The test
Our testing procedures can be found at this link.
Efficiency
| Load | 10pc | 25pc | 50pc | 75pc | 100pc |
|---|---|---|---|---|---|
| Efficiency | 82.1pc | 90.1pc | 91.3pc | 90pc | 89.6pc |
Efficiency is very good from 25 per cent to 100 per cent load. Our numbers show that the test HX850 manages to attain the required 90 per cent at 50 per cent load; it even manages it at 25 per cent and 75 per cent, too.
Regulation
In terms of regulation, we're looking at just how well the supply is able to hold to the various lines. The ATX spec. has a +/- 5 per cent leeway on all but the -12V line.
| Line/Load | 3.3V | 5V | 12V |
|---|---|---|---|
| 10 per cent | +1.0pc | +1.3pc | +1.1pc |
| 50 per cent | +0.2pc | +0.5pc | +0.5pc |
| 100 per cent | -1.1pc | -1.0pc | -0.5pc |
Most supplies over-volt with little load and under-volt when stressed. The HX850 does an above-average job of keeping the voltages in check when run at 100 per cent. The regulation is tighter - read better - than the similar GS800.
Regulation - cross-load
How about providing uneven loads that stress particular voltage rails? In the first attempt, we've put 65A on the 12V rails, and 1A on the 3.3V and 5V rails. This can actually be somewhat typical for a system heavy on graphics and CPU power. In the second, we've turned the tables and gone for 12A on both the 3.3V and 5V rails - highly unlikely in a real-world environment - and just 2A on the 12V - even more unlikely!
| Line/Load | 3.3V | 5V | 12V |
|---|---|---|---|
| Cross-load 12V focus | +1.2pc | +1.3pc | -0.5pc |
| Cross-load 3.3V/5V focus | -2.3pc | -1.4pc | +1.1pc |
Hammering one part of the PSU power delivery while using just a small portion of the other can throw cheaper supplies of out kilter. There's not a huge amount of variation going on here; you're looking at less than four per cent from a best-to-worst-case scenario.
Ripple
| Line/Load (mv - p-p max) | 3.3V | 5V | 12V |
|---|---|---|---|
| 10 per cent | 15mV | 15mV | 15mV |
| 50 per cent | 18mV | 20mV | 20mV |
| 100 per cent | 20mV | 20mV | 35mV |
The ATX v2.2 spec states that the maximum permissible ripple is 120mV for the 12V line and 50mV for others.
PSUs convert AC power into DC, but doing so requires the AC waveform to be suppressed. What we're really testing here is the quality of the supply's rectifier and any smoothing capacitors in getting rid of this unwanted up-and-down ripple.
Per-line ripple is less than half of what the ATX specification permits. The 12V ripple at full load is usually the worst figure, with AC suppression being hardest to control under this load, but the HX850's 35mV is very good for a high-quality supply.
Temps
| Temperatures | Intake | Exhaust |
|---|---|---|
| 10 per cent | 28°C | 44°C |
| 50 per cent | 32°C | 38°C |
| 100 per cent | 37°C | 44°C |
The fan isn't activated until the load hits 20 per cent and this is why the PSU is actually cooler at 50 per cent load than at 10 per cent. The supply was tested with an ambient temperature of 30°C, which is around 5°C higher than most other supplies we have looked at in the preceding four months.
Noise
As advertised, the supply is silent at low loads. However, the auto-switching mechanism isn't exact. We loaded the supply with 425W for five minutes and then reduced this to 100W, a figure where the fan should be switched off. The unit took approximately 90 seconds to determine it was safe to switch the fan off completely. Likewise, running from 100W up to 425W doesn't initiate an instant-on fan; the supply takes a while before kicking the 14cm fan into action.
Providing a little more context, it does remain switched off when most PCs are idling for extended periods. The transistor that controls RPM gradually increases fan speed up to around 70 per cent PSU load, performing much in the same vein as the reviewed GS800. After this level the fan noise is noticeable at all times, so do be aware of this if you plan to pull over 600W on a regular basis, though, of course, internal fan noise - graphics, CPU, fans, etc. - generally drowns out the PSU's.
