Check out the "Tectanic" (320W TEC + Homemade Waterblock)

[Particle]

Here's my test setup:


SEE THE MACHINE RUNNING (if you want to see how chilly it's running)*:
http://downloads.pcrpg.org/video/computer/Tectanic Running.mkv
(or for those of you who would rather just have an avi)
http://downloads.pcrpg.org/video/computer/Tectanic Running.avi



That's the result of my latest experiment dealing with computers. Last weekend I machined a water block from scratch with a drill press, a Dremel, ultra-strength epoxy, and about fifteen hours of time. (video coming later)



Today my 320W TEC arrived, so I prepped the Athlon XP (yeah--remember those? ) I was going to do the experiment on. After an hour or so, I was ready to roll. The video shows the result of my initial test run.



Running idle the setup pulls about 450 watts, and that's with the TEC only getting about 11V (around 160-180W). I'd give it full voltage, but I've got no supplies currently capable of doing so. I'm using a faulty Antec True550 for it currently. It's got a bad 5V rail (4.1V), but the 12V is fine. It's just sucking a lot of juice.

Total cost for experiment: $31
Total time: About 15-16 hours.

The more time you spend on something, the less it'll cost ya.

I'm not sure how high it will clock yet. I don't really get much time before the power supply's wires are ready to melt, but I do intend to find out one way or another. On air, this chip was known to be stable around 2200-2300MHz.

----------

* If you're having trouble playing that video but would like to see the machine running, you've got two options more or less.

1) Install a player with built-in support for Matroska containers (.mkv files) with xvid video and vorbis audio such as VLC.

And/Or

2) Install a DirectShow filter pack that adds support for Matroska containers with xvid video and vorbis audio so you can play it with DirectShow based players (such as Windows Media Player). Such a filter pack would be CCCP.

 

[Gautam]

Would you mind giving us screenshots and/or a jist of what was in the video?

 

[Particle]

I'll put it into an avi (sorry--I use matroska for all my video productions and I forget how exotic it is some times).

Be back in ~10 mins.

 

[Particle]

AVI version posted. It shows the rig running as well as the temp of the CPU. The BIOS apparently can't display anything below 0C is all I can say.

 

[THE JEW (RaVeN)]

Good to see a resurgence of interest in the cooling technologies "abandoned" by mainstream overclockers. This just may inspire me to build that bong I always wanted to.

 

[datura3]

Nice project I'm interested to see how cold you can go with that 320w peltier. May I ask where you got it from? If it's a 15v then you can get nice meanwell for 50 bucks off of ebay (PM me if you want me to point you to the auction). It may double your cost, but it may also double your performance.

Also are you running that in your bathroom? Multitasking? I'm assuming your just running cold tap water right through the block. Sweet

 

[Particle]

Since I don't have a radiator set up yet, I'm was just using the sink to hold tap water to run with. I recently moved it to a different room:



I got the TEC off of eBay--it's manufactured by Thermal Enterprises. (CP1-12726) I was quite happy to find a 320W unit in 50x50mm form.

If I understand properly, this TEC is rated for 320W of cooling (refers to max voltage at 15V and max current at 30A--450W, so it makes sense that the 320W number would be cooling power at 450W of input [70% efficiency if so]). This should yield temps of around -30C.

For the AMD Athlon XP 2500+ @ stock settings:
Typical CPU thermal output: 54W
Maximum CPU thermal output: 69W

Estimating no more than 100W thermal load after overclocking.

Maximum TEC voltage: 15VDC
Maximum TEC current: 30A
Maximum TEC input power: 450W
Maximum TEC cooling power: 320W
TEC efficiency at maximum: 71.1%
Current TEC voltage: 12VDC
Current TEC current: 17A
Current TEC input power: 204W
Current TEC cooling power: Unknown (145W if same efficiency as maximum, but it should be higher than at max)

Assuming hot junction temp of 80F and perfect insulation preventing heat creep from surrounding matter.

Stock 2500+ @ typical & expected maximum, temperature differential at current power:
Typical: (1 - (54 / 145)) * 69) = 43.3C (-17C / 1F)
Maximum: (1 - (100 / 145)) * 69) = 21.4C (-5C / 41F)

Stock 2500+ @ typical & expected maximum, temperature differential at maximum power:
Typical: (1 - (54 / 320)) * 69) = 43.3C (-31C / 23F)
Maximum: (1 - (100 / 320)) * 69) = 21.4C (-21C / -6F)

Since I run much colder water than 80F and the block seems to remove enough heat to keep itself at the water's temp, I should be able to get much lower temps. Also, I'm using a 1 gallon reservoir which can absorb about 45,000 calories of energy--that gives me about 10-15 minutes of run time before the water heats ~20 degrees to room temp.

 

[||Console||]

Nice setup I love seeing ppl useing tec's as im very interested to use one on my next Gfx card

 

[Particle]

One word of caution--they burn a lot of extra power. If you can live with the increased power bill, by all means. They're fun.

 

[||Console||]

It goes wb>tec>cold plate right ?

 

[Particle]

Yes, more or less.

Water Block
[Thermal Compound]
TEC
[Thermal Compound]
Cold Plate
[Thermal Compound]
Processor Die

Just make sure you test the TEC to see which side gets hot with which polarity before you try it. They aren't usually labeled. If you put the hot side down, you might end up with a toasty core before you know what's happening.

Also, be sure to protect the area around your card by sealing all the components with either an electronics lacquer or dielectric grease, then insulate the heck out of the area as well with a closed cell foam.

 

[||Console||]

I have head that vasoline works good and is cheeper than dielectric grease .

 

[Particle]

You can get enough dielectric grease for the project with about a $5 expenditure.

Coolant on this last run went from 43F to 86F in 38 minutes.