Built a Peltier Block with a Swiftech Apogee GT

[initialised]

I bought this:

It was covered in some green gunge, probably a combination of algae and copper sulphate.

I cleaned it up with a dip in some Cillit Bang.
This etches the surface the redeposits the copper
Then passivated the water contact surface using a blowtorch.

This forms red copper oxide and should act as a barrier to electrochemical corrosion by forming a thin surface layer without hampering thermal performance.
I put it all back together and lapped the base to 1200 grit, cut a 50 mm x 50 mm coldplate (only 2mm) and lapped this. I find a rubber glove stops it slipping.

Applied Arctic Silver Epoxy to the base of the block and did the same with some Ceramique mixed in on the cold side and clamped it using a spare acrylic top from and Alphacool block.

The Peltier is a Qc=226, Pmax=350

[NaeKuh]

algae and copper sulfate????

How could one ignore there loop maintence that much for that massive quanity of algae to get stuck on the block pins. I would like to guess more towards corrosion??

Do you have better pics?

[littleowl]

I would also like to see some more pics!

[Jigsaw]

... you lapped a Swiftech block?

[pH(x)]

I think that is the coldplate which he shows in the picture that he is lapping.

[Jigsaw]

I hope so.

[littleowl]

... you lapped a Swiftech block?

I think that is the coldplate which he shows in the picture that he is lapping.

if you read above the pic he said he lapped the base as well as the cold plate.

[Jigsaw]

That doesn't make it any better, there is no way he got that Swiftech block flatter than it was originally, there is not good that can come from lapping an Apogee.

[initialised]

algae and copper sulfate????

How could one ignore there loop maintence that much for that massive quanity of algae to get stuck on the block pins. I would like to guess more towards corrosion??

Do you have better pics?

No, sorry, but some more waiting to be uploaded on a 6mp Kodak rather than a 2mp Nokia (which I killed while stripping the solder of a radeon 'cooler').

I'm sorry, If i had of noticed I would have cleaned it up before I sent to you. This water block was the first time I used water cooling... I had it in the loop for about 2 days then some water got on my mobo and killed my pc so I never used again (The leak was not in the water block I sold you, that's fine.). I think i used distilled water and anti freeze or something although this was ran from an open tank of water so i suspect this is why it may be dirty inside.

Doesn't really explain much, there was no damage to the 'diamond pin array'. There were some strands it there, maybe hair, algae or PTFE. I suspect that these got stuck on the fins and slowly accumulated any particles that went through.

Would the bow that helps the thermal performance on CPUs with heatspreaders not reduce the surface area in contact with the TEC? Or is it a case of my bad?

Anyway, here's the rig I've knocked up to cool it:


226W Qc, (350W) TEC, 500W auxiliary PSU, white neoprene from a hiking bottle case, radiator from VAG motors, three loops one 3000l/h pump.

[Jigsaw]

Oooh! Nice radiator! I have been trying to pick up a transmission cooler but I don't have $500+ to toss around right now.

[initialised]

Oooh! Nice radiator! I have been trying to pick up a transmission cooler but I don't have $500+ to toss around right now.

£5.00 + £6 shipping, stacks of these cheap on ebay. The pink goo is more for leak testing than permanent, I'll get some shots of it under UV (room not PC), IIRC it glows nicely.

Learnt a neat trick for minimising the need for step down adapters while building this: Cut the tube diagonally, heat it with a hair drier/heat gun (no flames though) then you can stretch 12mm ID over a 19mm barb and as a bonus really tight fittings so less need for hose clamps. Anyone know if 19mm barbs are available and where?

[Jigsaw]

How does it compare to a PA120.3?

[initialised]

How does it compare to a PA120.3?

My pump wouldn't be able to handle it.

I tried it on a loop terminating with a single 10mm tube. It wasn't happy, sounded overloaded. Also Aluminium tube'n'fin is better than brass tube'n'copper fin.

