Guess you can cool an i7 with TEC's

[Elloquin]

Set Up:
Ultrasonic2 Dual 50mm Direct Die TEC Block
2 x Custom Thermoelectric TEC's : 226 Couples 331wQmax 17amp 31.5v @ 23v. PSU: Mean Well 600w 24v 25 amp. Cooling: 4 x Black Ice GTX 240 Rads, 16 x High Speed yates Push/Pull, Iwaki RD 30 24v Pump.

Ambient:



Base Shot:

Cheapo Laser Thermo doesn't go any lower :P Thermocouples will be added later to the hot and cold plates.

Prime Blend 1 Hour Hyper Threading Enabled.
i7 920 D0 @ 4.2 1.2Vcore 6 Gigs Gskill Trident Classified 759.


Prime Blend 1Hour Hyper Threading Disabled.
i7 920 D0 @ 4.2 1.2Vcore 6 Gigs Gskill Trident Classified 759.


Test Rig:

[QuadDamage]

Nice but can it handle a real o/c? Like 4.4Ghz+ and HT?

[Splave]

awesome! very very cool. Let see some big numbers!

[Elloquin]

@ Quad Dammage

I can run it for you this weekend. But honestly if you think 200mhz and an extra .2Vcore is going to make it magically crash and burn

Best I can do at the moment is LinX on 2 gigs Corsair Dom GT's. Just shy 1.4Vcore. The mount wasn't very good and was corrected for the Prime runs.

Hyper threading ON.

[Elloquin]

awesome! very very cool. Let see some big numbers!

Going to try and work on it over the weekend Bro. I wanted to pull it apart and drill the holes for the thermocouples this weekend but that can wait. Been a while and I am rusty So far boots 4.8 no problem. But that's worthless without pics.

[QuadDamage]

I didn't say it would "crash and burn" my point was 4.2Ghz and No HT with 1.2V is easily achievable on clean, average air ie Zalman CPNS9900.

I was curious if it can handle real o/c since i used to run a MCW6500-T with stock Swiffy block on an i7 at 4.5Ghz but it wouldn't handle full load.

[Elloquin]

Not sure what you mean here....First prime run is with HT enabled second is with it off. LinX run has HT enabled as well. As far as comparing it to air check the temps in the screens the cores are sub ambient Like i said i'll fire it up at 4.5 and rerun the test maybe I can get it done tonight. Although I don't expect much more than a 5c max increase in temps.

[imersa]

Will be nice to see the results. subbed

[Ultrasonic2]

Thanks for that man (Elloquin)




Ultrasonic2.com

[malkiewicz]

Thanks for that man (Elloquin)

http://thermo-electric-cooling.com/pic/50x2v2/50-7
http://thermo-electric-cooling.com/pic/50x2v2/50-8
http://thermo-electric-cooling.com/pic/50x2v2/50-6
Ultrasonic2.com

I had no idea you sold blocks.... Looks like your webpage is coming along nicely...

I think the blocks would look a little better if you used different screws.... Just a thought....

[Elloquin]

[malkiewicz]

How long did you run prime before recording temps? If you havent already I would let it run for at least an hour...

[Ultrasonic2]

He's been doing a proper job of testing. ( EDIT though i see that was for 30minutes )

I realises this is hard to believe but the block actually works LOL Mind you it should do with 2 of those TEC's in there

[Vampiyer]

Those results are pretty good

What's the power consumption like at full load? Also - may as well undervolt at idle, right? What are the idle power consumption numbers? It may not be too bad overall - I don't think anyone would run 100% load 24/7 on a setup like this.

Hmm... I guess we can expect a TEC-chiller on the site soon then too?

I'm not interested in anything right now, but if I had to choose one route, it'd be chilling the water. The expense and physical size is considerably greater, but I think that we could achieve similar temps at max load (4.5Ghz saw a rise in temp) with lower power consumption.

[Elloquin]

How long did you run prime before recording temps? If you havent already I would let it run for at least an hour...

The temp monitoring has always started prior to the start of the stress test. All diagnostic software is...just the way I do things. So the min temps are the temps on a cold boot after the block was allowed to "pull down" for 5 minutes. That is why you will see -22c as the min IHS temp and -8c as the min core temp (Core temp won't read below -8c. I think Eleet does but haven't gotten around to messing around with it yet.) The OS was shot as I had done several runs at different voltages be it Vcore, CPU PLL vcore, and VTT. I think I have found a happy sweet spot now. So yes I will obviously be running this again for an hour maybe more with screens. That run actually ran for 59 minutes and then blue screened lol so no screen shot. But just as with the other tests it really didn't change the max peak core temps were somewhere around 10 minutes in and then the cores never reach that level again. I really wasn't looking for a stable 24 x 7 overclock at this point just to answer the question at hand. "What happens when you run the block on a good overclock like 4.4+" I was a little surprised the one core went up that high TBH. It could be any number of things. At some point I have to readjust the TECs as they are slightly off center to one side. This is placing them too close to the mounting screws which will add thermal transfer through the screws to the hot and cold plates. How much dunno really just have to do it retest and see It could also be the increased stress in the IMC warming up the adjacent cores sensors. Which is why I started trying to finesse the voltages. I could just dump a load of vcore into it and brute force it but that wouldn't be fair to the block or the chip

