Originally posted by: Ginta
JPSJPS, since you seem well informed, I'm curious to your opinion of the testing methodology used for power supply reviews at
Silentpcreview.
For a design engineer, being blunt is usually a necessity. Opinions/guesses/ do not matter. However, diplomacy sometimes wrongly gets ignored so I apologize if I have offended anyone. In the course of a circuit design, the design engineer is critiqued by his peers (and critiques his peers' designs also) in several formal design reviews. This process is mandatory to find design errors/mistakes/shortcomings. You have to learn to deal with the facts without being personally insulted.
Anandtech is a HUGE commercial site and should not even have published a pathetic joke review like this!
On the other hand, silentpcreview is more of a hobbyist site and does not have the financial means to do a thorough report. In fact, silent mentioned that they would have liked to be able to vary the input mains 115/230 ACV input (simulating transient conditions too) but did not have the equipment. This shows that they recognize the importance of this test. Anandtech did NOT.
Following is a appropriate quote from the techreport.com comments section concerning their report :
"It's a real shame when kids write hardware reviews without knowing what they're doing. The lemmings rush out and buy THE WINNER and mfr's decide not to bother supporting the enthusiast community with review samples because the results are pure bullshit - why should they give out free samples when the results are a crap shoot?"
In fact, none of the on-line reviews that I have seen (including Uncle Tom's) produce any meaningful, conclusive, valid quality/noise/performance/evaluation test data.
This is because too many absolutely necessary factors/tests/analysis have been omitted that are required for a professionally done test. This is kind of like telling a story but omitting a selected detail which would change the whole meaning of the story.
One example: No analysis was done to determine the maximum heat sink temperature for a desired MTBF. To calculate this, we need to know the junction temperature specs for the high power semiconductor devices and the power dissipation of each. Then, we need to know the thermal resistance from junction to heat sink to calculate junction temperature. Alternatively, we could measure the case temperatures & dissipation of the active devices and do the calculations. Since these numbers will be different for different circuit designs, measured heat sink temperature data is meaningless since we do not know required/desired maximum heat sink temperatures. Thus, the measured cooling fan noise levels are also meaningless since we have no CFM requirements which would directly influence noise levels.
Other examples: Dynamic control of input AC Voltages & DC output currents over the full temperature operating range vs output voltage levels/transients, and ripple are not tested.
Also, no description of over voltage or over current protection design details.
Also, the importance of the PS "side fans" is not stressed nearly enough! The most important factor for convection cooling a heat sink is the velocity of the air and the turbulence it creates. This increases the Reynolds number which increases the convection coefficient. A fan blowing directly onto the heat sink will provide MUCH better/efficient cooling than the same CFM fan mounted as an exit air fan. Of course the heat sink design has to be optimized for airflow from this side direction.
A few of external factors are very important also:
1) Most folks have no idea of the amount of power (or current draw for each individual voltage) required by their system. DC clamp on ammeters are available for about $100 now and can be used to measure this. What total power is typically required for a system with a 9800 video card, several HDs and several CDR/DVDRs & 3GHz CPU?? If you do the measurements, you will see that it takes a stomping system to go much above 250W. Current draws from each individual voltage are generally more important than total power for comparison to PS capabilities.
2) How is the case cooled? Do we use the PS to cool the case or do we use several quiet low RPM case fans to provide a lot of cool case air? Obviously the PS requirements are COMPLETELY different for these two conditions since PS input air temperatures will be MUCH different.
3) Have we calculated case cooling air CFM requirements to minimize the PS air input temperature?
Simplified Equation at room temp and sea level:
Air Temp Rise (Deg C) = 1.7 times Watts/CFM
A couple of 80 mm L1As (24CFM) and 200W will produce about 7 Deg C air temp rise into PS. Thus, we may desire to mount one of the fans to blow air directly toward the PS input and CPU fan as well.
Design engineers hate to write test procedures, but most have to do it anyhow. For a PS test like this, several weeks would be required to do the analysis, write the procedure, and develop a required test equipment list. So, it may be too much to expect valid data obtained on a tight test budget.
If you have read this far, you have way too much stamina. Buy a 550 Watt PS and turn the fan speed down and live with it.