[TopherTony]

wow you just love the pink stuff dont ya
anyways nice rig post some temps when you get a chance

[initialised]

After running the system overnight I thought I'd check the potentials developed between the copper connectors and blocks and the aluminium radiator.

Pipe connector - radiator:


CPU block - radiator:


The readings are both in mV. If I don't control this I am going to get electrochemical corrosion. I guess this is the counter argument to using aluminium radiators. Hopefully a few ground straps will sort it out.

[initialised]

wow you just love the pink stuff dont ya
anyways nice rig post some temps when you get a chance

The van with my Q6600 in it pulled up, looked confused and drove off. Muppet.

So I guess tonight I'll be abusing my E4300 instead. I had it up to 380x9 (3.4GHz) on a TEC before, will be looking for a little more. Will see if I can get some temps (and charge up my camera).

[initialised]

I checked their tracking system again today, another failed delivery allegedly while I was outside. Didn't see one of their vans.

Anyway, it's probably far better that I cut my teeth on my P4 or E4300.

I wired up all the copper components to the radiator no voltage, no galvanic corrosion. Then solve the leaks this caused. Then I thought what happens to a battery when you short the terminals. It heats up. So there may be a trade off here: system reliability vs operating temperature.

I added a temperature probe to the system. When running with no load (no TEC) the system got to around 26C (ambient ~19C) that'll be heat from the pump. With the system wired up temperatures weren't noticeably affected. Will need a longer run (tonight perhaps) to get a more concrete picture. I'm thinking 2hours each way. Fortunately the lady of the house actually likes the noise the system makes so there's no problem running it 24/7 (apart from the electric bill 140W pump, 1350W of PSU, PC will probably use 500+ of this).

Anyway, something really annoying happened when I turned on the Peltier at 13.8V. The cold side got cold then slowly heated up from one end (-ve terminal). One end of the cold plate was hot and the other was cold. I swapped flow direction to see if this was the cause. No difference. I removed the passivation layer from the wet side of the water block. Slight improvement. Loosened the bolts slightly, improvement, but not enough to keep it cool at 12V. Dropped it to 5V, it stayed cold, but that's around 25-30W of cooling, not enough for a C2D. Blocking the other loops in the system helped a bit but I think that I need a block with less flow resistance and some air cooling on the radiator.

[initialised]

Did some digging and a little more testing. If I press hard on the cold plate it cools down. Not convinced about freezing the CPU with it just yet. What do I use to get the AS epoxy off acetone?

On the corrosion inhibition vs temperature issue:
With no load the system runs at 26.4C with the copper components wired to the radiator this goes up to 26.7C so there is a trade off but it's not a big enough heat load to worry about.

[littleowl]

Did some digging and a little more testing. If I press hard on the cold plate it cools down. Not convinced about freezing the CPU with it just yet. What do I use to get the AS epoxy off acetone?

On the corrosion inhibition vs temperature issue:
With no load the system runs at 26.4C with the copper components wired to the radiator this goes up to 26.7C so there is a trade off but it's not a big enough heat load to worry about.

yuck epoxy! I would say you will have to lap it off with sand paper. that is how I got it off of a heat sink I was playing with

[initialised]

OK, just run through CustomPC's Multimedia Benchmark achieving a rather disappointing 1098. Never mind, I'm more interested in the thermals right now.

Following from previous posts I got the Apogee/Peltier block mounted with sufficient pressure for it to keep the CPU cool at 12V by making a back plate out of two CDs. Clocked it back up to 3.0GHz. It's idling now at 21C, it peaked at 63C but came down very quickly afterwards. Unfortunately the water in the system is at ~54C and the result of this is that my HD2900XT (modded heat spreader, Danger Den Maze 4) is running at 67C (and hasn't come down), it was idling at 40C before I ran the tests. The card throttles at either 95C or 115C so it doesn't leave enough headroom for gaming.

I dug out my 62mm 350/540W Qc/Pmax TEC and am thinking of ways to use it on the card. I think I need the baseplate from a reference cooler to use as the coldplate but what can I use to cool the hot side? I have an over lapped DD Maze 4 on the card now so maybe that's a contender.