At some point the LinX purists will start chiming in and that test is very much temperature affected by how hard you stress the IMC. Which opens up a whole other conversation I am not ready to have yet

[Ultrasonic2]

primes more more heat than you ever get using the PC normally so im good with it

[Elloquin]

Those results are pretty good

What's the power consumption like at full load? Also - may as well undervolt at idle, right? What are the idle power consumption numbers? It may not be too bad overall - I don't think anyone would run 100% load 24/7 on a setup like this.

Hmm... I guess we can expect a TEC-chiller on the site soon then too?

I'm not interested in anything right now, but if I had to choose one route, it'd be chilling the water. The expense and physical size is considerably greater, but I think that we could achieve similar temps at max load (4.5Ghz saw a rise in temp) with lower power consumption.

I do have a plan to make a chiller with two of the same TECs. I would like to try and send chilled water to the TEC block and see what that nets. As it is Ambient rules the day here as the water temps from the rads being cooled by the mid range fan setup I am using is the biggest weak point of the setup. But at least I can hear myself think

As far as power consumption I will need to grab a kilowatt meter at some point. As it is the only information I can add at this point is the fact that the TEC's are running at 23v of 31.5 max ~73%. I tried voltages down as far as 22 and high as 25v which didn't net much of a difference in temps maybe 1 c or so with 25v. This is on a single 600w 24v 25 amp Mean Well rig and all connected to a 20 amp circuit in my living room lol (so yeah the tv is on the same line as a few other appliances ) The Mean Well is not struggling and the breaker doesn't pop under testing so i'll need to get a proper ammeter as well if these questions are to be answered properly. The Mean Well has a 2 wire sense connection on the front which in theory can be used to switch the outputs to the TEC's on and off which reminds me to PM someone about it Then you could run it 24 x 7 at a lower cost.

As far as costs go Most people already have water cooling before they get into this so I will rule that out. The block is up to Ultra for what he sells it for but is is significantly less than 200 USD add TECs (I use expensive TECs) 130 USD Mean Well 100 USD or less of of ebay. So the basic setup is ~300-400 USD depending how resourceful you are.

Lastly, there is the ease of setup to consider. Insulate the socket mount the block and roll.

I will say this the last thing you should be trying to save and skimp on here are the TEC's themselves.

[Elloquin]

primes more more heat than you ever get using the PC normally so im good with it

Stalker Just kidding...

[Ultrasonic2]

Stalker Just kidding...

yip . i have no life beyond TEC's

[Elloquin]

32M 4.8 Fresh OS imaged it yesterday no tweaks yet.



Things really started to warm up chasing this lol. Vcore is too high lol something else is holding back stability...I'll get to it later...

5G's

[npk]

Nice oc

Where did you get the tec´s from?

Do you know what your water temp is at full load? (even a rough guess will do) if you know your water temps you can quickly realize what´s holding you back i.e. good water temps= block or tec´s reached its limit.

[NotSoCoolJ]

I like it. When I was first reading it I didn't really think it would need four rads, but I guess it does.
Now you just need to build a controller so it doesn't drop to -8 at idle. That's wasted energy. Plus it's not that great for the chip to be going from -8 to +24C then back all the time. Less chances of condensation.. all of that.

[hipro5]

Nice going!....

TIP:
A thing that all guys that are testing Peltiers to take care off for BETTER temps is:
The screws that are holding the cold cooper with the hot cooper (waterblock) MUST BE MADE OF PLASTIC not to transport temperature from the cold surface to the hot surface and visa-versa.
By screwing those two plates with metal screws, you're LOOSING a LOT of peltier's efficiency....
Keep this in mind....

[Ultrasonic2]

Nice going!....

TIP:
A thing that all guys that are testing Peltiers to take care off for BETTER temps is:
The screws that are holding the cold cooper with the hot cooper (waterblock) MUST BE MADE OF PLASTIC not to transport temperature from the cold surface to the hot surface and visa-versa.
By screwing those two plates with metal screws, you're LOOSING a LOT of peltier's efficiency....
Keep this in mind....

that is all true but nylon or polycarbonate screws aren't strong enough . which is why as far as i know no one uses them. i also tested them and they were a fail

[TJ TRICHEESE]

Elloquin, which meanwell psu are you using to power those TEC's?