To avoid constant repeating of the same texts, here is a summary of all the information about how I evaluate the power supplies. This text will be updated if there are changes in methodology or instruments used. Actual article version: 1.33. This evaluation methodology is most likely one of the most complicated ever, on the other hand it should be able to reflect wide spectrum of parameters you can find in power supplies. For how we actually test the power supplies, refer to the other article.
History
1.0: Corsair VS450; pilot evaluation
1.1: Gold bar (zlatá cihla); lowering requirements for −12V rail
1.2: increasing the penalty for not meeting efficiency demands and increasing the penalty/reward for high-end segment units ripple
1.21: adding the penalties/gains for (non)functional protections and for burning during testing
1.22: adding the penalties/gains for (non)presence of saver-technologies and for wire thickness
1.25: modification of price categories due to tragic development of currency exchange
1.3: Cooler Master V Semi Modular V550S: measuring hold-up time which replaces the price in evaluation; instead, a second number (value/price ratio) is introduced
1.31: Zalman ZM600-GVM: measuring both voltage hold-up time and Power Good signal HUT
1.32: Antec High Current Gamer M: longer warranty expectations, smaller gain for quality capacitors in mainstream
1.33: SilentiumPC Enduro FM1 550 W: tweaking and better explanation of the build quality requirements
1.4: FSP200-GSV50-5K (85): different SFX price segment division
The basics remain at all times: the power supply must deliver rated power and must deliver the power within ATX specification tolerance to even be evaluated, otherwise it is not and will get the FAILED sticker. But the numerical evaluation itself is not self-redemptive, if the unit falls behind in several areas and excels in others, it may still get overall good evaluation even if I would personally not recommend such unit.
Therefore it is vital to take the numerical result into account together with the verbal evaluation in the final thoughts section. It is also possible that the ATX specification itself will change through time with technological progress, I personally expect the −12 V rail to drop out of the standard in time. For the time being, all the expectations are set up so high that really good power supplies will get high score, everything else that falls behind in some way will get less.
I postulate three categories according to price per watt ratio of ATX units:
- low-end: up to 1.99 CZK/W
- mainstream: 2–3.99 CZK/W
- high-end: 4→ CZK/W
SFX units have slightly shifted categories because of added costs:
- low-end: up to 2.99 CZK/W
- mainstream: 3–4.99 CZK/W
- high-end: 5→ CZK/W
There is a different set of criteria for each category. That means you cannot directly compare power supplies from different categories as there are different criteria for them! Average values of these criteria are pre-set, for which the unit gets basic amount of points. If it differs to one or other side (worse/better), it loses/gains extra points. All these groups are then summarized and the result is our final score value. One of the basics of this system is, that there are no absolute, un-crossable barriers like minimum or maximum. That means, it can happen that some units get better score than apparent maximum (100 %), or worse than apparent minimum (0 %).
The pre-set criteria are:
- component/technology quality (0–15)
- build quality (0–15)
- voltage regulation under combined load (0–10) and crossload (0–5, rated half)
- ripple under combined load (0–10) and crossload (0–5, rated half)
- efficiency (0–15)
- hold-up time (0–15)
- others, e. g. some special functions, accessories, longer warranty etc. (0–10)
Some of the expected values are specified under this paragraph, the basic amount of points for each set is in the brackets. If a unit meets the requirements exactly, it gets a value score of 60 %. So you can compare the actual value with different models even years after publishing the review when the price is completely different, there is second number introduced with version 1.3, the value per price ratio. It defined as vppr=100×value/actual_price (CZK). You may change the multiplier yourself depending on the value of your selected currency, I use 100× for CZK so we do not work with small decimal numbers.
I have been long thinking about letting pass or not letting pass units which do not meet ATX minimum hold-up time. Since it is not a thing which may immediately harm your computer and you can solve that for example with an UPS, I set the absolute minimum to 10 ms, where the shorter one of both hold-up times (voltage and power-good signal) is taken as the evaluation basis. If any unit fails to pass even that, it surely does not deserve to pass as that is usually the transfer time for many UPS. So even UPS will most likely not help you with such wonder. If it makes more than 10 ms but less than 16 ms (17 ms for PG signal), it shall be penalised for that fact itself and also per each ms under 16 ms (17 ms). Different market categories then have different segments for the minimum time to gain or lose points.