I put some fans blowing at the radiator, down to 8C on the CPU, 36C in the water, 48C on the GPU.

[littleowl]

initialised are you running all this stuff on your e4300?

I don't think it is a good idea to use a 62mm tec on the video card because you will end up freezing it. I would say make a 62mm water block for the cpu with custom hold down and cold plate. Then use the 226w one on your video.

I don't recommend using a maze4 on any tec that is over 226W. The maze4 design is not that great for tec cooling.

I would recommend looking for a fuzion water block. even though it is still only 50x50

[initialised]

initialised are you running all this stuff on your e4300?

I don't think it is a good idea to use a 62mm tec on the video card because you will end up freezing it. I would say make a 62mm water block for the cpu with custom hold down and cold plate. Then use the 226w one on your video.

I don't recommend using a maze4 on any tec that is over 226W. The maze4 design is not that great for tec cooling.

I would recommend looking for a fuzion water block. even though it is still only 50x50

I'm pretty sure I can mount the Swiftech block on the Radeon and will do if it's not enough for the Q6600 (which I picked up today after it was dumped in an alleyway and handed into a local police station). I'll do it using the mounting plate I pulled off the Radeon during it's destruction.

Anyway, my E4300 has been idling as low as 5C and temps on the GPU have been improved simply by pointing two 80mm fans at the radiator, they're keeping the water below 40C under load. When I dig out my pair of Sunon monsters (don't know where I hid them) this should improve further. On a slight downer (and against littleowl's advice) I have been running the Peltier off a spare PCI-E lead, the only problem that's cropped up is that the insulation on the wires near the terminal block and the terminal block's insulation has melted a bit so I may have to get a ceramic cased terminal block and use four wires rather than two. No power issues so far though so it looks viable for now.

[initialised]

http://cgi.ebay.co.uk/MCW60-B-GPU-Wa...QQcmdZViewItem
or this
http://cgi.ebay.co.uk/D-TEK-Customs-...QQcmdZViewItem

[littleowl]

http://cgi.ebay.co.uk/MCW60-B-GPU-Wa...QQcmdZViewItem
or this
http://cgi.ebay.co.uk/D-TEK-Customs-...QQcmdZViewItem

I like the D-tek myself!

I'm pretty sure I can mount the Swiftech block on the Radeon and will do if it's not enough for the Q6600 (which I picked up today after it was dumped in an alleyway and handed into a local police station). I'll do it using the mounting plate I pulled off the Radeon during it's destruction.

Anyway, my E4300 has been idling as low as 5C and temps on the GPU have been improved simply by pointing two 80mm fans at the radiator, they're keeping the water below 40C under load. When I dig out my pair of Sunon monsters (don't know where I hid them) this should improve further. On a slight downer (and against littleowl's advice) I have been running the Peltier off a spare PCI-E lead, the only problem that's cropped up is that the insulation on the wires near the terminal block and the terminal block's insulation has melted a bit so I may have to get a ceramic cased terminal block and use four wires rather than two. No power issues so far though so it looks viable for now.

This is what happens and if you don't notice it and leave it run to long it will catch of fire!! After that you can say bye bye pc and if you having a very unlucky day then you may have to say bye bye house!
Please watch it very close! You really should look into getting another meanwell!

[initialised]

I like the D-tek myself!




This is what happens and if you don't notice it and leave it run to long it will catch of fire!! After that you can say bye bye pc and if you having a very unlucky day then you may have to say bye bye house!
Please watch it very close! You really should look into getting another meanwell!

It'll be fine, I've got enough headroom on the PSU to run the 226W at 12V so theres no need to get a third PSU. It seemed like it was the terminal block that was heating up, it should be good for 30A but apparently not. As it heats up it's resistance increases so the current goes up and it heats up more, positive feedback. I just need to get some better cabling and use a stronger terminal block and maybe strap a heatsink to it.