case. That way your system isn't going to get a bath if your pump fails (there's a thread on this topic some months ago where a user's pump failed, then shattered under the pressure the water built up. Also toasted the CPU). Putting the relief point high in the system should allow water to be drawn past hot components on its way out giving them a few seconds respite - it should be trivial to arrange an instant power cut-out if water ventures through the relief tube as well

[Aedan]

Originally posted by Kaitain
That does look to be a perfectly closed system though... (no header tank AidanII? Not expecting the temperature to rise much?)

Nope, no header tank. Silicon tubing and a system operating below atmospheric instead should keep that in check. Of course, if the system hits a catastrophic temperature increase, then it's not going to help, but by then the system's toast anyhow. I'm just amazed at how elastic silicon tubing is!

If you can't find pipe-springs that small, then just plug one end of the tube with a rag, pack it with sand and bend it around your favourite vice/knee/local's head. You can get very accurate bends without much special equipment, and even less pipe-deformation than most mechanical benders (!)

Now, that's a hint worth knowing. Any idea what the minimum radius would be for copper 10mm plumbing pipe?

Since you're running a closed system and are going to move to solid tubing, consider also a system pressure relief valve that vents to a drainage point outside the case. That way your system isn't going to get a bath if your pump fails Putting the relief point high in the system should allow water to be drawn past hot components on its way out giving them a few seconds respite - it should be trivial to arrange an instant power cut-out if water ventures through the relief tube as well

The problem with the pressure relief valve is where to vent the water to. However, I'm having central heating replaced, so I'll nick the relief valve from that. I suspect that 4 bar might be a bit high though.

AidanII

[Daniel ~]

One thing I thought I'd mention:
Are you going to get the maximum blow through your radiator with it passing through you PSU?

You'll probably be fine if it's a bit off, just wanted to bring it up, enveloped as I am in base envy......

[Random Nonsense]

Kaitain, we have something even better in the aero engineering industry, a VERY low melting point solder, its really soft too, fill pipe, bend, heat, and it all drains away!

[Aedan]

Originally posted by Daniel ~
One thing I thought I'd mention:
Are you going to get the maximum blow through your radiator with it passing through you PSU?

No - the fan is completely the wrong type to really drive the radiator well. A blower would be far more effective, but I don't have one to hand.

The PSU part is little more than some ducting in reality, as the PCB is mounted upside down. With the fan removed from the PSU, the opening is fairly free.

AidanII