Owing to the significant progress in technologies that ensure reasonable temperature conditions for modern generations of processors, the primary characteristic of coolers (efficiency or cooling capacity) is slowly but surely losing its relevancy, while secondary parameters (noise, ergonomics, technical quality, etc) are becoming of paramount importance. This trend is clearly reflected in the segment of premium coolers, where marketing subcurrents oblige such products to combine technical parameter with practical results. The old sacramental question sounded like how efficient this or that cooler was, but now it's a little different, "How efficient is this cooler at minimum noise and providing decent usability?"
This transformation requires a more orderly approach to inspecting cooling systems, of course. Critical functional aspects should be highlighted. From this angle, according to the new methodological regulations, noise generated by coolers becomes a priority in our tests. Their comparative analysis will be done in 'functional domains' graded by noise.
We introduce six functional domains:
- Noiseless. A given cooling system does not generate noise, that is it's passive -- no fans or pumps, etc.
- Conditionally noiseless. Noise from a given cooling system, measured by the established procedure, is below 24dBA. This parameter is below the typical background noise in a quiet room (in the evening or at night). So the cooler does not contribute much to the noise level even of PCs with passive PSUs, noise-reducing enclosures for hard drives, passive GPU coolers, etc.
- Low-noise. Noise level of a given cooling system lies within 24-30dBA. That is the cooler does not contribute much to 'acoustics' of quiet PC configurations with low-noise PSUs, low-noise GPU coolers, low-noise hard drives, low-noise case fans, etc.
- Ergonomic. Noise level of a given cooling system lies within 31-36dBA. That is the cooler does not contribute much to 'acoustics' of typical Mid- or Hi-End computer configurations, but it may be described as loud in low-noise PCs.
- Conditionally ergonomic. Noise level of a given cooling system lies within 37-42dBA. That is, you will notice noise from the cooler in most computers, except for those using non-ergonomic components (10000rpm or faster hard drives, power supply units with high-speed fans, noisy GPU coolers, etc).
- Non-ergonomic. Noise from a given cooling system exceeds 42dBA. A cooler in these conditions will be the main generator of noise in practically any computer configuration. Home usage of such a cooler is not justified (it's limited to workplaces and offices with the background noise above 45dBA).
So, the main domains in our practice are 2-5. Cooling efficiency is compared at checkpoints using the most typical and very close noise characteristics (different by less 1dBA). For domains 2-4, checkpoints are chosen closer to the lower border, and for Domain 5 -- lying in the middle or closer to the upper border of the specified noise range. Noise is measured using this procedure adapted from GOST 12.1.026-80 and ISO 3744 (in Russian).
What concerns measuring thermal parameters of cooling systems, nothing has been changed: we have a number of tests to evaluate cooling capacity, they are performed on a special Intel LGA775 testbed.
- Motherboard: ASUS P5AD2-E Premium rev. 1.05
- Processor: Intel Pentium 4 550 (3.4GHz Prescott, HT Technology)
- OS: Microsoft Windows XP
The primary data used for the consequent calculation of thermal resistance are temperature readings of the thermal diode built into a processor. Besides, we take temperature readings of chokes in the CPU voltage regulator, as additional parameters that help evaluate 'collateral' cooling of near-socket components (temperature of chokes PL24, PL25, and PL26, which are installed close to the socket). We use S&M utility to generate more heat in the testbed. CPU voltage is raised to 1.525V (the resulting heat power is 150W.)
The testbed is installed in a semi-open test box with quasi-isothermal environment inside (25°C). A cooling system with the reference thermal interface Noctua NT-H1 and a CPU is heated with three test runs for at least 60 minutes, temperature readings are registered with SpeedFan utility and then processed for statistics.
The resulting temperature data are used to determine thermal resistance:
θja = (Tj - Ta)/Ph
Where Tj is CPU core temperature, Ta is the environment temperature (25°C in our case), Ph is the thermal capacity of a processor (150W in this case).
We also calculate an additional integral parameter -- efficiency-noise ratio (ENR):
ENR = DM*(Rt/TC)/(NL/Rn)
Where Rt is the reference temperature (the reference thermal resistance θja of the cooling system - 0.25°C/W); TC is the core temperature with the operating cooling system; Rn is the reference noise (the reference noise level is 20dBA); NL is the noise level, generated by the cooling system; DM is the denominate multiplier (10).
ENR is used to evaluate optimizations of the thermal design for ergonomic noise: the higher this ratio, the higher cooling capacity is demonstrated by a given cooling system at minimal noise.
That's about all. This approach is not the ultimate truth, of course. Nevertheless, it provides a crystal-clear comparison of cooling systems, relying on identical noise parameters in different acoustic conditions. We hope that our new form of representing test results will help you examine premium coolers and make the right choice for your preferences and requirements. As always, constructive criticism is welcome.