A noiseless system for computer cooling
The HTD Laboratory has developed and successfully tested noiseless (fanless) cooling systems for the central processors of desktop personal computers. The systems were created on the basis of Loop Heat Pipes. Processors Intel P4 with a clock rate of 2.8 GHz and Athlon XP 2500+ with a clock rate of 1.83 GHz were cooled. The capacity dissipated by the processors mentioned above at the maximum load was about 70 W. The total mass of the cooling systems, including the LHP and the radiator, depending on the modification, was from 0.85 to 3.2 kg. The total thermal resistance of the system was in the range from 0.5-0.8 К/W.
When the systems were tested as parts of computers by a test program providing the maximum loading of the processors, the temperature at their contact surface was in the range from 64-77 ºС at an ambient temperature of 26 ºС.
During simulation tests of the cooling systems temperatures of 82-95 ºС at the contact surface of a thermal simulator of the processor were achieved at capacities of 110-120 W.
Copper-water miniature loop heat pipes
About 20 different variants of copper-water miniature loop heat pipes (MLHPs) with a cylindrical and a flat evaporator intended for cooling promising CPUs of portable computers were developed and tested over the period of 2003.
The maximum capacity achieved by the best specimens by the devices was 160 W. The indicated value is not limiting. The limitation on the heat load was determined by the temperature of the evaporator wall Те, which should not exceed 100ºС. The minimum value of the total thermal resistance Te–Tamb/Q, equal to 0.46 K/W was demonstrated by an MLHP with a flat evaporator at a heat load of 140 W and an evaporator temperature of 88ºС. The maximum heat transfer coefficient in the evaporator equal to 61000 W/m2K and its minimum thermal resistance of 0.048 K/W have been achieved at a heat load of 160 W and an evaporator temperature of 98ºС.
The thermal characteristics of the devices are given in Fig.1 and Fig.2.
The table gives the main design characteristics of MLHPs:
| Characteristic | Cylindrical evaporator | Flat evaporator |
| Evaporator diameter, mm | 6 | -- |
| Evaporator thickness, mm | -- | 3,3 |
| Evaporator active-zone length, mm | 20 | 20 |
| Evaporator interface area, cm2 | 4,0 | 3,6 |
| Condenser length, mm | 61 | 61 |
| Vapor-line diameter, mm | 2,5 | 3,0 |
| Liquid-line diameter, mm | 2,5 | 2,0 |
| MLHP effective length, mm | 290 | 290 |
Fig.3 and Fig.4 present the external view of copper-water miniature loop heat pipes with a flat and cylindrical evaporator.
A cooling system for CPU of a desktop PC was developed and tested successfully in the laboratory conditions. The system was created on the base of a copper-water loop heat pipe (LHP) equipped with a flat-oval evaporator 7 mm in thickness and with lines for vapor and liquid 4 mm in diameter.
An aluminum radiator, the sizes of which correspond to the sidewall of the system block body “midi-tower”, was used as a heat sink (Fig. 5).
In the passive state at an ambient temperature of 22ºC the cooling system held a temperature of 70ºC on the thermocontact surface of the CPU thermal simulator at a heat load of about 90 W.
In the active state, when the radiator was blown by a low-noise fan, a temperature of 70ºC was reached at a heat load of 160 W.
At present in the Laboratory of Heat-Transfer Devices a new compact cooler AC-1 for desktop PCs created on the base of a copper-water LHP is being tested. An axial-flow fan (92x92x25 mm), which has an flow rate of 40 CFM and creates a noise at the level of 25 dBA at 2200 RPM, is used in the cooler. The cooler sizes are 100x100x75 mm, its weight is 585 gram.
The cooler was tested with a CPU thermal simulator at an ambient temperature of 22ºС.
The results of tests are presented in the table.
| Heat load, W | Temperature on the LHP thermal interface, ºС |
Temperature on the simulator thermocontact surface, ºС |
| 50 100 150 200 250 |
36,7 40,4 47,1 54,5 62,8 |
37,4 41,8 49,1 57,1 66,2 |
The system total thermal resistance “thermocontact surface of the CPU simulator - ambient air” decreases from 0.3 up to 0.18ºС/W at a heat load increase from 50 up to 250 W. No signs of a heat-exchange crisis were observed at a maximum value of 250 W reached during the testing.
