Input required

[Bravo]

Greetings :)

I am looking for some feedback in improving a block thats been designed/produced.

http://bravo.ausgamers.com/waterblock2.jpg

The block above features the "spiral" design (as it is called in these forums, at OCAU, its called a 'G' channel) with the center barb acting as an inlet. 1/2" inch barbs, with a water channel width of 12mm/a height of 12mm, and a base thickness of 3mm. Believe it or not, this block was designed around mounting on a GPU or northbridge.

A couple of questions i have this is:

1. How effective a design is this ? In arguing with aussies at home at OCAU (www.overclockers.com.au) the main principle around designing blocks is the more surface area it has, the more effective it is at cooling.

2. I intend to adapt this design for CPU usage. However, because i intend to use a 226watt TEC (15v @ 23.1amps, same as tekheads), it has been suggested that any locally available retail block would not be able to copt with the heatload. Would dividing the water channel into two seperate 5x12 channels (seperated by a 2mm cooling fin bored out of the base of the block) be effective at drawing heat away from the base, and also properly cooling the block at the same time?

3. Finally, these questions are more suited to other forums, but i would like to keep this in a single post.
a) How effective are 2x 120watt TECs vs. a single 226watt TEC? (Forget assumptions that 240 is better then a single 226watt TEC, im talking in terms of powering, surface area, and cooling capacity)
b) Are there any articles about building DIY power supplies involving toroidal transformers or rod transformers? I know that running 2x AT PSUs in series seem to be quite popular, however, im trying to fit all of this in a single case, and hoping to take up as minimal space as possible (e.g. using a surplus ATX PSU case).
c) Another power supply question, could you supply any links to 12/240v AC relays to power on pumps when a PC is turned on? Looking through this forum's previous threads, i have had little luck.
d) Memory/Northbridge active cooling. Would actively cooling these components allow for a higher overclock via the FSB? I've already got some plans for this, aside from insulation, im worried about minimal gains.
e) What seems to be the stepping of choice in the US with XPs? i've heard a couple of good things about the AGKGA stepping (XP1600s) but very little of other steppings.

Thanks

P.S Heres the links for the plans for the block above (forgive its simplicity)

http://bravo.ausgamers.com/TypeI/channel.jpg
http://bravo.ausgamers.com/TypeI/crosssection.jpg
http://bravo.ausgamers.com/TypeI/lid.jpg

[Bravo]

Bump?

nobody?

[Diggrr]

That's one beautiful block!

Yes that's a great design, with the inlet directly above the cpu die.
It brings the coolest water to the hottest part of the block.

Yes, adding a 2mm fin down the center of the water channel would add nearly 1/3 more surface area exposed to water, thusly increasing it's capacity to remove heat.

And thirdly, please don't say you are considering stacking the peltiers...not a good idea. I'd rather go with side by side peltiers on a larger block. Stacking would burn the top most one out first because it would have a heatload on it equal to the processor plus the first peltier. The lower one would then shortly burn out because of insufficient cooling. A side by side configuration at the wattages you posted would be a safer alternative. If one peltier were to go bad, your processor at worst would lock up (if overclocked harshly) or simply raise temperature, allowing a safe shutdown. (Set the bios temp alarm, or use a digidoc with alarm)

Very nice workmanship on the block, I like the removable top idea as cleanout is very difficult on the sealed units.

*edit..sorry* Welcome!

[Stephen Castles]

if your gonna make a CPU block foir side by side pelts, just make sure channels cover the block so you are cooling all of the peltier. next, make sure the channels are just as big as the 1/2 inch tubing, if not, you will see a huge pressure drop.

[Bravo]

Diggrr, thanks :) While i wish i could take credit for production of the block, it was jointly produced with a mate back home. With regards to peltier stacking, i wouldnt do this because of the sheer cost of TECs, at the moment, im opting for a single 226watt TEC.

Stephen, using some grade 8 math, i've calculated the area of the inner diameter of the 1/2" barb (0.5x3.142x(6.25x6.25)=266.5mm squared) to compensate for it. Basically the block had as wide channels and height as was physically possible, in this case 11x12.5 (137.5mm squared). While i do expect a pressure hike, because the flow of the pump will be equally divided between two circuits (GPU + Northbridge + System Memory and the CPU).

Thanks :)

[Stephen Castles]

i've got a design in the making that i have never seen before, i am not going to release the design yet. If it's a piece of crap, yeah, i'll post it up here, but if it's a good performer, i may want to patent it. I havent decided yet.

[Bravo]

Originally posted by Stephen Castles
i've got a design in the making that i have never seen before, i am not going to release the design yet. If it's a piece of crap, yeah, i'll post it up here, but if it's a good performer, i may want to patent it. I havent decided yet.

Hehe, sounds secretive :)

A guy on the OCAU forums back home introduced this block made of 2mm thick pieces of copper, basically, the water was pumped into a cavity, and pushed through 14x 2mm spaces and into another cavity, where the outlet was located. Pretty good design, but hard to produce :(

[Stephen Castles]

got pictures of it ?

