English 
Views: 15 Author: Site Editor Publish Time: 2024-06-26 Origin: Site
The improvement of CPU/GPU computing power and power continues to drive the demand for heat dissipation, and there is an accelerating trend.
According to PassMark ratings, Intel/AMD CPU chipshave continued to increase in single/multi-core performance2001 to 2020.
Meanwhile, according toTechspot's research, much of the arithmetic (or processing power) of CPUs and GPUs is determined by their transistor density, and each transistor generates heat ascurrentpasses through it , so the increase in transistor density brings with it an increase in heat;
Meanwhile, the majority of CPUs on the market are capable of running at speeds higher than their base frequencies.
On the other hand, due to the rapid increase in the demand for AI computing power, the power increase of related CPUs/GPUs also shows an accelerated trend.
Taking GPUs as an example, thepowerGPUs carrying image processing capabilities, which are mainly used in gaming and other fields,increased by nearly 5x to about 450W during the 18-year periodfrom G70 in 2004 to AD 102 in 2022;
comparing to V100/A100/H100 used in the field of AI, the power of V100/A100/H100 used in the field of AI will be increased by 1.6x/1.75x in every three-year interval from 2017 to 2023 to 700W
The evolution of chip-level cooling materials and processes, air-cooled to liquid-cooled.
CPU Heatsink manufacturers produced the first air-cooled way to appear, the early one-piece extruded aluminum down-pressure radiator, aluminum is cheap and easy to process, but the thermal conductivity efficiency (237W/(M-K)) is only one-half of copper (401W/(M-K)), so the emergence of plugged-copper radiator.
In terms of the radiator form, the early downward-pressure radiator, the newer tower radiator can improve the cooling efficiency by means of side blowing
With the increase of CPU power, the radiator began to be paired with heat pipes, fins and other devices to form a higher-performance cooling module, and the emergence of water-cooled radiators with high cooling efficiency wind.
Chip cooling (including CPU/GPU) mainly has two solutions: air Heatsink and water Heatsink. Under the same specification and power consumption, water cooling is more powerful, but expensive, and is used in the high-end market.
As CPU power increased, heat sinks began to be paired with heat pipes, fins, and other devices to form higher-performance cooling modules, and water-cooled heat sinks with high cooling efficiency emerged.
Chip cooling (including CPU/GPU) mainly has air-cooled and water-cooled solutions. Under the same specification and power consumption, water-cooled cooling is more powerful, but expensive, and is used in the high-end market.
Comparison between air-cooled and water-cooled chip cooling methods
| Cooling Method | Cooling Principle | Advantages | Disadvantages |
| 空气散热器Air-Heatsink | Heat generated from CPU/GPU operation is transferred to the heat sink and heat transfer with surrounding air under the action of the fan | Low price, high security | Noisy, large size, limited heat dissipation efficiency, dependent on chassis ventilation |
| Water Heatsink | CPU heat is absorbed through the liquid in the pipeline under the action of water pump, cooled and circulated repeatedly. | CPU heat is absorbed through the liquid in the pipeline under the action of water pump, cooled and circulated repeatedly. | High price, difficult to repair fully enclosed, risk of water leakage |
Water-cooled Heatsink module mainly contains cold plate, water pipe, fan, cold head and other components, divided into one-piece water-cooling and split water-cooling, both of them work on the same principle, but there are differences in the way the components are assembled. Take the integrated water cooling as an example, the cold head contains a pump inside, when working, the cold head side is in direct contact with the CPU surface, and the other side adopts the CNC process to knock out a large number of grooves (micro-channels), the cold water flows through the micro-channels are heated by the CPU heat, through the pump to drive the water flow, flow through the water pipe to enter the cold row, the cold row has a lot of internal waterways, there are a large number of fins inlaid in between the waterways, the heat is transferred to the fins through the cold row, the upper fan dissipates the heat, cooling down the cold row, the water pipe, the fan and so on.
The heat is transferred to the fins and then dissipated by the fan on top of the cold platoon, and the cooled cold water flows back again, and the cold liquid circulated by the water pump takes away the heat absorbed from the core of the cold head. The size of the cold plate greatly affects the cooling efficiency, the common sizes in the market are 120mm, 240mm and 360mm.
At present, CPU cold head and graphics card cold head are mainly applied to the copper bottom heat conduction, there are two main CPU water-cooling cold head designs on the market, one is the ordinary copper column type, the other is the jet type, the jet type cold head in the copper column based on the ability to water through the narrow nozzle quickly spray to the bottom of the copper plate, to enhance the local flow rate and the formation of turbulence, so that the water-cooled liquid heat-absorbing efficiency is greatly improved, the water resistance of the jet type cold head is greater, the water pump The water resistance of jet cooler is higher, which requires higher head of water pump. Graphics card cold head is divided into single-core, half-coverage and full-coverage cold head.
The working principle of the air cooling module is that a large amount of heat generated by the CPU/GPU during operation is transferred to the heat sink, causing the heat sink to heat up. Under the action of the fan, heat is transferred between the heat sink and the surrounding air, and the temperature reaches an equilibrium and stable state when the heat emitted by the heat sink is equal to the heat generated by the CPU during its maximum power consumption. The air cooling module consists of three main components:
1. Heat Pipes or Vapor Chamber;
2. Heat dissipation fins Fins;
3. Cooling fan or turbine;
These three parts are often coupled with thermal conductive materials and processes such as soldering to accelerate heat transfer. There are two main ways of contacting the heatpipes with the fins, one is "through-fin", where the heatpipes are inserted directly into the fins, and the other is metal soldering.
Heat pipe and CPU / GPU core contact method there are mainly two kinds, one is direct contact, directly to the heat pipe processing to the approximate shape of the bottom surface of the grinding and pasted on the core, this processing method is low-cost, but prone to deformation, mainly used for low-end; high-end copper bottom is more, the processing of the heat pipe interspersed with the welding of the heat pipe in a copper block, or with a groove of the copper will be sandwiched in the middle of the heat pipe, the heat is absorbed by the copper bottom before transferring to the heat pipe, and then transferred to the heat pipe. Heat is first absorbed by the copper bottom and then transferred to the heat pipe, this way of long service life, uniform heat transfer, higher cost.