Comparative testing of a new cooler and a production run aluminum body cooler Intel/Sunyo Denki supplied with a fan with a number of rotations of 2700 RPM was conducted at the Ural computer engineering plant (Ekaterinburg).
The testing was carried out outside the body of the system unit in the combination of a motherboard ASUS P5LD2-VM/LGA 775 Socets with a two-core CPU processor Intel Pentium®4 640 (3,2GHz/2M) at an ambient temperature of 27,2ºС. A testing program S&M was used for testing.
Testing results are presented in the table.
| Cooler | Tcore1, ºС | Tcore2, ºС | Tmb, ºС | Tev, ºС |
| Intel/Sunyo Denki AC-1 |
56,7 51,0 |
47,3 47,5 |
38,3 34,4 |
- 50,0 |
Tcore1 is temperature of the 1st processor sensor
Tcore1 is temperature of the 2nd processor sensor
Tmb is temperature of the sensor on the motherboard
Tev is temperature on the thermal interface of the LHP evaporator.
It follows from the comparison of the indicated temperature values with the results, which were obtained earlier during the testing of a new cooler with a CPU thermal simulator, that the maximum power dissipated by a real processor makes up a value of about 130 W. Testing with a real processor demonstrated that the new cooler reduced the temperature of the first CPU core by 5,7ºС and the temperature of the motherboard – by 3,9ºС in comparison with a run production cooler. The temperature of the second processor core remained approximately at the same level at this.
The comparative testing of the cooler AC – 1a and a cooler “Cooler Master” (Taiwan) supplied with aluminum finning with the area of 3000 cm2, three heat pipes 6 mm in diameter each and 92 mm fan PL92S12M-5 was conducted. The sizes of “Cooler Master” with a casing are 145x125x100 mm, its weight is 804 g. The coolers’ testing was being conducted with a CPU heat simulator and with a fan “Cooler Master” at 1800 and 3200 RPM at an ambient temperature of 22ºC .
The results of the tests are represented in the table which shows the values of the temperature measured on the thermocontact surface of the CPU heat simulator with varied heat load changing within the range from 50 to 250 W.
| Heat load, W | 50 | 100 | 150 | 200 | 250 |
| 1800 RPM | |||||
| Cooler Master AC – 1a |
36,8 35,3 |
50,7 46,2 |
64,7 57,2 |
78,2 67,5 |
91,7 77,8 |
| 3200 RPM | |||||
| Cooler Master AC – 1a |
33,1 32,6 |
42,5 39,7 |
51,9 46,5 |
60,5 52,9 |
70,0 59,9 |
Within the range of heat load from 100 to 250 W at 3200 RPM the total thermal resistance of “Cooler Master” was changing from 0, 20 K/W to 0,19 K/W. For AC – 1a these values were making up 0,19 K/W and 0,15 K/W accordingly.
 |
External view of the cooler AC-1a
The tests of the modified cooler АС-2а with the 120mm ZM-3F fan were conducted. The cooler’s mass is 380g. This cooler as well as the AC-1a is designed on basis of copper-water LHP with the flat-oval evaporator of 7mm thick. As a heat source the CPU heat simulator with the 32x32mm thermal interface was used.
The tests were conducted both under the normal number of fan rotations 1800RPM with the noise of 34dBА and under the reduced number of rotations - 900 RPM with the noise level of 20dBА.
Test results are presented in the table.
| Heat load, W |
|
|||||||||||||
| 50 | 41,3 | 41,5 | ||||||||||||
| 100 | 46,1 | 46,5 | ||||||||||||
| 200 | 60,1 | 64,3 | ||||||||||||
| 300 | 74,8 | 82,1 | ||||||||||||
| 400 | 86,0 | 95,0 | ||||||||||||
| 500 | 98,1 | 110,8 | ||||||||||||
| 550 | 107,8 | - | ||||||||||||
| 600 | 117,1 | - | ||||||||||||