[Bravo]

Originally posted by Stephen Castles
got pictures of it ?

i will in a minute.....

[minute later]

Like most of my better ideas, possibly it's been done somewhere before, but this is my latest heatsink or waterblock or both design. Please excuse the dodgy drawings, but as you can see, it's a laminated design. It incorporates considerably more surface area than the current milled blocks, with what should be a reasonably good rate of flow. It requires nothing more than basic hand tools but you will need a good heat source to make the soldering joints sweat together. By varying what you cut out, you can make it either a waterblock, a heatsink or a combined unit. For a straight heatsink, cut part A and Part B off level with the top of Part B's coolant void. For a straight waterblock, cut the top of part A off level with the top of part B.

http://members.ozemail.com.au/~flyingwing/heatsink.gif

Don't use copper sheet that has been bent or recycled. The pain is not worth it. I got three kilos for 10 dollars at a recyclers and it is straight enought to squeeze together easily.
Please note that the drawing is not to scale and is only to impart the concept of construction. The rest is up to you. Modifiy it to put pipes and mounts where you need them.
How to mount it is also your problem as I can't account for all situations. A straight waterblock can be done as most are mounted, and the straight heatsink design can be modified to let a standard heatsink mounting design work too. Heatsink/waterblock combinations will be more interesting but I'd suggest making use of the board holes if you're an AMD man. The alignment bolts will let you hang brackets for mounting attachments also, so make up something reliable! It would not be good if your kick ass heatsink fell off. It is probably going to be reasonably heavy too, if you make a combined block/sink...
I would also strongly suggest making a mock up in cardboard or some other substitute so you can check your sizes and clearances for pipes and mounts etc.
Mine will have half inch inlet pipes and the top of part A will be flared up to 80 mm to take a nice QUIET panaflo. It will also use .6 mm copper sheet for all construction apart from the top fan mounts which will be slightly thicker.
Keep in mind that you'll need at one Part A with only the alignment bolt holes if you're making a waterblock. If you intend to pressure test it with mains pressure, I'd suggest making that one piece out of at least 1mm copper.
The inlet and outlet pipes would be made of copper and I would suggest that you make them go full depth to seat against the other side for strength when you're attaching hoses, and cut or grind half of the wall of the pipe away so that the coolant can escape easily. The pipe comes back to full wall just before it exits and is soldered in place. If you decide to use Brass Barbs, make the first Part A out of at least 3 mm sheet. (I wouldn't recommend it though)
I was initially concerned with the heat transferance ability of solder joints, but the latest copper heatsinks with soldered fins made up my mind that it wouldn't be a problem as they currently are testing the best. The edges of the copper plates will be touching the processor core anyway, so the solder joints aren't so important.
You will need to carefully clean, flux and tin the whole of the mating faces of part B and Part A with solder in a THIN layer. I would suggest scouring them gently with a brillo pad or something similar to get them a bit shiny, and then handle them by the edges until they're tinned. Assemble the tinned pieces with the alignment bolts, and then using two bits of sacrificial material, like wood, clamp them together in a vice and hit them with the blowtorch. (The wood is to stop the vice acting as a heatsink. Keep a wet rag handy to put the wood out, or the paint on the vice for that matter)
When it is hot enough you'll be able to tell because solder will melt on it and squeeze out as you tighten the vice up. Give it a bit more heat for insurance. Pretend that you're Chopper Reid and the heatsink is the drug dealer. When you're happy, let it cool. Now would also be a good time to install the inlet tubes before it gets too cold.If they need to be bent, I hope they have been bent before you put them in... Just make sure your grind outs are facing each other before you solder them. Making your inlet tubes longer than you need may help you during the testing phase. You can always cut the longer bits off later. Let the block cool naturally. You've just laminated about 70 or more bits of metal. Tossing it in a bucket of water can do bad things. Trust me....
Test it by jamming one side onto the garden hose and clamping it with a hose clamp. If you manage to crush the pipe, it's okay because you made it extra long like I suggested.....
Block off the other end with another bit of hose, or something similar and give it mains pressure. Fix any leaks with the torch and more solder. Eventually it will tell you where it's hidden it's drugs..
When you're happy that it doesn't leak, now it is time to sand and polish the bottom. Now your careful work in cutting the pieces accurately will pay off.
Before anyone asks, you can't do it with aluminium. Well, you can, but solder won't work and it gets a lot more painful to achieve.
BE careful drilling copper sheet. The normal drill angle is 110 degrees which works well for steel, but it needs to be from 80-60 degrees for copper. Copper is well known for its grabbing ability in the drill press if not secured. You could end up being more like Chopper than you bargained for. Be safe and have fun with your blowtorch

While the above design is a hybid heatsink/waterblock, this picture best explains the surface area inside the block

http://members.ozemail.com.au/~flyingwing/waterblock2.gif

For the rest of the info, visit this thread:

http://forums.overclockers.com.au/showthread.php?s=&threadid=14577

Cheers