Low-end
- components must be adequately chosen (minimum for rated power); input filtration in minimum configuration must be present; output filtration coils and capacitors must be present; 18AWG wires at least for main connectors, 20AWG for the rest; over-power protection (10)
- soldering can contain small flaws, manual repairs are allowed without restrictions; electrical safety must be at least bare minimal; maximum 5 tiny solder balls allowed (and not easily movable) (10)
- voltage regulation up to maximum allowed by ATX (10)
- ripple up to maximum allowed by ATX (10)
- average efficiency at least 75 % (10)
- hold-up time 14 ms (10)
- nothing (0)
Mainstream
- components must be adequately chosen (approx. 30+ % overbuilt); input filtration in minimum configuration must be present; output filtration coils and capacitors must be present; quality capacitors at least on effortful +5 V SB rail; 18AWG wires for all connectors; over-power, under-voltage and over-voltage protections (10)
- soldering can contain small flaws, good cleaning, good electrode trimming; two manual repairs are allowed (properly cleaned); electrical safety must be at least average; maximum 2 tiny solder balls allowed (and not easily movable) (10)
- voltage regulation up to 3 % difference (−12 V up to 6 %) (10)
- ripple up to 60 % of the ATX specification interval (10)
- efficiency set by 80+ Bronze (@115 V) certificate (average is then taken for the evaluation) (10)
- hold-up time ATX minimum (16/17 ms) (10)
- sleeved cables for main connectors (Main ATX, ATX 12 V, PCIe), at least 3-year warranty (0)
High-end
- components must be adequately chosen (approx. 80+ % overbuilt); input filtration in minimum configuration must be present with at least one extra energy-saver technology (X-capacitor discharge IC, thermistor bypass etc.); output filtration coils and capacitors must be present; only quality capacitors allowed; 16AWG wires at least for main connectors, 18AWG for the rest; over-power (over-current), under-voltage and over-voltage protections; gold-plating at least on main connectors (10)
- soldering, cleaning and electrode-trimming must be flawless, only one manual repair is allowed (which must be done well and cleaned after the process); electrical safety must be exceptional (drilling and creepage slots between primary and secondary, extra component and leg insulation etc.); no solder ball allowed (10)
- voltage regulation up to 2 % difference (−12 V up to 4 %) (10)
- ripple up to 40 % of the ATX specification interval (10)
- efficiency set by 80+ Gold (@115 V) certificate (average is then taken for the evaluation) (10)
- hold-up time 18 ms (10)
- all cables sleeved, at least 4-year warranty (0)
Here are also some examples of the point evaluation, some areas are then more individual rather than exactly pre-set:
| Low-end | Mainstream | High-end | |
| Efficiency | ±1 p/1 % | −1 p/not passing, ±1 p/1 % | −2 p/not passing, ±1 p/1 % |
| Hold-up time | ±1 p/ms | −1 p/not passing, ±1 p/ms | −2 p/not passing, ±1 p/ms |
| Voltage regulation (−12 V: −half) | +1 p/0.4 % | ±1 p/0.2 % | ±1 p/0.1 % |
| Ripple (−12 V: −half) | +2 p/10 % | ±1 p/10 % | ±2 p/10 % |
| Non/quality capacitors | +1 p/piece | ±0.5 p/piece | −1 p/piece |
| Warranty | +3 p/year | ±2 p/year | ±1 p/year |
| Saver technology | +3 p/piece | +2 p/piece | ±1 p/piece |
| Solder balls | −1 p/3 pieces | −1 p/2 pieces | −1 p/piece |
| Sleeving | +2 p/connector | ±1 p/connector | −1 p/connector |
| Unplugging clips on connectors | +2 p/connector | +1 p/connector | +0.5 p/connector |
| Thinner/thicker wires | ±2 p/connector | ±1 p/connector | +0.5 p/−1 p per connector |
| Gold-plating | +2 p/connector | +1 p/connector | +0.5 p/−1 p per connector |
| Modular cabling | +15 p | +10 p | +5 p |
| Semi-fanless operation | +10 p | +5 p | +3 p |
| Burned | — | −5 p | −10 p |
| (Non)working protections | +7 p | ±5 p | −5/+3 p |
Discuss the article in our forums.
Loading ...