DrRom’s 172 Watt Peltier Cooling

ED Note: What follows is DrRom’s successful supercooling of a T-Bird 1100, with detailed pictures:

Cases

All built, and ready to travel 🙂

UPDATE 2/6/01:

Everyone has been asking me what gear is in my system, so here goes:

  • T-Bird 1100
  • Abit KT7A
  • 2x 128MB Kingmax PC150 Rev 1.2
  • IBM 30 Gb 75GXP
  • Pioneer DVD-105s 16x DVD
  • Generic GF2MX – just getting me through until the GF3 release
  • SB Live! DE5.1
  • 2 x Intel EE Pro/100 NICS

The next two most asked questions are: “What type of power supply runs the peltier?” and “Where can I buy one?”

The PSU is a 0-30v, 20A switchmode model, manufacturer unknown. It has been on my lab bench for years now. Being that it runs from Australian 240VAC, it would be useless to 99.9% of people reading this anyway.

2/4/01:

I finally got my 172 watt peltier from TE Distributing. The dimensions are 40x40x3.3mm, so it’s nice & easy to insulate compared to the bigger 48 or 50mm ones. However, the downside is that you will need a dedicated PSU (ATX need not apply) to run it due to the voltage requirements. I’m running the 172w with 18v from a switchmode PSU – higher voltages offered no cooling improvements for me at all.

I’m currently running my Tbird 1100 @ 1400MHz, with -24C at the CPU back as I type this. Max temps get up to +5C to +7C at an ambient of 24C.

With ducted A/C, CPU temps drop to -32C to -35C (maintains sub-zero under load) with an ambient around 12-14C; at these temps, the CPU gets up to 1524MHz – Not too bad.

Here’s the setup:

* 172w TE Distributing peltier, 11.4A @ 24v – currently running with 18v from a dedicated switch mode PSU mounted inside the tower. Higher voltages offered no improvement in CPU temp, as the water temp seemed to increase proportionately to the peltier temp decrease.

* DangerDen maze style copper waterblock with a custom CNC machined alloy clamp that uses the 4 holes around the socket – gets *real* good clamping pressure.

* Generic silicon heat transfer compound – Arctic Silver has big problems with the clamping pressure & temps – it ends up forming a “foil” layer on all surfaces which CAN’T be good, as it is adding another layer for heat to transfer through. Waiting for some Danfoss compound (as used in Vapochill systems) to arrive in the next few days.

* Liquid is cooled by a Senfu 16-pass radiator mounted in the front of the case with a home-made fiberglass fan duct that directs the heated air out through the side panel (no hot air in box). Home-made variable 5v – 15v regulated PSU with speed control runs 2x 34 cfm fans for radiator cooling.

The liquid used is a mix of 95% demineralised water + 5% Redline Water Wetter. Other parts in the system are a 1 litre alloy reservoir (inside case) and a Maxi-Jet 750 LPH (~200 gph) pump with a modified impeller wheel. The entire cooling system is self-contained within the case, and is 100% portable – in fact I took it Lanning just last night

With ducted AC, it maintains sub zero under load, and now that I’ve had a bit more patience I’ve seen it drop to -40c at idle 🙂

Here’s what this particular CPU does with various cooling (for the record it’s a Cu 1100 week 51):

  • Agilent Arcticooler: 1200 @ 61C under full load
  • Alpha PAL6035: 1200 @ 60C under full load
  • Water Cooling: 1333 @ 32C under full load
  • Water Cooling with ducted A/C into radiator: 1401 @ 18C under full load
  • 172w peltier + water: 1400 @ 5C under full load
  • 172w peltier + water + ducted A/C: 1524 @ -5C under full load

Ambient is a controlled 24C for all ‘normal’ tests, dropping to 9-12C when ducted to radiator as the 4C cold air is being dumped right next to the box. As you can see, temperature has played a VERY big part in the overclockability of this chip, contrary to even my own earlier AMD chips.

In summary I can say: Supercooling does work on my CPU, and I’d speculate the same for most if not all GHz+ Cu Tbirds.

PICTURE GALLERY

Cases

An overview of the cooling system; nice, tight and totally portable.

CPU

Close up the the CPU used for these tests; note that pads have been removed so as not to impede insulation. The L1 bridges have been joined using a conductive pen and all bridges have been insulated with tape.

Nuts

Nylon nuts have been epoxied onto the mainboard, and it has been coated with a heavy layer of CPL for water-proofing.

Foam

10mm thick neoprene is sandwiched between the mainboard & back plate to exclude humid air from behind.

Gasket

This shows the home-made neoprene insulation gasket that completely covers the socket, down to the mainboard.

More Pix on page 2

DrRom


More pix:

Block

Custom machined alloy clamp for the Danger Den waterblock.

Block

Here we see the clamp installed; hiding underneath is the DangerDen block, 172w peltier and 1/4″ copper coldplate.

Cases

This image shows the front of the fiberglass fan-duct, along with the 1L Alloy reservoir.

Cases

Rear view of the fiberglass fan duct.

Fan Control

0-30v, 20A switchmode PSU that runs the peltier and fans. To the right you can see the blue-top knob that controls fan speed.

Cases

Home-made 5-15v variable voltage regulator for fan speed control.

Performance Data on page 3

DrRom


Even more pix:

Everyone has been asking me what gear is in my system, so here goes:

  • T-Bird 1100
  • Abit KT7A
  • 2x 128MB Kingmax PC150 Rev 1.2
  • IBM 30 Gb 75GXP
  • Pioneer DVD-105s 16x DVD
  • Generic GF2MX – just getting me through until the GF3 release
  • SB Live! DE5.1
  • 2 x Intel EE Pro/100 NICS

The next two most asked questions are: “What type of power supply runs the peltier?” and “Where can I buy one?”

The PSU is a 0-30v, 20A switchmode model, manufacturer unknown. It has been on my lab bench for years now. Being that it runs from Australian 240VAC, it would be useless to 99.9% of people reading this.

CPUID

WCPUID of the system @ 1524 MHz, with -27C at the CPU back.

SiSandra

SiSoft Sandra memory score @ 150 FSB, CAS2.

DrRom

ED NOTE: Many, many thanks to DrRom for taking the time to share with us his fine accomplishment!


Be the first to comment

Leave a Reply